- Permanent Link:
- A Unified Human Interaction-Based Theory and Framework for Simulation Modeling and Visualization Design
- Ezzell, Zachary Peter
- Place of Publication:
- [Gainesville, Fla.]
- University of Florida
- Publication Date:
- Physical Description:
- 1 online resource (152 p.)
- Doctorate ( Ph.D.)
- Degree Grantor:
- University of Florida
- Degree Disciplines:
- Computer Engineering
Computer and Information Science and Engineering
- Committee Chair:
- Fishwick, Paul A
- Committee Members:
- Ho, Jeffrey
Dankel, Douglas D
Lok, Benjamin C
- Graduation Date:
- Subjects / Keywords:
- Animation ( jstor )
Domain ontologies ( jstor )
Graphic design ( jstor )
Graphics ( jstor )
Mathematical variables ( jstor )
Modeling ( jstor )
Ontology ( jstor )
Semantic models ( jstor )
Simulations ( jstor )
Three dimensional modeling ( jstor )
Computer and Information Science and Engineering -- Dissertations, Academic -- UF
design -- modeling -- simulation -- visualization
City of Crystal River ( local )
- bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Computer Engineering thesis, Ph.D.
- Visualizing dynamic phenomena by leveraging computer graphics and simulation is of increasing interest to society. In the classroom, such dynamic visual modules can motivate students and give them a holistic view of a given event. Industry practitioners and researchers also leverage such modules to aid in comprehending simulation results and communicating ideas. Current methods used to construct these customized views,however, are expensive. Content experts must employ engineers to code the mathematical model that defines the dynamic behavior of the scenario. Engineers must then somehow visualize the output of the model. These tasks are performed with engineering software tools or just pure computer programming. This work is towards defining an interface and interaction model that will compress this engineering overhead, and thus narrow the user-interface gap between mathematical modeling and 3D interactive visualization design and consumption. Interaction within the proposed interface centers on a visualization of a 3D semantic network, or ontology, in which domain concepts are represented by nodes and an edge between two nodes represents a semantic relationship between two concepts in the domain space. To create custom modules, visual attributes can be added to nodes within the semantic graph that point to graphical resources, or define dynamic behavior. In this dissertation it is demonstrated that through interactions with this semantic graph, a designer can sculpt and annotate simulation models into meaningful interactive, animated visualizations that integrate 2D and 3D information to relay the dynamics of an event to an observer. The theoretical interface and interaction techniques will be presented along with a case study visualization constructed using a prototypical implementation of the theory. ( en )
- General Note:
- In the series University of Florida Digital Collections.
- General Note:
- Includes vita.
- Includes bibliographical references.
- Source of Description:
- Description based on online resource; title from PDF title page.
- Source of Description:
- This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
- Thesis (Ph.D.)--University of Florida, 2012.
- Adviser: Fishwick, Paul A.
- Statement of Responsibility:
- by Zachary Peter Ezzell.
- Source Institution:
- Rights Management:
- Copyright Ezzell, Zachary Peter. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
- LD1780 2012 ( lcc )
This item has the following downloads:
1 A UNIFIED HUMAN INTERACTION BASED THEORY AND FRAMEWORK FOR SIMULATION MODELING AND VISUALIZATION DESIGN By ZACH EZZELL A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2012
2 2012 Zach Ezzell
3 To Mom and Dad
4 ACKNOWLEDGMENTS I thank my research advisor and supervisory committee chair, Dr. Paul Fishwick for teaching me how to conduct research and giving me the freedom and time to truly explore ideas during our numerous projects I would not have purs u ed a Ph.D. if not for my experience working with Dr. Fishwick as an undergraduate I thank my supervisory committee members, Dr. Douglas Dankel, Dr. Jeffrey Ho, Dr. Samsun Lampotang and Dr. Benjamin Lok for their time, ideas and support. I also thank my collaborators at the University of Central Florida, Dr. Juan Cendan and Dr. Jonathan Kibble for their expert feedback from a medical education perspective. I thank my Mother, Father, and two sisters for their love and support. My family instilled me with an intellectual curiosity at a young age and has always provided much needed emotional suppo rt, especially during the ups and downs of graduate school. I would also like to thank my loving girlfriend, Mo lly for her endless support and kindness. Finally, I would like express further gratitude to wards Molly and my F ather for helping me polish thi s document.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 LIST OF OBJECTS ................................ ................................ ................................ ....... 1 2 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 15 ................................ ................................ ................................ ........ 17 1.1.1 Ontological Models ................................ ................................ .................. 17 1.1.2 Dynamic Models ................................ ................................ ...................... 18 1.1.3 Graphical Models ................................ ................................ ..................... 19 1.2 Thesis Statement ................................ ................................ .............................. 19 1.3 Contributions ................................ ................................ ................................ ..... 20 1.4 Broader Impact ................................ ................................ ................................ 21 2 BACKGROUND ................................ ................................ ................................ ...... 30 2.1 Ontologies ................................ ................................ ................................ ......... 30 2.1.1 O ntologies: A Wide View ................................ ................................ ......... 30 2.1.2 Ontology Visualization ................................ ................................ ............. 31 2.1.3 Ontologies in Simulation ................................ ................................ .......... 33 2.2 Visualization Frame works ................................ ................................ ................. 40 2.3 Integrative Multimodeling ................................ ................................ .................. 44 2.3.1 An Integrative Model Blending Environment ................................ ........... 45 2.3.2 The Augmented Anesthesia Machine ................................ ...................... 45 2.4 Simulation and Visualization Design in Practice ................................ ............... 46 2.4.1 Graphics Programming ................................ ................................ ............ 47 2.4.2 Engineering Simulation Tools with Visualization Add ons ....................... 47 2.4.3 Game Engines and 3D Modeling Packages ................................ ............ 48 2.5 Summa ry ................................ ................................ ................................ .......... 49
6 3 ONTOLOGY CENTERED INTERACTION THEORY AND FRAMEWORK ............ 54 3.1 The Base Structure ................................ ................................ ........................... 55 3.1.1 Concepts ................................ ................................ ................................ 55 3.1.2 Attributes ................................ ................................ ................................ 55 3.1.3 Rel ationships ................................ ................................ ........................... 56 3.2 Structural and Semantic Affordances ................................ ................................ 57 3.2.1 Meta attributes ................................ ................................ ........................ 57 3.2.2 Attributes as Simulation Equations ................................ .......................... 59 3.2.3 Taxonom ies for Object Creation ................................ .............................. 60 3.2.4 Creating Transform Hierarchies and Vertex Groups ................................ 60 3.2.5 Influence for Ontology Based Animation ................................ ................. 62 3.2.6 Designed Ontology Representations ................................ ....................... 63 3.2.7 Ontolog y Guided Particle Systems ................................ .......................... 64 3.2.8 Attribute Expansion ................................ ................................ ................. 65 3.2.9 Ontology Based Interaction Modalities ................................ .................... 66 3.2.10 Semantic Styling ................................ ................................ .................... 67 3.2.11 Ontology Portability through RDF ................................ .......................... 68 3.3 Technical Requirements ................................ ................................ ................... 69 3.4 Methodology ................................ ................................ ................................ ..... 71 3.4.1 Ontology Acquisition ................................ ................................ ................ 71 3.4.2 Simulation Model Building and Visualization Construction ...................... 72 3.4.3 Simulation Execution and Visualization Animation ................................ .. 73 3.5 Classification, Comparisons and Limitations ................................ ..................... 73 3.5.1 Classification ................................ ................................ ........................... 73 3.5.2 Comparisons ................................ ................................ ........................... 74 3.5.4 Limitations ................................ ................................ ............................... 75 3.6 Summary ................................ ................................ ................................ .......... 77 4 C ASE STUDY ................................ ................................ ................................ ......... 88 4.1 Cardiovascular Modeling ................................ ................................ .................. 88 4.1.1 The Beneken Model ................................ ................................ ................ 89 4.1.2 The Goodwin et al. and S Couto et al. Models ................................ ...... 90 4.2 Case Study ................................ ................................ ................................ ....... 91 4.2.1 Software Prototype ................................ ................................ .................. 91 4.2.2 Solving the Compartmental Model ................................ .......................... 92 4.2.3. Construc ting the Executable Simulation and Visualization ..................... 93 188.8.131.52 Constructing the base model ................................ ......................... 94 184.108.40.206 Adding dynamic visualization ................................ ......................... 96 220.127.116.11 Sculpting relationships ................................ ................................ ... 98 18.104.22.168 Creating part icle systems ................................ ............................... 99 22.214.171.124 Defining the viewer module ................................ .......................... 100 4.2.4 Extensions for Hypovolemic Shock ................................ ....................... 101 4.3 Preliminary Human Considerations ................................ ................................ 102 4.3.1 Student Survey ................................ ................................ ...................... 103 4.3.2 Expert Survey ................................ ................................ ........................ 105
7 4.4 Summary ................................ ................................ ................................ ........ 107 5 CONCLUSIONS ................................ ................................ ................................ ... 123 APPENDIX A EXAMPLE LESSON ................................ ................................ ............................. 128 B EXPERT SURVEY RESULTS ................................ ................................ .............. 137 LIST OF REFERENCES ................................ ................................ ............................. 140 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 152
8 LIST OF TABLES Table page 3 1 A list of suggested attributes for assigning the influence between simulation variables and visualization parameters ................................ ............................... 79 3 2 A list of suggested attributes for assi gning particle systems properties to ontology concepts and relationships. ................................ ................................ .. 80 3 3 The characteristics of ontologies well suited for use in the proposed framework. ................................ ................................ ................................ .......... 80 4 1 The semantics required to recreate the 10 compartment Beneken model of the human cardiov ascular system. ................................ ................................ ... 109 4 2 I nfluence attributes added to the Right Atrium concept to drive the animation of the right atrium portion of the heart mesh. ................................ .................... 109 4 3 Particle system attributes added to flows to relationships. ............................... 110 4 4 The effects of h ypovolemic shock as illust rated by Lawrence et al. (2006) ...... 110
9 LIST OF FIGURES Figure page 1 1 simulation variables to visualization parameters. ................................ ................ 27 1 2 The ontological spectrum ................................ ................................ ................... 27 1 3 A portion of the foundational model of human anatomy ontology ....................... 28 1 4 A dynamic compartmental model for cardiovascular physiology ........................ 28 1 5 Different representations of graphical models. ................................ ................... 29 2 1 Protg (Noy et al., 2001) displaying an ontology for a newspaper. ................... 50 2 2 A partial view of DeMO (Miller et al., 2004). ................................ ....................... 50 2 3 A high level view of the Semantic Web enabled simulation design .................... 51 2 4 Transforming a SE DRIS file into an OWL ontology ................................ ............ 51 2 5 High level depiction of the translator written by Lacy (2006) to demonstrate PIMODES. ................................ ................................ ................................ .......... 52 2 6 The ontology engineering, storage and search framework cre ated by Bell et al. (2008). ................................ ................................ ................................ ........... 52 2 7 An example of transforming and rendering 3D objects based on simulation variables in Ptolemy ................................ ................................ .......................... 53 3 1 A concept Bunny with graphical attributes of a 3D m esh and a 2D image (xray.jpg). ................................ ................................ ................................ ........... 81 3 2 A high level sketch of a user interface to meta attributes. ................................ 82 3 3 Two different syntax approaches for accessing attributes across multiple concepts within equations. ................................ ................................ ................. 82 3 4 A simple 3D scene with an ontology visualization serving as an in terface to mesh transformations ................................ ................................ ......................... 83 3 5 A simple illustration of using a has a relationship to form vertex groups. ........... 83 3 6 A depiction of a particle traversing a r elationship curve. ................................ ..... 84 3 7 The action of expanding an attribute. ................................ ................................ 84
10 3 8 A depiction ................................ ........................... 85 3 9 A simple example of applying a gr aph style sheet to an ontology to create a graphical representation used in systems dy namics modeling. .......................... 85 3 10 Different approaches to morphing the ontology structure used in the prototype into th ................................ ...................... 86 3 11 A sketch of a potential ontology pruning interface where concepts can be dragged into the 3D design environment from a 2D list based view. .................. 86 3 12 A high level diagram depicting the interfaces between the various required components of the proposed framework. ................................ ........................... 87 4 1 A block diagram model created by Warner (1959) to simulate blood circulation. ................................ ................................ ................................ ........ 111 4 2 The Beneken (1965) compartmental model for blood flow. .............................. 111 4 3 Pathology m odels by S Couto et al. (2006) ................................ .................... 112 4 4 The architecture of the software prototype created to de monstrate the proposed theory ................................ ................................ ................................ 113 4 5 A screen shot of the prototype. ................................ ................................ ......... 113 4 6 A snap shot of building the Beneken model. ................................ .................... 114 4 7 The complete Beneken model created with the software prototype ................. 114 4 8 The Beneken model executing in the software prototype ................................ 115 4 9 The result of adding a Heart concept and assigning it a mesh and material attribute. ................................ ................................ ................................ ........... 115 4 10 A snap shot of the prototype while a designer p ositions the Right Atrium concep t to be within the heart mesh. ................................ ................................ 116 4 11 The influence of the Right Atr ium concept over the hear t mesh ....................... 11 6 4 12 Three snap shots of a designer forming a curve from the flows to relationship between the concepts of Right Ven tri cle and Pulmonary Arterial Tree ............. 117 4 13 The complete Beneken model co located with the heart mesh in 3D. .............. 117 4 14 The frame of a curve ................................ ................................ ........................ 118 4 15 The result of using the rotation minimizing fram e technique. ............................. 118
11 4 16 A 32x32 pixel image, cell.jpg, used in the creation of particles for the case study ................................ ................................ ................................ ................. 119 4 17 A particle system animates between the Extrathoracic Arteries and Extrathoracic Veins concepts. ................................ ................................ .......... 119 4 18 The complete Beneken model co loca ted with the heart mesh in 3D ............... 120 4 19 A snap shot of the designed visualization ren dered in the prototype viewer .... 120 4 20 A snap shot of the Beneken model co located with heart and human body 3D geometry ................................ ................................ ................................ .......... 121 4 21 An illustration of two influences added to the concepts Heart and Human Body within the shock visualization. ................................ ................................ 121 4 22 A simulation and visualization of hypovolemic shock over time ........................ 122 4 23 The viewer prototype with lesson plan that was distributed with the student and expert surveys. ................................ ................................ .......................... 122 5 1 Various engineering processes for constructing simulation based 3D interactive visualizations ................................ ................................ ................... 127
12 LI ST OF OBJECTS Object page 4 1 Simulation and visualization design demonstration video ................................ ... 93
13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy A UNIFIED HUMAN INTERACTION BASED THEORY AND FRAMEWORK FOR SIMULATION MODELING AND VISUALIZATION DESIGN By Zach Ezzell December 2012 Chair: Paul A. Fishwick Major: Computer Engineering Visualizing dyna mic phenomena by leveraging computer graphics and simulation is of increasing interest to society. In the classroom, s uch dynamic visual modules can motivate students and give them a holistic view of a given event. Industry practitioners and researchers also leverage such modules to aid in comprehending simulation results and communicating ideas Current methods used to construct these customized views, however, are expensive. Content experts must employ engineers to code the mathematical model that defines the dynamic behavior of the scenario. Engineers must then somehow visualize the output of the model. These tasks are performed with engineering software tools or just pure computer progra m ming. This work is towards defining an interface and interaction model that will compress this engineering overhead and thus narrow the user interface gap between mathematical modeling and 3D interactive visualization design and consumption. Interaction within the proposed interface centers on a visualization of a 3D semantic network, or ontology, in which domain concepts are represented by nodes and an edge between two nodes represents a semantic relationship between two concepts in the domain space. To create custom modules, visual attributes can be added to
14 nodes within the semantic graph that point to graphical resources, or define dynamic behavior. In this dissertation it is demonstrated that through interactions wi th this semantic graph, a designer can sculpt and annotate simulation models into meaningful interactive, animated visualizations that integrate 2D and 3D information to relay the dynamics of an event to an observer. The theoretical interface and interaction techniques will be presented a long with a case study visualization constructed using a prototypical implementation of the theory.
15 CHAPTER 1 INTRODUCTION Content experts leverage visualizations of simulated scenarios during a variety of aca demic and industrial endeavors For example, consider visualizations of plant growth (Tang et al 2011) and human lung dynamics (Santhanam et a l. 2008). Practitioners create visualization s t o analyze and tune simulation models and communicate ideas to inves ted third parties (e.g., management, students) Visualizations of these types are often based on an underlying mathematical model that conforms to the modeling rules within a paradigm (e.g., the systems dynamics paradigm in ecology the compartmental mode ling paradigm in physiology ). The creation of these customized visualization modules can be quite resource intensive. During creation, the content expert will create an abstract simulation model using the domain modeling practices for which they are familiar. In addition to defining dynamic behavior, the expert will often create a primitive pre visualization or sketch of the envisioned graphics and user interaction. Engineers and computer graphics specialist s will then realize this design with a var iety of tools and approaches. The engineer will need to map the abstract model into an engineering software paradigm (e.g., using Simulink or Lab VIEW ) to create an executable simulation. After the simulation model is made executable, a graphics specialis t will link the simulation output to visualization parameters (e.g., using scripts in a 3D modeling package or a graphics programming API) to create the desired animation when the simulation executes. Someone will also be responsible for encoding the desi red interaction including the ability to tweak simulation parameters in the interface a nd see the results in real time
16 Attempts have been made to compress the visualization construction pipeline through the creation of various simulation and visualizati on frameworks. These frameworks however, are lacking in options for visual output require engineering training or operate across multiple user interfaces that employ different interface metaphors. Fo r example, in Simulink ( 2012 ) an industry leading m odel based simulation platform, to link a 3D visualization with an executable simulation model, a user must (1) encode the simulation model within the Simulink graphical simulation modeling window, (2) build the virtual 3D scene in VRML editor window and (3) use a third connective interface to link visualization parameters to simulation variables. An abstract sketch of this interaction model is shown in Figure 1 1 The motivation behind this work is to define an interaction model that shrink s the human computer interface gap between simulation model building and visualization construction activities while remaining general enough to be applied across different content modeling paradigms. Ideally, such an interface would afford the content expert who is not necessarily an engineer, the ability to sculpt interactive and expressive 3D visualizations with minimal technical training The interaction model proposed is structured around a semantic network, or ontology, which is graph based encoding of k nowledge understan dable by a machine and a human ( Berners Lee et al., 2001 ) It will be demonstrate d how the ontological structure can be leveraged to meet user interface and interaction requirements of both the design and consumption of executable simulation models and their corresponding visualizations. This dissertation will present background in several applicable areas, the theoretical interface model and visualization design methodology and a case study built
17 using a prototypical implementatio n of the theory. The remainder of this chapter will explain the three foundational models that are synthesized in the proposed interaction theory, articula te the research question posed, and discuss the potential broader impact of this work. 1.1 The term model background. The definition of model most applicable to this research is "an economical physical surrogate so that we may employ different notations and metaphors as a way to increase ou r understanding of the phenomen a being modeled" (Fishwick, 1995 ). Three core model types are encountered when constructing dynamic 3D visualizations: ontological, dynamic, and graphical. T o fully understand the interface synthesis this work aims to achieve, it is important these three base model types are well understood. 1.1 .1 Ontological Models The concept of ont ology is phrased by G ru In a given field, an underlying set of semantics is always present. For example, the field of psychology contains the sub field cognitive psychology which contains the paradigm information processing which was first described by the person Donald Broadbent in the year 1958 With proper attention to detail, on tological models can be constructed and used to formalize these concepts and semantic relationships into a format which is understandable by humans and machines (Woods, 1975). The ontological spectrum, shown in Figure 1 2 was defined by Daconta et al. (2 003) to highlight the relationship between the expressive power of a knowledge model and the cost of creating and maintaining it.
18 An ontology can be represented mathematically as a graph with nodes represent ing concepts in a domain and edges represent ing relationships between concepts. Figure 1 3 shows a visualization of a portion of the foundation al model of human anatomy ontology (FMHAO) ( Rosse and Mejino, 2007 ). The portion shown pertains to the human heart, and the concept Human body serves as the r oot of the ontology. The Human B ody has a Cardiovascular S ystem which in turn, has a Heart Notice also the relationships denoting dynamics (i.e., blood circulation) such as has arterial supply An ontology is sometimes referred to as a semantic web a term popularized by Berners Lee et al. in 2001 1.1 .2 Dynamic Models Dynamic models are used to recreate or further understand the dynamics of phenomena. Due to their abstract nature, they can also serve as blueprints for the design of simulation and visualization applications. The compartmental model is a specific type of a dynamic model and will be used in the exploration of this thesis Compartmental models represent systems which can be abstracted into interconnected com partments containing some substance (e.g., fluid). The levels of substance at each compartment are typically governed by a set of differential equations. For instance, Beneken (1965) created a compartmental model for cardiovascular physiology. The model, shown in Figure 1 4 is composed of compartments (e.g., left atrium, pulmonary arteries) which are typically drawn out as circles. Directed edges, drawn as lines with arrows, connect the compartments and represent the flow of blood. This physiological model is explored furt her in Chapter 4.
19 1.1 .3 Graphical Models attributes. Graphical models are used in a variety of applications, from visual effects in film to computer aided design in engineerin g M any tools exist to create and edit this model type, such as the 3D modeling package Blender ( Blender 3D 2010) Graphical models consist of material information (e.g., color, reflectivity) and a geometric model composed of vertices, edges and faces. T he information contained within a graphical model should be sufficient to create a render (i.e., a 2D image created using the model as input) from a given perspective. T hree representations of graphical models are explored in Figure 1 5. 1.2 Thesis Statem ent Thesis statement: A three dimensional visualization of the domain ontology can serve as a central user interface structure to connect simulation modeling and visualization construction activities and allows domain specific semantics to guide the interactive modeling process. I believe exploring the interface defined in this thesis will ultimately result in new interaction techniques that will allow experts to leverage their preexisting domain modeling knowledge to create interactive 3D visualizati ons of dynamic processes The resulting visualizations would synthesize ontological models, dynamic models, and graphical models. These visualization s have many uses, including the analysis of simulation results and education. specific sema modeling semantics used by practitioners in a given domain Modeling semantics include domain specific terminology and the graphical properties of visual s imulation models (Zeigler et al., 2000). To support this thesis statement, an o ntology based
20 interaction theory will be defined and this theory will be situated within a framework that defines a complete simulation modeling and visualization construction tool. 1.3 Contributions This dissertation provides two primary contributions to the field of m odeling and s imulation Both contributions pertain to the visualization of simulation model output Contribution 1: The definition of a domain ontology based interaction model to connect simulation modeling to visualization construction activities Current simulation modeling tools allow users to build visualizations controlled to varying degrees by simulation output. However, vi sualization design features in these tools are tacked on in separate interface mod alities and require custom coding for any advanced visual output ( e.g., mesh deformations, particle systems). There is current work on creating advanced to focused within the InfoVi s community and is based on data, not simulation models and often does not utilize 3D graphics The ontology based interaction model and framework presented this dissertation : (1) serve s as a first step towards unifying simulation modeling an d 3D visuali zation construction at the graphical user interface level and (2) leverage s semantics familiar to the arbitrary domain practitioner so that simulation models will not have to be translated to the general engineering modeling paradigm used in current tools and (3) in contrast to current applications of ontologies, demonstrates how the 3D visualized ontologic structure c an serve as a virtual instrument to guide user interactions Contribution 2: The definition of the technical framework required to implement the interaction model. The interaction theory from C ontribution 1 is an essential requirement for integrated simulation and visualization tools, but it is not enough on its
21 own to build the envisioned system The system can only be built after the inter action model is placed within a complete framework. A theoretical framework is described in this dissertation and implementation details surfaced through the coding of a case study inform the requirements of the framework. 1. 4 Broader Impact If the f ramework outlined in this dissertation is employed, the broadest potential benefit is the creation of general user interface that affords simulation and visualization construction and consumption activities to practitioners in various domains and makes the construction process more efficient for engineers that use current methods (see Section 2.4 ). This applies to any domain in which 3D visualizations are leveraged. I believe the particular domain of education, however, could benefit in spe cific ways that are detailed in the remainder of this section 1. 4 .1 Education There are benefit s in using 3D visualizations to help educate someone about a spatial arrangement (e.g., anatomy) or process (e.g., the pumping of oxygen through the body). For studies on how 3D visualizations can be beneficial in an educational setting, see the work by Prinz et al., (2010); Beermann et al., (2010); Quarles et al., ( 2 008); or Siln et al., (2008). Collectively these studies demonstrate the breadth of potential visualization applications in education. There are several educational concepts I would like to explore further and in doing so highlight how the ontology centered framework could provide some foundational technology to support these concepts. The educa tional concepts are systems thinking, constructivism, and scaff olding. A high level description of how the proposed theoretical framework could
22 be leveraged to support these concepts is provided in the following sections along with an overview of the res pective concept 1.4 .1.1 Systems Thinking The abstract models that drive the visualizations may be leveraged as an educational affordance. This is in line with systems thinking, a concept that has experienced a recent proliferation in education Systems thinking is the practice of viewing a system as a set of components and interrelations which influence the becoming part of curricula in fields s uch as ecology (Jrgensen et al ., 2007) and biology (Klipp et al., 2009). In physiology, an area used in a case study for this research, there is a push to emphasize general models when teaching instead of presenting topics as Differe nt ways in which to emphasize models in the physiology class room have been explored. One example is an educational software package created by Rothe and Gersting (2002) which contains a minimal complexity compartmental model for cardiovascular circulatio n. The model drives a simulation and is visualized with the were issues with the usability (some of which have been addressed in revisions), 75% of first year medical students who used the packa ge reported that their understanding of circulation was improved. Another example is a physical respiration model created by Kueble r et al. (2007) for use in a physiology lab course. Of the students that participated in the course, 70% reported the inter active teaching afforded by the model increased their understanding of respiratory mechanics.
23 The proposed ontology centered framework could enable new educational software that falls within the systems thinking paradigm. This could be achieved by crea ting 3D visualizations that are co located with the abstract dynamic model s that drive them In the proposed framework, these dynamic models are encapsulated within the visualized ontology because they are utilized in the design phase and are therefore, readily available to be shown in any resulting visualization module. A n interactive 3D visualization with this characteristic is presented as a case study in Chapter 4 1. 4 .1.2 Constructivism Others have int egrated a model based approach when teaching by employing the constructionism learning paradigm Constructionism is an extension of constructivism, a theory that states we learn by constructing knowledge models. A core principle of constructionism is that the construction of these knowledge models may be aptly facilitated by building tangible things in the real world (Papert and Harel, 1991). Aldridge (2009) applied this theory when teaching a course on the physiology of congenital heart defects by having his students build physical models of defects out of materials such as Play Dough, cardboard and Styrofoam. Students often claimed the model building exercise increased their understanding of the pathology in question. Because one intention of the proposed framework is to support visualization desi gn based more on domain knowledge and less on engineering knowledge, students could potentially use the resulting tool to build animated visualizations of dynamic processes, such as the hea rt defects explored by Aldridge, and thus take part in a constructi vist learning exercise.
24 1. 4 .1. 3 Scaffolding Scaffold learning is the process of providing educational guidance proportional to based scaffold learning tools, supplemental information fades t he subject grows (Jackson et al., 1998 ; Oliver and Herrington, 2003 ). Scaffolding is known by many in education and psychology to be an effective teaching strategy (Roehler and Cantlon, 1997). For quant itative study results on the efficacy of scaffolding see, for example, Toth (2000) or Chang et al. (2002). In the proposed framework content embedded in the visualized domain ontology (e.g., classification information, diagrams) is the supplemental inform ation which could fade away as the students understanding grows. This could be achieved by an educator setting the domain ontology concepts and attributes that are to be displayed in a given module. 1. 4.2 Healthcare In healthcare, human error is one of th e main causes of preventable deaths and For example, a study showed that human error accounted for 82% of anesthesia related p reventable deaths (Cooper et al., 2002). strides have been made to increase patient safety by reducing human error. However, there is still much need for improvement, especially in th e area of training (Leape et al., 2009). One reason new training paradig ms are being suggested is because some junior physicians feel they have had insufficient exposure to uncommon conditions and procedures before embarking on independent practice (Mason and Strike, 2003). Some suggest a new medical training paradigm should integrate competency based
25 knowledge with clinical skill and expose trainees to more uncommon conditions (Rodriguez Paz et al., 2008). Consider a common source of human error: the administration of drugs to patients. Adverse drug effects (ADEs) are a pr imary reason patients die in a hospital (Bates et al., 1995). As a result of the intricate and complex nature of the cardiovascular system, cardiovascular drugs are known to cau se the most ADEs (Ebbesen et al., 2001). Even more difficult for trainees to understand is the varying effects drugs and other medical interventions have on patients with abnormal circulatory physiology. For example, hypoplastic left heart syndrome (HLHS) is a congenital heart defect in which some chambers of the heart fail to dev elop. HLHS results in abnormalities of blood flow and oxygen delivery which have been identified by medical educators as difficult for learner s to visualize and, therefore, understand. The ability to identify subtle differences between pathophysiol ogical scenarios is critical because interventions that may be helpful in one scenario could be harmful in another (e.g., the administration of oxygen may be harmful to an infant with HLHS) New training paradigms that integrate simulation with traditional teach ing methods will help trainee physicians avoid making harmful errors because of their lack of understanding of a rare and complex pathophysiology. T he ontology centric methodology outlined in this dissertation could provide important technology to help ad dress some of the needs of this newly proposed medical tra ining paradigm. Starting with a visualized domain ontology, such as one representing anatomy, medical educators should be able to construct a dynamic model, such as the Beneken (1965) model for card iovascular physiology shown in Figure 1 4 and then create a 3D visualization of the physiological process by augmenting the visualized
26 ontology with attributes, such as 3D meshes. The 3D interaction within the resulting integrative visualizations could e xpose medical trainees to more uncommon conditions and integrate their textbook knowledge of anatomy and physiology with interactive 3D renderings, potentially increasing their spatial understanding of the anatomy and physiology in question (Beer mann et al ., 2010).
27 Figure 1 simulation variables to visualization parameter s list s all possible visualization attributes and allows the user to cho ose which parameters can be influenced by the executing simulation model. Figure 1 2 The ontological spectrum proposed by Daconta et al. (2003).
28 Figure 1 3 A portion of the foundational model of human anatomy ontology ( Rosse and Mejino, 2 007) Figure 1 4. A dynamic compartmental model for cardiovascular physiology (Beneken, 1965).
29 A B C Figure 1 5 lists the vertices of a mesh and connectivity information to form edges and faces. B) A visualization of the geometric model of the heart C) A render created using the complete graphical model with material information.
30 CHAPTER 2 BACKGROUND This research builds on previous work across several computer science sub domains. In Section 2.1, background on ontologies, including those leveraged in simulation, is presented. In Section 2.2, background is given on visualization frameworks. Integrative m ultimodeling can be applied to link these two are a s in the user interface, an d is described in Section 2.3. Finally, Section 2.4 provides an overview of the state of the art tools used in industry 2.1 Ontologies 2.1.1 Ontologies: A W ide V iew cle, depicted a future of the Internet where reasoning is employed over vast ontologies, composed of classes and relationships, in the support of a more intelligent browsing and searching experience for users (Berners Lee et al., 2001). A variety of new kn owledge based applications would also emerge. It is over a decade later and the proliferation of the Semantic Web via such standards as the Resource Description Framework (RDF) is debatable, but ontologies have still became useful in many ways In biolog ical sciences, for instance, the Gene Ontology Project, created by a consortium of biological scientists to aid in the construction of a unified vocabulary across different biological domains, has been very successful and continues to grow (The Gene Ontolo gy Consortium, 2000 ; The Gene Ontology Consortium, 2010 ). Another example of an defines how mechanical components of systems fi t together to automate engineering model co mposition.
31 The increasing adoption of ontologies may be attributed in part to the graphical tools which have been developed for creating and editing them. Perhaps the most prolific of these tools is Protg, which was developed at Stanford and maintains an active community of thousands of users. Protg allows users to create knowledgebases with a hierarchical class structure and supports adding relations hips between classes (Noy et al., 2001). It also allows users to view their knowledgebases in several graphical layouts, including a frames based view and a graph based view of classes and attributes. After the substantial adoption of Protg by practitioners in a variety of fields, the Web Ontology Language (OWL) plug in was developed (Knublauch et al., 2004). The plug in adds support for creating and editing ontologies encoded in the OWL and support for reasoning using description logics. Reasoning functions supported by OWL include co nsistency checking (i.e., determining if a class could be instanced), and classification (i.e., determining a conceptual hierarchy among classes). Figure 2 1 shows a screen shot of the Protg loaded with an ontology representing a newspaper. 2.1.2 Ontolo gy Visualization The concept of the Semantic Web propagated through the computer science community, which prompted some researchers to explore the visualization potential of V an Harmelen et al. (2001) w ere some of the first to demonstrate the expressive potential of a visualized ontology. They created an ontology visualization repr esenting the Dutch job market by employing an augmented version of the spring force graph visual ization technique (Eades, 19 84) in which spring force values are mapped to semantic relationships resulting in semantically similar concepts being visualized near each other. Currently, Protg
32 provides some mechanisms for 2D graph based visualizations of the ontologic structure. TG Viz Tab (Alani, 2003) is a Protg plug in which leverages TouchGraph technology built with Java (TouchGraph, 2010). Due to the shortcomings of the 2D layout (e.g., clutter, overlapping), some began exploring visualizing ontologies in three dimensions. O ntoSphere3D ( Bosca and Bonino, 2006 ) allows users to browse a dynamic yet fixed height 3D tree and click on and change the level of detail concepts are visualized on a sphere whic h the user may rotate to see the big picture and ded Another example is the work done by Butain (2008), where semantic search results are visualized in 3D. Attributes of a semantic arranged in a radial fashion about the result. Conversely, the hierarchy of which the result is a member is arranged vertically. For more approaches to 2D and 3D visualization of ontologies, see the survey performed by Katifori et al. (2007). One final visualization effort worth mentioning is the Int elligent 3D Visualization Platform (I3DVP) created by Kalogerakis et al. (2006) They defined a methodology and constructed a platform that facilitated the merger of an ontology for graphics primitives with an arbitrary domain ontology. Part of their moti vation was to provide more visual results when performing Semantic Web enabled functions, such as semantic search. As a demonstration of the methodology, ontology based graphical representations were merged with a portion of the Gene Ontology I3DVP also supports visualization based
33 2.1.3 Ontologies in Simulation The work of Tolk and Tu rnitsa (2007) provides a suitable introduction to the idea of applying ontological means to the field of modeling and simulation. They focus on the theory of applying different parts of the onto logical spectrum (Daconta et al., 2003), shown in Figure 1 3 to challenges in the field. These challenges include the formal definition of agreed upon simulation concepts and primitives, multi resolution modeling, the creation of a universal language in support of simulation system composabili t y and execution, and the automation of simulation application development. Various research efforts address these issues and are described later in this section. The Semantic Web was described by visionaries in the infancy of semantics based application development and repre sentations. Much in the same vein, Tolk requirements for the Dynamic Web: (1) s ervices must provide a self description in a standardized format; (2) s ervices must be searchable based on their functionality; (3) Communication with the services must be possible; (4) s ervices must be able to be initiated with data; and (5) s ervices must be able to exchange data w ith other services. The remainder of this section is dedicated to highlighting research efforts that utilize ontologies in different ways to benefit the field of modeling and simulation. 2.1.3. 1 Ontologies to d efine s imulation c oncepts The Discrete event Modeling Ontology (DeMO) was created as a prototype ontology for modeli ng and simulation (Miller et al., 2004). DeMO was built using the OWL plug in for Protg. The ontology contains three top level classes: DiscreteEventModel, ModelComponent, and Mode lMechanism. Under these three
34 base classes is a collection of subclasses representing the necessary concepts within discrete event modeling. For instance, two subclasses of DiscreteEventModel (the 2 2 ) are State OrientedModel and Event OrientedModel. ModelComponent has corresponding subclasses such as state and event. The authors of DeMO concede that the classes such as state and event are fundamental concepts and will require further formal specifications which are implement ation specific for a given simulation environment. However, once users of DeMO provide these specifications for the "primitive" types, they can leverage the entire DeMO semantic network that is machine understandable. This will permit automation of certa in areas of simulation application development. The decrease in application development time through inference and reuse of ontology components afforded by the use of DeMO is touted as a core offering of ontologies to the field of modeling and simulation ( Lacy and Gerber, 2004). 2.1.3. 2 Ontologies for semantic web based simulation s ystem s Proposals have been made to utilize web components and online knowledgebases as the foundation for new component based simulation tools (Chandrasekaran et al., 2002). Th is approach is appealing because it allows simulation designers to choose from state of the art components when building a simulation system. An ontology could be utilized to organize the available components and allow sophisticated searching and semi aut omation when piecing together a simulation system. Chandrasekaran et al also suggest the use of simulation and modeling ontologies, such as DeMO, to create simulation components with a high level of intero perability. JSIM (Miller et al., 1999) is sugges ted as web based language to code the simulation components. Figure 2 3 depicts the different simulation components needed in the Semantic Web. Some are
35 entirely decoupled (e.g., databases and visualization tools) and others are coupled to varying degrees (e.g., animators and optimizers). The completely decoupled modules will integrate into simulation systems well because internet throughput will have a minimal effect. With smart design, however, the coupled modules could be also be integrated. 2.1.3. 3 O ntologies for environmental data r epresentation Authors of the Synthetic Environment Data Representation Ontology (sedOnto) project aimed to create an ontology for synthetic environments based on the Synthetic Environment Data Representation and Interchang e Spec ification (SEDRIS) (Bhatt et al., 2004). Synthetic environments (e.g., a virtual town square) have many applications in simulation (e.g. military simulations and vehicular piloting simulations). SEDRIS (2012) provides a representation of environm ental data and a loss less, non proprietary interchange format. The latter makes SEDRIS a natural fit for ontological formalization. The sedOnto is created after the SEDRIS UML notation is mapped to an OWL ontology. The authors of sedOnto eventually su pplemented their approach with a SEDRIS to OWL tran sform tool (STOWL) (Bhatt et al., 2005). STOWL uses assertions to automate the transformation of a SEDRIS defined Synthetic Environment into one defined in the OWL language using the Jen a 2 Ontology API ( Carroll et al., 2004). The transformation process is illustrated in Figure 2 4 along with a snippet of the resulting sedOnto OWL file. 2.1.3. 4 Ontologies for model t ranslation The Process Interaction Modeling Ontology for Discrete Event Simulation (PIM ODES) was developed to support the open interchange of simulation models in commercial simulation software pac kages (Lacy, 2006; Silver et al., 2006). PIMODES
36 is a vendor neutral intermediate model format encoded in OWL. The simulation software packages tested in the development of PIMODES were ARENA, ProcessModel, AnyLogic and ProModel. The process interaction data structures of which were composed to develop the final onto logy. To demonstrate the potential of PIMODES, Lacy created a translator program. The flow and function of this program is depicted in Figure 2 5 the conversion between proprietary model fo rmats. During tests, however, translations did not always work flawlessly as some semantic information was lost due to lack of access to the inner workings of the commercial (closed source) software packages. 2.1.3. 5 Ontol ogies to support r e u se As a lig hter alternative to distributed simulation solutions, Bell et al. (2008) developed a framework for simulation model re use that utilizes an inter organizational semantic network. The framework defines the ontology engineering process and the semantic search t echnology required to intelligently locate models. All models are based on representations found in commercial off the shelf simulation packages (CSPs), such as Simul8. The ontology engineering process starts with the anal ysis of models created in CSPs This analysis leads to the creation of OWL ontology classes to represent the necessary simulation components, a process known as Component Typing. These classes are then merged with DeMO. Dependencies are determined between classes in DeMO and concept s in the domain. The result is a simulation model ontology which contains domain specific knowledge of use to practitioners. This resulting ontology is termed the Discrete Event Simulation Component (DESC) ontology and is exported as
37 an OWL file and store d on a SEDI4G server. SEDI4G (Bell and Ludwig, 2005) is a research effort into grid based systems for semantic discovery. A component discovery system is also defined as a semantic search program that intelligently traverses the SEDI4G server given a text string as input. Results are returned to the searcher and the appropriate model can be selected. In the software prototype developed, the models are downloadable to an XML file which can be imported into Simul8, but this approach could theoretically be u sed with any CSP. A high level view of the framework defined by Bell et al. is depicted in Figure 2 6 2.1.3. 6 Ontologies to s upport f rameworks Benjamin et al. (2006) identified distributed, federated simulation systems as being difficult to build and requiring a long development time. Ontologies play an integral role in a framework which was developed to address these issues. Benjamin et al. describe several ways in which an ontologically centered design philosophy is beneficial to simulation model d evelopment. Ontologies serve to harmonize terminology, promote a common knowledge and language, and help relate organizational goals to simulation goals. They can also aid in model construction by clearly and unambiguously separating levels of abstractio n and determining simulation model objects, structure, and logic. Finally, they can help identify data sources and can inform data mining techniques by extracting semantics from textual descriptions. Ontologies can also be incorporated into the infrastruc ture required for distributed simulation systems. This approach has several advantages. Creating a simulation model ontology can provide a common language which can be used to share semantics between commercial simulation packages. Ontologies can also b e useful when constructing simulation systems composed of federated components by ensuring
38 component congruency and semantic interoperability. Additionally, support for multi resolution modeling (i.e., models representing the same phenomena but with differ ent levels of detail) can be added to ontological representations. With this support, rules can be applied to model ontologies ; and models of different resolutions can be merged, or new hybrid models can be created from multiple models representing the sa me phenomena but in varying capacities. The ontology centric design philosophy described in this section was applied to create the Framework for Adaptive Modeling and Ontology driven Sim ulation (FAMOS) (Benjamin et al., 2005). Modeling in FAMOS entails s electing and editing templates and composing models together from library components including ontologies for different domains and process models. Simulation and domain ontologies can be utilized in in support of the composition of distributed, federated simulations systems. Miller and Baramidze (2005) also noted the ability ontologies have for supporting multimodeling and multi faceted reasoning to generate and interpret simulation results. 2.1.3. 7 Merging simulation ontologies with domain o ntologies Of the previous applications of ontologies to simulation systems, the methodology proposed in this document is most akin to those which utilize domain specific ontologies. S ilver et al. (2007) developed the ontology driven simulation (ODS) framework. Their framework utilizes domain ontologies along with DeMO and consists of an ontology mapping tool, a model markup generator, and a simulation code generator. The ontology map ping tool allows users to map classes of the domain ontology to classes in the simulation ontology. Their design suite was demonstrated starting with the problem oriented medical records ontology (PMRO) to build a clinical
39 simulation. Examples of mapping s in this scenario include mapping the class Patient in PMRO to the class Resource in DeMO, and mapping the class ClinicalExamination in PMRO to the class Activity in DeMO. After the domain ontology is aligned with the simulation ontology, an intermediate markup file can be generated that represents the simulation instance as either a Petri Net model or a process interaction model. This intermediate markup file can then be translated into executable models that run in JSIM (Miller et al., 2000) or ARENA. Silver et al. describe four main benefits of the ontology driven approach: (1) the use of a shared vocabulary facilitates good communication between modelers and end users, (2) models created for different simulation packages using different simulation wor ld views can be represented in a unified language, (3) models are encoded in a Semantic Web compatible language, thus supporting potential web based model repositories and web component based simulation systems, and (4) model development can be sped up by allowing modelers to assemble models by using a set of predefined classes represented in the domain specific and simulation ontologies. Together, these projects illuminate the roles ontologies have played in the field of modeling and simulation. They are often used to formally define programmatic structure as a way to encourage simulators to speak the same language when coding systems. This supports interoperability and model composability. Further, various levels of reasoning have been applied to simula tion ontologies resulting in the partial automation of simulation application development. Some also explored the use of domain ontologies for constructing simulation systems. The methodology proposed in this document will compliment these previous eff orts in utilizing ontologies in modeling and
40 simulation by manifesting the ontologies as 3D graphs within a visualization environment for authoring of simulation models and visualizations. The 3D graphs can then be used as part of an integrative visualiza tion or as an interface affordance. 2.2 Visualization F rameworks Since researchers began to explore the potential of computer graphics, the visualization community has called for scientific visualization frameworks that would abstract low level graphics manipulation and rendering functions away from the user, and would provide a more intuitive integration between data sources and visualization design features for domain practitioners (Treinish et al., 1992) Some of the first attempts at creating such a theoretical framework were by Haarslev and Moller (1988) and Upson et al. (1989) Since these seminal projects, much research has been dedicated to designing visualization frameworks that employ the increas ing power of hardware and leverage our increasin g understanding of what makes interaction techniques effective (Tory and Moller 2004). This continued work has provided a set of visualization tools to those working in many scientific and engineering domains. For the remainder of this section, select wo rk towards creating visualization design systems is presented in chronological order. Visualization of E xperimental S ystems ( Haarslev and Moller, 1988) o ne of the first efforts in visual dataflow progra m ming to take advantage of the mouse and keyboard. This allowed user s to directly manipulate dataflow graphs. The system was originally designed to support image processing techniques but was extended to allow construction of more general algorithms, and to allow visualization of their output. Ap plication Vi sualization S ystem (Upson et al., 1989) a system oriented towards engineers and scientist who did not necessarily have experience in computer
41 programming. Within the framework, a direct manipulation interface was employed to allow the user to construc t dataflow networks that ultimately fed into one of the provided visualization types (e.g., voxel or iso surface renderings). There is a similar approach to visualization construction in many m odern tools ( s ee Section 2.4 ) Object Oriented 3D Graphics Too lkit (Lucas et al., 1992) a framework oriented towards visualizat ion application programmers. The framework allowed programmers to more efficiently develop visualizations by providing features such as a graphics for rendering, direct manipulation interface functions and code templates for 2D widgets. VISAGE (Schroder et al., 1992) a n object oriented visualization design library. Objects were built in LY MB (Lorsen and Yamrom, 1989) an animation progra m ming library coded in the C language. A collection of visualization objects exist in the library (e.g structured grids, unstructured point sets). Each object in the library suppo rted the notion of time, allowing a visualization des igner to animation their representations. VISTA ( Senay and Ignatius, 1994) a knowledge based system for visualization design. Domain knowledge is mapped to visualization primitives (e.g., 3D scatter plot, shape, size). Once a visualization is automatic ally generated based on input data, the designer may interactively modify the results to further suit their needs. The Visualization Toolkit ( Schroeder et al., 1996) a C++ library of visualization objects. VTK is intended to be executed in a dataflow e nvironment and can utilize a variety of rendering APIs that were prolific at the time of development (e.g., OpenGL,
42 DSR Open Lab ( Kolodnytsky and Kovalchuk, 2001 ) an interactive visualization system with a focus on the graphical user interfac e and usability. The software used OpenGL to render a multitude of 2D and 3D histograms. The user interface code was multi threaded, permitting new levels of interactive graphical manipulation. Davis (Huh and Song, 2002) a statistically oriented J ava b ased tool for creating advanced 2D data plots. Davis was unique at its time of release because it supported sophisticated statistical visualizations techniques (e.g., flipped empirical distributions, Grand Tours ) through user interface controls. The InfoViz Toolkit (Fekete, 2004) a Java 3D based code library that allows a programmer to easily encode traditional information visualization structures. Data is obtained by importing tables with named columns. prefuse (Heer et al., 2005) a robust jav a based toolkit for creating information visualizations. The toolkit provides classes to support typical information visualization tasks and renders visualization output using the Java2D API. The creators of prefuse performed several user surveys and exp eriments that allowed them to identify issues with the toolkit design and vindicated perfuse as a useful tool for visualization constructions for those with Java programming experience. VIS STAMP ( Guo et al., 2006 ) a visualization system for space time a nd multivariate patterns VIS STAMP provides geovisual analytic tool s with support for cartographic based visualizations. It also supports analysis of spatio temporal patterns by employing self organizing maps, parallel coordinate plots, recordable matric es, and recordable map matrices.
44 simulation systems. The interface is specified in an XML file and then parsed and created by GiCK. Behaviorism ( Forbes et al., 2010 ) a Java/OpenGL based framework for creating 2D and 3D visualizations. Users of the framework create visualization s by designing around three graph structures: the data graphs that define data sources, the scene graph that defines graphical resources and their transformations, and the timing the visualization scene. A collection of interactive visualizations were created with behaviorism that demonstrate the versatility of the approach. Edi t Flow (Benzaken et al., 2011) One of the first i nformation visualization framework s built on top of a database management system (DBMS). The DBMS supports strong scalability and persistence across multiple users features not inherently present in frameworks that operate by loading all required visualization data in memory. 2.3 Integrative Multimodeling F ishwick (2004) presented the challenge of integrative multimodeling, which requires using multimodels and adhering to new HCI criteria which did not traditionally apply to simulation. Most models of use are multimodels since individual model types are oft multimodeling calls for a closer linkage between the abstract model and the phenomena which it represents. One way to achieve this is to graphically represent the abstract model and phe nomena within one cohesive interface and allow user interactions on both the abstract and concrete graphical representations. This new modeling technique also has four main HCI requirements: usability, emotion (by means of connotation), immersion, and cust omization.
45 2.3.1 An Integrative Model Blending Environment An early application of the principles of integrative multimodeling is the work done by Park and Fishwick (2005). A framework was created that allowed for the integration of different model types w ithin a 3D simulation environment. Dynamic, interaction, ontological, and geometric model types were included. The framework provides a Model Explorer for dynamic modeling and an Ontology Explorer with OWL integration so users can create OWL classes and su bclasses. The framework runs inside of Blender (Blender3D, 2012) and requires the RUBE simulation library (Hopkins and Fishwick, 2001) to build dynamic models and execute simulations. Users can create geometry using Blender or import pre existing meshes. D ynamic model components (e.g., blocks within a Functional Block Model) can be selected from RUBE and associated with geometry in the scene. Using Blender Game Engine functionality, an interaction model can be created and linked with geometry. Shim and Fish wick (2008) demonstrated the use of integrative multimodeling by co locating a s ystem d ynamics model for ecological phenomena and behavior. 2.3. 2 The Augmented Anesthesia Machine A recent attempt to apply integrative multi modeling to anesthesia training has been successful by employing mixed reality. Quarles et al. (2008) created a mixed reality training tool that co l ocates abstract models used to teach anesthesia machine operation with the components on the physical anesthesia machine. The tool, known a s a tracked tablet PC which serves as a window into the real world. The tablet is tracked with six degrees of freedom using infrared cameras and small reflective objects. Users hold the tablet PC in front of the physical anesthesia machine and see a 3D rendering of
46 the machine on the display from the perspective of the tablet. On top of this 3D rendering, the abstract model is visualized using iconography, and color coded p article systems are used to visualize invisible gas flow within the machine. User studies were performed that compared the AAM to traditional training methods (e.g., using only an instruction manual and the anesthesia machine) and web applets which used ab stract visualizations (Fischler et al., 2008). The results of the study showed that the AAM was most successful at bridging the gap between abstract and concrete knowledge, thus proving the approach is a potentially viable educational technique which shoul d be explored further. The author of the AAM points toward the use of ontologies as a way to define a theoretical framework for such mixed reality systems (Quarles, 2009). I view the proposed methodology as a natural evolution of the work done by Quarles et al. (2008). The methodology will provide a theoretical foundation for software tools such as the AAM. By using domain ontologies (e.g., an ontology for anesthesia equipment) as the starting point for dynamic model building and visualization construction users will not have to hard code visualization solutions. 2.4 Simulation and Visualization Design in Practice This section is dedicated to providing a brief overview of the state of the art for simulation based visualization design. This includes existi ng techno logy that provides some sort of interface to visualization functions that can be used create to reasonably sophisticated visualizations based on collected real world data or simulation output These technologies range from pure graphics programmi ng to purely graphical user interfaces for non programmers
47 2.4.1 Graphics Progra m ming Graphics progra m ming is the most powerful and customizable way to create visualizations. Graphics programmers leverage low level APIs such as OpenGL or DirectX, or high er level programming environments such as OGRE. When employing graphics program m ing for visualization design simulation models that drive the visualization can be co ded in a separate package or along with the graphics in the same code base An example of employing graphics progra m ming is the detailed segmented cardiac visualization coded in OpenGL ( Wang et al., 2011 ) Tools that abstract away graphics progra m ming from the end visualization designer (including the goal of this work) still must build upon some graphics API provide the visual output to the user. A limitation of the pure graphics programming approach is that it, of course, requires programming aptitude which practitioners operating in most domains do not possess. In some cases graphics programming is even overkill for those with programming expertise and in this case the designer will choose to use o ne of the visualization design tools listed in the following sections. 2.4.2 Engineering Simulation Tools with Visualization Add ons I classify engineering simulation tools as tools that provide a graphical user interface to advanced mathematical modeling. E xamples of such tools are Simulink, LabV IEW and Ptolemy. The Simulink method of creating visualizations based on the encoded mathematical model was described in Chapter 1 and depicted in Figure 1 1. LabVIEW provides an interface similar to Simulink fo r mathematical modeling, but integrates 3D visualization differently Within LabVIEW 3D visualization objects known as SceneObjects, are represented as blocks and can be connected to the greater mathematical model on the LabVIEW canvas ( National Instrum ents 2008) OpenGL is
48 the underlying graphics technology used in LabVIEW to render SceneObjects Transform nodes can be added to the canvas and used to link simulation variables to graphical parameters Visualization output (all of the SceneObjects) ar e rendered in a 3DPicutreControl window when the simulation model is executed. LabVIEW also provides extensions for graphics programmers to code custom visualization objects to be utilized in comprehending simulation output. Ptolemy, a Java based simulati on platform for simulation modeling and execution operates in a manner similar to LabVIEW. The focus of Ptolemy, however, is more on merging heterogeneous model types within a single modeling environment ( Lee and Seshia, 2011 ). With regards to visualization, Ptolemy leverages t ransformation objects that may be placed on the Ptolemy canvas ( this is similar to LabVIEW). These transformation objects may be chained toget her to create a final composite transform that can be based on simulation output to varying degrees. An example of this is shown in Figure 2 7. A basic simulation of orbital dynamic s between the Sun, Moon and Earth was created in Ptolemy and visualized with 3D shapes that Ptolemy provides (spheres in this case). Th e portion of the model shown in the figure pertains solely to the graphical output; simulation calculations are performed at a different level of the model. 2. 4.3 Game Engines and 3D Modeling Packages Game engines are software tools that enable designers t o create interactive virtual environments with a focus on customizable graphical presentation and diverse user input. Examples include The Unreal Engine, the Source Engine and Unity3D ( Petridis et al., 2012) Similar to simulation packages, each game en gine requires users to learn the specific modeling paradigm defined within the engine. Often a game engine will support scripting in some pre existing program m ing language. Using such tools can be
49 viewed as engineering practice because any content m odeli ng performed will need to be translated through the game engines modeling paradigm and script will potentially need to be coded. Certain engine designers are aiming for ways to open their technology to practitioners in domains outside of engineering and c omputer science but the current state of the art still requires substantial engineering effort ( Thiriet et al., 2011 ) Modern 3D modeling packages are similar to game engines in that they provide 3D rendering capabilities and often include scripting features for the creation of customized plugin s However, 3D modeling packages focus on providing a tool set for the creation of 3D content (modeling, t exturing and animating 3D objects), not the synthesis of content and interaction to create a game simulation dynamics can be encoded in 3D modeling packages, but this requires custom scripting or selection from a pre determined set of simulation features such as the fluid or cloth dynamics options provided in Blender. 2.5 Summary In this c hapter, relevant background was present ed from the areas of ontologies, simulation visualization frameworks, and integrative multimodeling Ontologies have been used to approach some challenges in simulation, but have been applied mostly on the computational level. Visualization frameworks have progressed recently but have been oriented mainly towards information visualization, not simulation systems. Finally, integrative multimodeling is a new area concerned with linking abstract and real world representation and is applied in this work towards the definition of new theoretical interface State of the art tools used in practice for simulati on based visualization construction were also presented in this chapter.
50 Figure 2 1 Protg (Noy et al., 2001) displaying an ontology for a newspaper. The attributes) of a Columnist are shown next to a blue icon. Figure 2 2. A partial view of DeMO (Miller et al., 2004). Model Concept types are enumerated.
51 Figure 2 3. A high level view of the Semantic Web enabled simulation design system proposed by Chandrasekaran et al. (2002). Figure 2 4. Transforming a SE DR IS file in to an OWL ontology (Bhatt et al., 2005).
52 Figure 2 5. High level depiction of the translator written by Lacy (2006) to demonstrate PIMODES. Figure 2 6. The ontology engineering, storage and search framewor k created by Bell et al. (2008)
53 Figure 2 7 An example of transforming and rendering 3D objects based on simulation variables in Ptolemy. Textured spheres representing the Sun, Earth and Moon are transformed using a series of Ptolemy transform objects (translate rotate, scale). A ViewScreen3D objec t collects the transformed spheres as input and renders them in a separate Java3D view shown to the right.
54 CHAPTER 3 ONTOLOGY CENTERED INTERACTION THEORY AND FRAMEWORK The theoretical interface requirements and design methods described in this chapter rep resent the core contribution of this work. A case study presented in the following chapter may help ground the high level concepts presented here through real world examples. The design of the proposed framework should facilitate simulation model buildin g and integrative visualization construction within a single interface model that leverages pre existing domain semantics when possible. The defined interaction theory should demonstrate that a visualized ontology could be the central connective interface structure through which these different activities can be afforded. The ontology centric design can also support executing the designed simulation models and, The interface requirements described in this chapter support a simulation and visualization design environment. This is a 3D graphical environment with controls to start, pause and reset the simulation and visualization at any time. The three main c omponents of an ontology -concepts, relationships and attributes -all play important role s in the proposed interaction theory The role of each component will be described in Section 3.1 A detailed look at the various design affordances of the theo ry will be presented in S ection 3.2 A technical framework that can house the proposed interaction theory is presented in Section 3.3. A methodology for using the proposed framework is pre sented in Section 3.4 Finally, a classification, comparisons, and limitations of the framework are discussed in Section 3. 5.
55 3.1 The B ase S tructure 3.1.1 Concepts In a domain on tology, concepts represent high level objects in the domain space (e.g., in the medical domain anatomy could be represented by a set of interconnected organs ) Likewise, when an ontology is visualized, a concept can serve as a high level interface handle to parameters required for simulation and visualization. That is, the visual ization of the ontology c oncept can serve as a handle to more information about the c oncept in the form of attributes ; and can serve as a visualization design interface structure that allows users to perform graphical t ransformations When a node is transformed, any 2D or 3D visu al objects that are assigned to the concept as attributes will be transformed as well. The visualization of the concept object is the graphical parent of all visual objects assigned through attributes. A depiction of this basic idea is given in Figure 3 1 The concept is visualized as a gray sphere. When the sphere is selected, a graphical widget appears to allow for 3D translations. A 3D mesh (the bunny ) and image (xray.jpg) are assigned to this particular concept. An example graphical transform of translation is depicted in Figure 3 1c but a rotation or scale widget could be root ed at the ontology concept as well. As the concept is translated, the 3D mesh moves with the concept and the 2D image maintains the same 2D offset. 3.1. 2 Attributes A basic visualization of a concept should serve as an interface to attributes of that concept When a concept is selected not only should graphical transform options become available, a list of all attributes relevant to that concept should be displayed D epending on the interaction modality (different modalities are described in Section 3.2.9) attributes can either be read only or modifiable. In the proposed framework,
56 attributes play a critical role ; they can be used to define simulation variables, visua lization parameters, and any required influence between simulation and visualization. T o have consistency in the interaction model and to massage the ontological structure into something that is well suited for a visual interface to graph based simulation models and visualization design attributes may be added to relationships. In the common notion of ontology relationships do not hold attributes and, therefore, a translation will need to occur if the ontologies leverage d in this framew ork are to be ut i lized elsewhere. Attributes of relationships are discussed in the next section and the required trans lation is discussed in Section 3.2.11 3.1. 3 Relationships Relati onships in an ontology denote a unidirectional semantic co nnection represented as an edge that connects two ontology concepts. In the proposed framework, relationship edges can be used to just relay basic semantics as in typical onto logical solutions or can be sculpted into a meaningful curved path within a 3D v isualization environment. The resulting curved edge can be used to denote the path of flow within a dynamic system, or allow further expression and serve as a guide for anima ted objects, such as particle systems. The physical structure of the relationship as defined interactively by the visualization designer can be encoded by a n attribute assigned to the edge. The attribute should be a list of ordered points that define the edge and a potential extra that is piece wise linear, or a curved edge that should be interpolate d between the points It may be beneficial to add attributes to edges as well depending on the modeling paradigm being
57 Attributes for a relationship should be displayed when the edge is selected, as is the case with attributes attached to concepts. 3.2 Structural and Sema n tic Affordances A collection of design affordances will be presented in this section. These affordances are diverse, ranging from the ability to stor e sim ulation equations as attributes, to using relationships to form the path of particle systems. The defin ition of affordance used in this context is : (Gibson, 1979) in support of simulation modeling and visualization design. The following 11 sections each describe an affo rdance. C oll ectively, I view this set as a showcase for the use of the ontology centric approach described in this dissertation. 3.2.1 Meta attributes Meta attributes are attributes of attributes that are required in the proposed framework to define properties significant to simulation modeling and visualization design activities. These meta attributes include values to determine if an attribute should be visible in a resulting visualization viewer module and if an attribute should be a numerical variable or text based If an attribute is deemed a variable, meta attributes exist that can define the range of acceptable values for the variable, or a discrete set of acceptable values if the variable is crisp. When adding ontology member attributes required when creating an envi sioned visualization module (e.g., the 3D position of an ontology concept ), a designer may want to limit visibility of such attributes in the final visualization. Design centered attributes will not be relevant in most viewing contexts and likely distract from the information the designer intends to convey. To this end, a meta is added
58 to each ontology concept attribute. This is a true or false property whose value determines if an attribute will be visible in the viewer module. O ther meta level attributes in the framework allow a designer to attach interactive controls to numeric attributes. These controls allow viewers to tweak visualization properties in real time. Attributes could be linked to simulation dynamics (e. g., coefficients in the underlying model) or purely graphical features (e.g., the c olor of a material assigned to a 3D mesh). Regarding numerical simulation and visualization parameters, there is a true or false value that determines whether an att ribute is marked a scal ar or 3D vector type. 3D vector types are well suited for some visualization parameters (e.g., position: x, y, z and color: red, blue, green). Further, attributes tha t are marked as a variable can al Adjustable attributes can be interactively modified. The control for adjusting should allow modifying a variable attribute to values within a defined range. This range is bounded by a level attribute One way to support this is thr ough a linear slider control With respect to the slider behavior, a designer may not always want this control to be smooth and continuous, but allow only for discrete values from a set. To this end, the proposed When an adjustable variable is marked as crisp, a set of possible values may be assigned to the variable. These values should be sorted and the interactive control (e.g., the slider) s crisp. All meta attributes are depicted in the user interface sketch presented in Figure 3 2
59 3.2.2 Attributes as S imulation E quations While most simulation modeling paradigms can be made to conform to the ontological graph structure, there will always be mathematical customizations that are desired in certain simulated scenarios (Krahl, 2002) Often these additions are made through custom scripts. In visual environments (e.g., Simulink, Ptolemy) equation blocks can be added to the canvas that consider any input as a variable to use within the equation and output a calculated value. Another approach in these environments is to leverage a functional block model notation to build a graph that represents the desired formulation. In functional block model notation operators are blocks within a flow graph with operands as input, and the resulting value of the operation as output. Representing e quations as graphical objects (e.g., blocks on a canvas in Simulink or Ptolemy ) within a dynamic model representation does not mesh well with the ontological graph visualization approach Because t he purpose of an equation is to calculate a value which, ultimately, is just an attribute of some concep t with in a domain (e.g., blood volume is an attribute of an organ in the domain of physiology), the equation can be treated as such an d its representation can be encapsulated within that of the concept s. Along with a de cluttering of the visual design environment defining simulation equations as attributes of concepts has the additional benefit of allowing equations to be represented in a manner close r to their native text based form as compared to the to graph based structures. Any implementation of an ontology based equation solver will need to define some base syntax to denote the concept of time, both total time and the current time slice; and to determine from which concept a variable is being referenced. For the latter, the dot notation used in object oriented programming could be employed Figure
60 3 3 explores the dot notation with two approaches for retrieving a variable value from a foreign concept. In Figure 3 3A, syntax is shown of the following structure : concept(). In Figure 3 3B, a simple function output() is used to find a node adjacent to an out going edge with a given name. This syntax has the following structure: output(). A function of input() can be used in a similar manner. The second syntax style has the advantage of being general enough so that a single equation attribute could be distributed across an entire dynamic mode l because no explicit concept names are required. 3.2. 3 Taxonomies for Object Crea tion O ntology inheritance is represented by an is a relationship between two concepts. A tree of concepts connected by is a form s a taxonomy This taxonomic structure can be useful when creating simulation o r visualization objects. If a root concept can serve as a template for other concepts then any de sign choice applied to the root can be automatically propagated to any concepts connected through out going is a relationship s During visualization construct ion, certain graphical properties can be applied automatically to a collection of sub concepts (e.g., all items of a certain type should be colored red). In simulation models, given a certain modeling paradigm, model blocks may all have the same variable a ttached (e.g., in compartmental modeling all compartments have simulation variables volume, pressure and flow). 3.2. 4 Creating T ransform H ierarchies and V ertex G roups The has a relationship is a common semantic notion and can serve as a useful interface af fordance. The proposed framework leverages this relationship type to create
61 graphical hierarchies (i.e., scene graphs) and to define vertex groups within a mesh to create vertex animations. A has a relationship between two ontology nodes will form a paren t child pair. This mimics common parenting or grouping functionality present in 3D modeling and animation software ( Derakhshani 2010) With has a relationships present, transformations applied to a source node (the parent) of the relationship are propagated down to the destination nodes (the children ) This technique can be applied to transform a set of disjoint meshes or to define regions within a single mesh to be transform ed Figure 3 4 illustrates the use of the has a relationship in a simple scene containing a tea pot on top of a table Because a has a relationship connects the two concepts, when the table is transformed, the tea pot will b e transformed as well. The tea pot will be transformed about the location of the table concept node (i.e., the tea pot Vertex animations are animations in which a subset of vertices is deformed within a mesh over time. Different transforms may be applied to different vertex subsets, resulting in complex animations. This is in contrast to animation in which a transform is applied uniformly across an entire mesh. For example, compare a n animation of a beating heart in which the heart scales uniformly to an animation in which the different chambers of the hear t scale independently over time Vertex animations are often n Vasconcelos 2011). Some sort of skeletal structure is provided in the user interface that can be registered with a mesh and manipulated to create vertex animations. Designers can create simula tion based vertex animations in the proposed framework b y leveraging the ontology visualization as the required skeletal structured.
62 In the proposed framework, r igging 3D meshes by leveraging the ontology structure requires has a relationships be present in the ontology. The has a relationships, which should b e unidirectional, along with their adjacent ontology nodes form a hierarchy of parts and subparts. If a mesh is assigned to the root of this hierarchy, then in some cases its subparts will represent sub regions of the assigned mesh. The ontology nodes re presenting subparts can be positioned within the root mesh and serve as the interface handle to creating vertex groups of the root mesh. These groups should be centered at the ontology node and can be bounded by a variety of geometry. Further, the span o f this bounding geometry can be controlled by the addition of an attribute named influence radius Once vertex groups are formed by positioning ontology nodes, affine transformations may be applied to each vertex set. Transformations should treat the onto vertices are defined). These transformations may be linked to simulation variables of a node through the a ddition of influence attributes. is shown in Figure 3 5 The bounding geometry for this group is a transformed sphere. 3.2. 5 Influence for Ontology Based Animati on The ontology can serve to connect simulation variables to visualization parameters resulting in dynamic 3D visualizations based on simulation behavior. By adding influence attributes to ontology concepts, a designer can link the value of one attribute to another by mapping between linear ranges during simulation execution. To create the
63 semantic link in the ontology, attributes influence source influence destination influence source range and influence destination rang e are required. Ranges can be bounded by either vector or scalar values. A list of suggested attributes that could be used to create and tune influences is shown in Table 3 1 Adding influences should be supported during simulation execution, resulting i n the desirable feature of immediately seeing the effect of mapping a simulation variable to a visualization parameter ( Shneiderman and Plaisant 2005 ). Combining influences with adjustability ( see Section 3.2.1 ) can result in a simple way for a designer to interactively fine tune a visualization parameter to achieve a desired animation effect over time. 3.2. 6 Designed Ontology Representations In some cases, visualization of the graph based ontological structure, with concepts as nodes and relationships as edges, may be useful on its own with few additional enhancements ( Katifori et al., 2008 ). For example, concepts could be color coded or relationship visualizations could be set to a n edge width that maps to a simulation coefficient. Special attributes can be defined to support the graphical customization of the ontology visualization. This is required to differentiate between what is an additional graphical element (to be visualized alon g with the ontology) and what graphical elements are to replace or augment the standard ontology components visualization. A way to allow the designer to delineate between these two options is to allow a set of attributes with a prefix to determine that i t refers to the graphical properties of the ontology visualization itself. For example, the prefix presentation
64 could be used. In this case the attribute presentation mesh refer s to the 3D mesh used to represent the concept and the attribute mesh would r efer to the mesh assigned to the concept that is still visible even if the ontology is hidden. 3.2. 7 Ontology Guided Particle Systems A particle system is a collection of duplicated 2D or 3D graphical objects that are animated and subjected to external fo r ces over time. Particle systems are used in real time computer graphics to simulate a variety of effects (Hastings et al., 2007), including fluid flow and lighting effects. Because of their increasing application, tools for the creation of particle syste ms are included in most modern game engines and 3D modeling packages. Particle systems can be useful when visualizing simulation output and are an important consideration when designing general purpose visualization software. The graph based ontology vis ualization can serve as an interface handle to connect particle appearance and dynamics to simulation variables. The ontology concepts can serve as particle emission points. From the emission points, particles can be ejected in a certain direction or foll ow the path defined by the curvature of an out going relationship edge. Particle properties (e.g., speed, size, color) can be set by adding attributes to the appropriate concepts and relationships. Particles that follow the path of a relationship can hav e constant properties derived from attributes assigned to the relationship, or have dynamic properties that change as the particle moves along the relationship (e.g the color of the particle could fade from blue to red overtime). Dynamic properties can be encoded by attaching attributes to both source and destination ontology concepts. For example, if an attribute particle size is attached to both the source and destination concepts of a relationship, then the
65 size of the particle can be can linearly i nterpolated as it traverses the connecting relationship, thus causing the particle to shrink or grow overtime. The inte rpolation of particle behavior along a relationship can be facilitated by checking attributes types on relationships and adjacent concep ts. If a particle attribute is not of variable type (i.e., numerical) then it can be assigned th e name of an attribute assigned to adjacent concepts. In other words, the particle property is either assigned a constant numerical value at the relationship level, or bound to adjacent concept attributes and subjected to interpolation across the relationship This is illustrated in Figure 3 6 where the particle property particle speed is added as an attribute to a relat ionship and assigned the value speed Because this attrib ute is not of variable type, the value for particle speed is calculated by considering the numerical values of the attribute na med speed that should be present in the two adjacent concepts A collection of suggested particle attributes are described in Table 3 2 3.2 8 Attribute E xpansion Each concept has a collection of attributes. These attributes may be of varying interest to someone consuming the final visualization module As part of the proposed framework, two styles of attribut e visualization are defined: docked and expanded. Attributes that are docked are presented along with all other attributes within a list configuration. This representation of attribute s should be of the lowest detail, taking up the least space. For exam ple, if an attribute is an image, then when it is docked it should only be displayed as the image name or the name accompanied by a small thumbnail. If the attribute is a simulation variable, then when it is docked it can be
66 displayed a s its real time nu merical value, and an interactive 2D variable plot when it is expanded. If presented with a collection of docked attributes after selecting a concept or relationship, a designer or consumer can make a choice a s to which attributes are of interest, and the interesting attributes can be expanded. Generally, a n expan ded attribute should be represented at a higher fidelity than docked attributes within the visualization while maintaining a visual link to the concept. An image attribute, when expanded, could b e displayed as the full image along with an edge pointing to the concept. This image could be moved about the screen space on a 2D plane parallel to the camera can be thought of a visualized. In Figure 3 7 the dashed arrow depicts the process of expanding a docked attribute. Figure 3 8 view. 3.2.9 Ontology B ased Interaction Modalities A simulation and visualization designer will most often utilize semantics during the design phase that are irrelevant to the end consumer (e.g., co worker, student ) and, if visualized, would result in unnecessary clu tter. End consumers also have no need to add or remove attributes due to their more passive relationship with the visualization and simulation that they are consuming. Further, transforming the visualization objects in a designed module is unnecessary a nd m a y confuse users presented with transform widgets, and relationship curvature control s To this end, two interaction modalities are defined and supported by the ontological structure: a designer modality and a vie wer modality.
67 Basic interactions s hould be supported across both modes. These interactions include: 3D camera controls, selection of concepts and relationships, attribute expansion and the adjusting of attributes that are marked adjustable. Within the design modality, additional interac tions should be supported including : the ability to add and remove concepts, the ability to add and remove relationships between concepts, the ability to transform concepts (and any assigned visualization attributes), the ability to sculpt the curvature of any relationship, the abil ity to add and remove attributes of concepts and relationships, and the ability to view and modify all meta attributes. 3.2.10 Semantic Styling Most ontology visualization efforts provide a 2D or 3D node and edge based approa ch with minimal visual diversity or customization tailored to the domain from which the ontology was derived (Pietriga, 2006). To address this, graph style sheets were created that map the semantics embedded in an ontology into meaningful graphical repres entations One example of this is STOOG (Style sheet based Toolkit for Graph Visualization), created by Artignan and Hascot (2010), which provides an interface for user s to define 2D icons to represent ontology concepts and to specify which attributes ar e rendered in the final ontology visualization. GraphViz (Ganser and North, 2000) was leveraged in STOOG for automated graph layout. In simulation modeling graph style sheets can be used to tai lor the ontology representations to be consistent with the visual simulation modeling practices used in a given domain. Many graph based simul ation modeling paradigms exist and leverage a different set of visual syntax (Page, 1994). Being able to express dynamics in a visual paradigm familiar to a practitioner could help ease the visualization designer into the 3D virtual environment. In the given scenario, a practitioner could use a customized
68 version of the propos ed framework with a graph style sheet applied that was created based on the models used in that domain. Figure 3 9 demonstrates a simple transform from a graph based ontology representation with two concepts and a relationship into that of a systems dynamic model with source flow rate and lev el objects. 3.2.11 Ontology Portability through RDF The R DF (Resource Description Format) was created to support the envisioned Semantic Web The abstract RDF structure can be understood by machines and humans. An RDF object consists of a subject, an object and a predicate ( Miller, 1998 ). For example, in the in RDF the This RDF format maps cleanly to the ontological structure, which is composed of concepts connected by unidirectional relationships. Attributes in an ontology can be stored in RDF as terminal objects. Due to the fact that the ontological structure is augmented in the proposed framework with meta attributes and a ttributes attached to relationships, a simple translation phase is required to map the augmented ontology into the RDF standard. T o perform the translation from the augmented ontology concepts and relationships can be elevated one level. This elevation w ill create concepts from elements that were previously attributes and create attributes from elements that were previously meta attributes. A technique to account for attributes of relationship s is to morph the relationship in to a concept that connects t o the previous destination concept and has all the attributes attached to the original relationship ( Gomez Perez and Corcho, 2002 ). Objects of this type are referred to in RDF as "blank objects" because they have no reference to online semantic repositorie s but serve to link various well
69 defined semantics in scenarios that demand a more complicated structure than the simple object, predic ate, subject triple can provide. A depiction of the meta attribute and relationship attribute morphing is depicted in Fi gure 3 10 Depending on the desired application of the ontology, this mapping for meta attributes and relationship attribute s may obfuscate the core domain semantics. To address this, in stead of performing a mapping, a supplemental file structure could be defined that houses all semantics defined in the framework that do not map cleanly to the base ontological structure This file could be exported along with the base ontology for permanent storage and interoperability. 3.3 Technical Requirements In this section, the high level technical requirements will be presented. This will help summarize the proposed framework and determine the various modules required to implement the simulation model builder and visualization designer defined in this chapter. The framework requires the following components: an ontology parser, an ontology ex porter, a simulation solver, an ontology solver, a graphics renderer, a designer user interface and viewer user interface. An ontology pruning component and a domain sty ling component are optional. The ontology parse r serves to load the ontology, encoded in a certain format (e.g., RDF), from a file. The parser should create the necessary internal ontology data structures in memory to be used by the other components of th e framework. The ontology exporter saves the ontology for later use or use in other ontology based frameworks. T o execute any simulation model, a simulation solver will need to be created. The simulation solver should: read the current values of simulat ion variables from the
70 ontology; perform the necessary calculations, perhaps domain specific calculations; and then write the new values for the simulation variable back into the ontology. There should also be an ontology solver whose purpose is calculati ng the values of any influences and solving any attributes that are equations. Any variable, simulation or visualization oriented, that is not listed as just purely numerical (i.e., it references another attribute by name), will have its value resolved by the ontology solver. The simulation solver can be updated at any interval but will need to know the time since the last update (i.e., the simulation time slice). The effect of large time slices when solving certain simulation models should be considered, because large time slices can be problematic in models with sensitive coefficients. For graphics rendering, a component dedicated to the rendering of 3D textured objects is required. This graphics renderer component can be coded in a low level graphics language, or one utilizing a higher level graphics library. The graphics component will need to communicate with the ontology solver to obtain any ontology concept attributes linked to graphical properties within the visualization scene. These properties range from the standard affine transforms applied to meshes, to the speed or rendering component should be updated at an interval conducive to smooth animation and high user interface responsiveness (e.g., 30 times a second or greater). Ideally, the simulation solver will run in a thread separate from the graphics rendering to ensure the highest precision simulation possible (i.e., have the simulation time slice not be limit ed by a graphical refresh rate).
71 The final required components pertain to user interaction. For simulation and visualization design, there will need to be a designer user interface For consumption a, viewer user interface will need to be created. The d istinction between these interfaces is defined in Section 3.2.9. The optional ontology pruning component (see Section 3.4.4) will interface with the ontology importer and allow users to hand pick concepts to be placed in the 3D design environment. The op tional semantic styling component ( see Section 3.2.10 ) interfaces with the graphical renderer and dresses the basic ontology representation with graphical objects used in abstract simulation modeling in a given domain. The entire architecture described i n this section is illustrated in Figure 3 12. In the architecture, the ontology component serves to store the current state of the ontology in memory. 3.4 Methodology In this section, the high level methodology is presented. The methodology is defined t o provide a ste p by step account of how t he theoretical tool defined earlier in this chapter could be leveraged by a designer. There are three primary steps to the methodology: ontology acquisition, modeling building, and execution. 3.4 .1 Ontology Acquisition T o leverage an ontology visualization as an interface affordance, one must first obtain an ontology encoded with the desired semantics. Naturally, there are two ways to fulfill this requirement. The first is to seek out a pre existing ontology which may be published by experts in the given field. For example, there is the Modelica ontology (Pop et al., 2004) that supports simulation of physical systems (e.g., those containing electrical, thermal, or hydraulic components). Other pre existing onto logies include those that are intra organizational and often not publically accessible. These ontologies
72 enable knowledge sharing and reuse across departments within a single organization (Blomqvist and hgren 2008). The second option is to simply constru ct the ontology oneself. In turn, the newly constructed ontology can be used as a starting point in future efforts. If a pre existing ontology is employed, then it should be analyzed and modified if the semantics required by the modeler are not already pr esent. Examples of modifications include adding new sub concepts (i.e., adding semantic precession) necessary for the desired simulation and visualization, adding entirely new concepts, or adding relationships between pre existing concepts. 3. 4 .2 Simulati on Model Building and Visualization Construction To eventually execute a simulation and corresponding visualization, system dynamics need to be defined. To achieve this, relationships specific to simulation dynamics should be added between existing concept s in the ontology, if they are not already present. If more than one relationship type can denote dynamic flow, then the specific dynamics of interest need to be defined to avoid ambiguities when it is time to execute the model. Concurrently with simulati on model building, expressive 3D visualization can be created by adding visual parameters as attributes to ontology concepts. Examples of visualization attributes include 3D meshes and materials. When a visualization attribute is added to an ont ology conce pt, it should be ren dered in the ontology visualization and rooted at the node of the given concepts. Further, influences can be added to ontology members to link visualization attributes to simulation variables (e.g., attribute A in fluences attribute B, w here A is a dynamic variable and B is a visualization parameter).
73 As a final construction requirement of the methodology, the graph based ontology visualization itself can also be modified to express certain ideas. This requirement supports integrative mu ltimode ling. For example, the visualization of a relationship denoting dynamic flow can be morphed into a curve that traces the path of the flow within the 3D visualization environment. 3. 4 .3 Simulation Execution and Visualization Animation A correctly fo rmed model can be executed once the needed ontological concept s are established, and the simu lation model and visualization attributes are defined within the ontolog y. The method of execution will depend on the mathematical paradigm in which a given domain ontology exists For example, in some physiol ogy simulation models, blood flow is represented by the hydrodynamic metaphor where blood volume in compartments is updated by calculating pressures, resistances, and valves between adjacent compart ments. The execution requirement implies that the specific solver is connected to the interface and is coded depending on the desired application domain. As the defined simulation model is executed, simulation parameters should be updated at each time inte rval. In turn, visualization parameters that are linked to simulation attributes through influences are updated. This results in a real time 3D animation based on simulation output. 3. 5 Classification, Comparisons and Limitations 3. 5 .1 Classification In Ta ble 3 3 the proposed usage of ontologies in the interface is codified according to the classification system presented by Paulheim and Probst (2010). The ontologies that are well suited for the proposed methodology lev semantics but th ere is no semantic complexity requirement on them. The visualized
74 ontology can be leveraged during design time (e.g., to build a dynamic model) or run time (e.g., to label components and denote relationships). Further, ontologies are visualized as graphs, in the proposed methodology, interaction is allowed as a means to change which ontology members are visible and to modify the ontology within the visualization environment. 3. 5 .2 Compari sons A core p rinciple of integrative multimodeling is that the synthesis of multiple models may support new visualization and interaction techniques beneficial to the areas of design and education. This synthesis can be achieved in the human interface la yer, as is the case with two notable integrative multimodeling efforts: the work of Park and Fishwick (2005) (see Section 2. 3.2 ), and the work of Quarles (2009) (see Section 2. 3 .3 ). to further articulate the proposed methodology, the remainder of this sect ion presents a comparison of these previous works against the proposed ontology centric approach, as well as a classification of the particular usage of ontologies the proposed methodology will employ. Park and Fishwick (2005) developed an integrative mult imodeling framework that leveraged ontologies. They created a holistic ontology composed of classes pertaining to concepts in simulation, visualization, and the domai n of interest (e.g., military, chemistry). Portions of the ontology were represented in the interface as lists. This enabled users to fill in some required information (e.g. of model: Functional Block Model or Finite State Machine). However, the ontology was not displayed within the visualization environmen t, leaving the integrative visualization to be composed solely of graphical an d dynamic models. The proposed methodology
75 builds on this framework by placing the domain ontology with in the integrative visualization along with the dynamic and graphical mode ls. I believe this placement represents the most important aspect of what is presented in this dissertation : the ontological structure can serve as the foundation for a new interface model supporting simulation model building and visualization design activities. The Park and Fishwick framework provides an environment for model blending (i.e., the blending of graphical models with dynamic models), but model design and execution required the traditional engineering edifice such as Blender Game Engine logic blocks and Python scripts. Quarles (2009 ) showed the Augmented Anesthesia Machine ( AAM ) to be effective in some areas of training. However, no formal method was provided to create similar style visualizations for different p rocesses. The proposed methodology is a first step toward defining a conceptual framework for building tools such as the AAM. With domain ontologies manifested in the visualization environment, dynamic models can be constructed by supplementing the domai and visualizations can be built by adding visual attributes to the members of the ontology. The 3D collocation required by the AAM will come naturally in the proposed methodology because the abstract knowledge ( located in the ontology) and dynamic model are used as interface affordances and, therefore, are manifested in the 3D visualizat ion environment from the start. 3. 5 .4 Limitations It is important to consider the potential limitations of the theoretical framework described in the previous sections. Thus far, I believe there are two major limitations : one pertain ing to parsing and visualizing large pre existing domain ontologies; and
76 another pertaining to simulation complexity of models outside of the r ealm of equation based and continuous. Domain ontologies can grow quite large. For example, the Foundational Model of Anatomy (FMA) o ntology contains over 75,000 concepts and over 2,000,000 relationships While there is active research in automated and pa rtially automated large graph layout (e.g., see the work of Diaz et al. (2002)) and there is research in to what makes a graph easily comprehendible ; large graph structures are still widely considered network (von Landesberger et al., 2011) U sers of the proposed framework will need to delve into deep semantic detail to build executable models and visualizations with a desirable level of precision. Due to these restraints, there will most likely need to be a pruning phase in which a subset of semantics are chosen from the larger ontology, and then this subset is imported into the framework. The semantic pruning phase can be handled in many ways. External ontology editing tools (e.g. Protg ) could be used to edit a large ontology down to the desired concepts and relationships, or the framework coul d be mo dified to include an ontology representation that is not 3D. The latter option may be more desirable becau se pruning could occur at any time in the design process and within the same interface in which the design activities are taking place. The 2 D text based hierarchal list (leveraged in Protg and the FMA online viewer) could provide a clean interface into a vast ontology, A high level sketch of such an interface is depicted in Figure 3 1 1
77 Regarding simulation model complexity, there are certain domain modeling paradigms that map cleanly to the base ontological graph structure (e.g., compartmental modeling, system dynamics, finite state machines). These model ty pes are often flow based. There are less constrained model types, however, that do not have such a straight forward semantic map onto an ontological graph. One such hard to map model type is the rule based model Consider the simulation and visualization of a manufacturing assembly line. In manufacturing simulation, there are many factory components driven by different rule sets Further assembly line visualization can be very intricate. For example, there may be conveyor belts, robotic arms and human workers The dynamics between h uman worker s and other objects are particular ly difficult to animate because the human may interact with another object changing the position and orientation of the object as the human holds it in their hands Complex, mult i model environments such as the manufacturing assembly line may be best served by o ther simulation platforms that are coded with the appropriate functionally and graphical representations 3.6 Summary In this chapter, an interface theory was presented along with a framework to house the theory and define a software tool that could be implemented. The central structure of the proposed theory is a visualized 3D ontology. The three main structures of the ontology (concepts, attributes, and relationships) all play an important role in the theory. These roles were detailed along with 11 semantic and structural affordances of the visualized ontological structure with respect to simulation modeling and visua lization design. Further, a methodology was provided to describe how a designer would use the theoretical tool. Finally, the tool was classified using the Paulheim and Probst (2010)
78 scheme for describing the role of ontology in the user interface and th eoretical limitations of the tool were discussed.
79 Table 3 1. A l ist of suggested attributes for assigning the influence between simulation variables and visualization parameters Attribute name Attribute value i nfluence source The name of the simulation attribute to drive the influence. influence source min The minimum source value permitted for the influence. This value can be scalar or 3D vector. Values less than the min value should be clamped to the min value. This attr ibute must influence source max The maximum source value permitted for the influence. This value can be scalar or 3D vector. Values greater than the max value should be clamped to the max value. This attribute must be of the t influence destination The name of the attribute which will be influenced influence destination min The minimum destination value permitted for the influence. This value can be scalar or 3D vector. Values less than the min value should be clamped to the min value. This attribute must influence destination max The maximum destination value permitted for the influence. This value can be scalar or 3D vector. Values greater than the max value should be clamped to the max value. This influence bounds The type of bounding geometry to define the region of influence. i nfluence radius The scale of the bounding geometry. This attribute must be of the i nfluence falloff The type of falloff calculation applied to influence vertices based on their distance from the central influence point (i.e., the 3D location of the parent concept).
80 Tabl e 3 2. A l ist of suggested attributes for assigning particle systems properties to ontology concepts and relationships. Attribute name Attribute value particle emission The numerical value or name of the variable to determine the rate of particle emission from the source concept. particle emission factor A numerical value that can be used to scale the particle emission. This attribute must be of the particle color The color of the particles. This attribute m ay be linked to other attribute s value s either within the relat ionship, or in source and destination concepts. If the color value is to be hard coded, particle image An image used to render a single particle. This image may be tinted by the value of particle color particle speed The speed at which the particle is animated (e.g., travels along a relationship curve). This value can be provided in unit lengths per sec ond, where the unit length is determined within the 3D environment. particle random offset A numerical value to serve as the upper bound of a random offset of a particle from a predefined path. A random unit vector can be calculated and multiplied by a ra ndom value from the range [0, < particle random offset >] to find the 3D position of the offset from a point along a relationship curve. p article wander interval A numerical value to represent the time, in seconds, that one random point to another in the region defined by the particle random offset. Table 3 3 The characteristics of ontologies well suited for use in the proposed framework. This classification scheme was defined by Paulheim and Probst (2010). Characteristic Classification Domain Real world Complexity Informal, low, medium and high Usage Design time and run time Visualization Graphical and verbalized Interaction View and edit
81 A B C Figure 3 1. A concept Bunny with graphical attributes of a 3D mesh and a 2D image (xray.jpg) A) A sketch of a concept with a mesh and image attribute. B) A translation widget appears rooted at the conc ept when the mesh is selected. C) The translation of a concept when an axis of the translation widget is grabbed and dragged The mesh is transformed along with the concept and the image maintains its 2D offset.
82 Figure 3 2. A high level sketch of a user interface to meta attributes. A B Figure 3 3. Two different syntax approaches for accessing attributes across multiple concepts within equations.
83 Figure 3 4 A simple 3D scene with an ontology visualization serving as an interface to mesh transformations The relationship between Table and Tea Pot is has a and therefore when the table is transformed, the tea pot will be as well treating the Table concept location as the origin point. Figure 3 5 A simple illustration of using a has a relationship to form vertex groups. The H ead concept is assigned a sphere as bounding geometry. This sphere was then transformed by a designer to encompass the desired set of vertices.
84 Figure 3 6 A depiction of a particle traversing a relationship curve In this example, the particle speed is determined by interpolating between the sp eed attribute assigned to the source and destination concept. Figure 3 7 T he action of expanding an attribute. The attribute image represented by the image file name and a thumbnail in the attribute list, is dragged in to the 3D environment and then represented by a higher resolution version of the image that maintains its own visual link to the concept (the edge connecting the imag e and the concept).
85 Figure 3 8 parallel to the view plane that should hold all expanded attributes of a given concept. Figure 3 9 A simple example of applying a grap h style sheet to an ontology to create a graphical representation used in systems dynamics modeling. The concepts shown above should contain domain appropriate semantics to be interpreted by the style sheet. The system dynamics model contains a source (source co ncept), a flow (the relationship), a rate (the hour glass icon), and a level (destination concept).
86 A B Figure 3 10 Different approaches to morphing the ontology structure used in the A ) A sketch of t he morphing of an ontology supporting meta attributes to a standard ontology with just concepts, relationships, and attributes. B) A sketch of the morphing of an ontology supporting relationships with attributes, to a standard ontology. Figure 3 11 A sketch of a potential ontology pruning interface where concepts can be dragged into the 3D design environment from a 2D list based view.
87 Figure 3 1 2 A high level diagram depicting the interfaces between the various required components of the proposed f ramework.
88 CHAPTER 4 CASE STUDY A software prototype of the theoretical tool defined in Chapter 3 will be presented in this chapter. To demonstrate key functionality of this prototype, a simulation and visualization was constructed that showcases the impo rtant features. To create this case study t he Beneken (1965) cardiovascular model was chosen as the base simulation model from which to build an interactive integrative 3D visualization of the human cardiovascular system. In Section 4.1, the model will be described in detail lineage and reasoning for why this model was chosen The software prototype and the construction of the simulation and visualization are presented in Section 4.2. In Section 4. 3 preliminary human fa ctors are considered including an example lesson based on the resulting simulation and visualization that was created by a medical educator. 4.1 Cardiovascular Modeling Many credit Euler (1775) with the creation of the first mathematical model to simulate blood flow. Weber (1850) is often credited for creating the first circulation model to employ the hydrodynamic metaphor ( i.e ., using valves and pumps). However, t his introductory discussion will be limited to the history of modeling efforts in c ardiovascular physiology in the computing age. One of the earliest examples of leveraging formal modeling techniques and computer technology to simulate a physiological process is the work done by Warner (1959). He employed the power of the electronic an alog computer to create two novel simulations: one of the carotid sinus (an area of bifurcation in the carotid artery), and one of the circulatory system. The latter is worthy of further explanation due to the
89 implementation of a compartmental model and t he adherence to a simulation design Warner is shown in Figure 4 1 A set of equations are used at each block to solve for blood pressure and blood volume. Because this model was implemented on an analog computer, each equation had to be built as an electric circuit and the continuous outputs of these circuits were fed into oscilloscopes for visualization. 4.1. 1 The Beneken Model Soon after Warner published his analog computer compatible circulation model, Beneken submitted a compartmental hydraulic metaphor model as his PhD thesis (1965), which was also primed for analog computation. Beneken found the functions that drive the beating of the heart in the Warner model to be probl ematic because they transition too abruptly from systole (contraction) to diastole (relaxation). As a result, the Beneken model contains a smoother cardio controller function. The model Beneken created is shown in 4 2 The typical set of equations for a compartment in the Beneken model account for blood pressure, flow, and change in volume. The pressure in a compartment at time is defined by, (4 1) where is the elastance of the compartment (a time variant function for compartments representing heart chambers, a constant for the other compartments), is the volume at time of the compartment and is the unstressed volume of the compartment. Th e flow into a compartment at time is defined by
90 (4 2 ) w here is the pressure of the compartment at time is the pressure of the input compartment at time and is the resistance between the compartments. The change in volume in a compartment, at time is defined by, (4 3 ) where is the flow into the compartment, and is the flow out of the compartm ent There is also influence from inertia between the concepts intrathoracic arteries and extrathoracic arteries. For a complete description of the remaining equations and 4.1. 2 The Goodwin et al. and S C outo et al. Models The Beneken model is still used today as a foundation for new models used in simulations for training scenarios. One such model is for infant cardiovas cular physiology Goodwin et al. (2004) adult circulation model to estimate the circulation of an infant. As part of the tuning, f or example, the total blood volume was decreased from 4740 milliliters to 685 milliliters and the heart rate parameter was increased from 72 beats per minute to 129 beats per minute. Other coefficients, such as arterial and venous resistances, were changed as well. Also using the Beneken model as a mathematical foundation, S Couto et al. (2006) built models representing four pathologies. Many pathological models share the same compartments as the base physiological model and differ only in connectivity of
91 compartme nts and placement of resistances and valves. The four pathologies modeled by S Co uto et al. are shown in Figure 4 3 4. 2 Case Study 4.2.1 Software Prototype A software prototype needed to be created to demonstrate the proposed interaction theory using a real world example The prototype was coded in C++ and leverages the Open source Graphics Rendering Engine (OGRE, 2012) for managing the rendering of textured 3D geometry. OGRE was chosen because it is open source, leverages either the DirectX or OpenGL hardware accelerated graphics libraries and supports the powerful and fast program m ing language C++. The prototype runs on computers using the Windows 7 operating system. A depiction of the architecture of the softwar e prototype is shown in Figure 4 4 This architecture is one possible instantiation of the high level architecture presented in Chapter 3. In the prototype model solver because this is required to execute t he Beneken model. Another implementation choice was to encode the ontology in the XML format (notice the XML importer and exporter components in Figure 4 4). XML was cho sen because of it s universal adoption and because it can be easily understood by future researchers continuing work on this topic. A screenshot of the created prototype is shown in Figure 4 5 with interface components labeled. environment and selected so a translation w idget is displayed rooted at the concept. The S imulation C ontrol allows users to play, pause, reset, and change the time scale of the simulation. The time scale ranges from 0.01 (one one hundredth real time) to 1.0
92 (real time). The I mporter imports onto logies encoded in the XML format. Conversely the E xporter exports ontologies from the prototype into the XML format. Concepts are created using the concept creator by entering the concept name in the text field and then pressing the s no relationship creator because relationships are created in the prototype by the user holding the left mouse button down on a concept and then dragging to another concept and releasing. Relationship types are determined through an attribute name d type assigned to relationships automatically upon creation. The V isibility S ettings control the visibility of concepts and relationships. Within the visibility settings are : particle visibility that dete rmines if particle s are rendered; knowledge vis ibility settings that determine i f the ontology structure is visible ; a show all option to determine if concepts are visible even if they have no visible adjacent relationship s ; and a list of current relationships in the ontology with the option to set a given r elationship type visible or change the color coding of the relationship type The A ttribute E ditor displays all of the attributes of a selected concept (e.g., the selected 5) along with the meta attributes of the selected att ributes (see the right portion of the attribute editor) and allows for editing of the attributes and meta attributes by a designer 4.2.2 Solving the Compartmental Model The simulation solver used in this prototype is tailored to solve the Beneken model. With more development effort, this solver could be made to solve any compartmental model. For this case stud y development effort was spent towards showcasing the design affordances of the visualized ontology and to create a complete interactive, integrat ive visualization of the human cardiovascular system, as oppo sed to
93 creating a general tool. I consider the creation of a more general tool a promising direction f or future efforts (see Section 5.2 ). The implemented solver looks for the required semantic s as attributes attached to concepts and relationships in the scene. If these semantics are present when S imulation C ontrol, the simulation will execute and, in turn, the visualization will animate. The required attributes are sh own in Table 4 1. The solver uses these attributes as coefficients to solve the necessary differential equations of the Beneken model using the forward Euler method. The solver runs in a thread separate from the main graphics rendering and user interface thread with a time step of one millisecond. 4.2. 3 Constructing the Executable Simulation and Visualization In this section a step by step account of the process of constructing an interactive integrative visualization of the human cardiovascular system is presented. This visualization will include a heart with four chambers that according to the value of the simulation variable volume Particle systems will also be attached to the relationships of the simulation model to illustrate the dynamics of blood in the arteries and veins. The construction of the visualization will demonstrate the interaction a designer can have with the ontological structure in support of design activities. To fully relay the interaction and visual dynamics involved in the design process described in this section a video is provided as Object 4 1 Object 4 1. Simulation and visualization design demonstration video (.wmv file 120 MB).
94 126.96.36.199 Constructing the b ase m odel The base executable Beneken model is constructed in the prototype by adding the 10 required concepts (i.e., compartments in the compartmental model) using the concept creator, and connecting them with flows to relationships denoting the path of blood flow. The flows to relationship type must be used because the solver looks for this specific relationship type to solve the model. In the prototype, relationships denoted as flows to are automatically styled with a peach color, but any relationship can be colo red with any color by assigning an attribute presentation color to the relationship and marking it as a variable that is of the vector type to provide red, blue, and green values. A snap shot of constructing the base model is shown in Figure 4 6. This sna p shot was taken as the designer adds a relationship between the P ulmonary A rterial T ree and the P ulmonary V enous T ree concepts. This model was constructed through collaborations with a medical educator, wit h the end goal being to deploy the constructed visualization in a classroom (see Section 4.3 ). The educator suggested changing several concept names to be consistent with current teaching terminology. This accounts for the difference in some of the concept names when comparing the model depicted in Figure 4 6 to the model depicted in Figure 4 3. Further, abbreviations are used in the prototype to minimize visual clutter. If a concept has an attribute named abbr then the value of this attribute will label the concept in the visualization environmen t (e.g., abbr = RV is assigned to the Right Ventricle concept ). Figure 4 7 shows the completed Beneken model with all the required concepts, relationships and attributes added. The Left Ventricle concept i s selected and its
95 attributes are rendered as a list in the Attribute Editor. The required attributes of simulation concepts to properly solve the model are enumerated in Table 4 1 Some of the attributes are shown in the Attribute Dock in Figure 4 7 (flow, volume, pressure, elastance m in, elastance max). One technique for the efficient creation of these attributes is to : create a super ; add the required attributes to this concept ; and then attach this concept to the 10 concepts of the Beneken model wi th an is a relationship. This technique employs the taxonomic approach discussed in Section 3.2.3. A distinction should be made on the is a relationship so just the attribute types and names are propagated down and not the attribute values as all 10 conc epts require different values for these attributes. A way to achieve this implementation detail is to leave attributes blank whose va lue should be over written One could imagine, however, additions to the theory where attributes are added to the is a re lationship to precisely define the nature of the attribute propagation. Once the Beneken model is made executable, dynamics can be explored. Figure 3 8 shows the result of expanding the attribute pressure that is attached to the Left Ventricle concept. presentation space (see Section 3.2.8 for a full definition of attribute expansion and the display space ). In this case study variable attributes are rendered as numerical values when docked, and interactive plots when expanded. Expansion is done by the user by pressing the left mouse button on the gray pad to the left of the attribute names, and dragging this value into the 3D scene, and then releasing. Once the expanded display (the plot in this case) is in the scene, its 2D offset from its parent concept can be
96 changed by dragging the displ ay with the right mouse button. T he display may be 188.8.131.52 Adding dynamic v isualization Once the simulation model is defined and its execution provides output that is deemed correct, visualization elements can be added to the scene to illustrate behavior of certain simulation variables. This is achieved by adding visual elements as attributes to the appropriate concepts, and then linking these visual properties to simulation variables through influence attributes. The visualization to be created for the primary case study shall include a beating heart and particles syst ems attached to the relationships between simulation concepts to illustrate the dynamics of blood flow. To begin constructing the visualization, a mesh of the heart needs to be placed in the scene. To achieve this, a Heart concept is added to the scene to hold the mesh. A mesh attribute is added to this concept and assigned a value of heart.mesh and a material attribute is added to this concept and assigned a value of hear t _new_mat Within the prototype, meshes and material must conform to the OGRE stan dard of .mesh and .material files. A 3D heart mesh was purchased online and converted to the .mesh format using the free 3D modeling and animation package, Blender. The material file heart_new_mat contains a simple definition of material properties that account for the red shading of the heart. This simple coloring may also be achieved by adding an attribute of color to the concept. The result of adding the Heart concept and the two visualization attributes ( mesh and material ) is shown in Figure 4 9. Af ter the heart mesh is added to the scene, it needs to be semantically linked to the simulation model to achieve the beating effect. This will require linking the heart to
97 its four chambers, all present in the Beneken model. These chamber concepts are the Left Atrium, Left Ventricle, Right Atrium and Right Ventricle. For demonstration purposes, the process of rigging the right atrium of the heart will be detailed. To start the process of rigging, the Heart is linked to the Right Atrium concept through a has a relationship. The designer then positions the Right Atrium concept near the perceived center of the right atrium of the heart mesh. The positioning of the Right Atrium concept is depicted in Figure 4 10. There is a collection of vertices in the hea rt mesh that approximate the region of the right atrium. It is this vertex group that should be deformed according to the volume attribute of the Right Atrium concept to create a realistic animation of a beating heart chamber. A vertex group can be create d in the prototype by adding a set of influence attributes (introduced in Section 3.2.5) to the Right Atrium concept. When all required influence attributes are present, the vertices bounded by the influence are highlighted. This is depicted in Figure 4 11. A sphere is used as the default bounding volume in the prototype, but as discussed in Section 3.2.5, more complex geometry could be used to bound sets of vertices. Quadratic falloff is employed so vertices closer to the boundary of the sphere (i.e., further away from the concept node), will be influenced less. This creates smooth boundary conditions for influence regions and, therefore, no hard edges will be created when the heart mesh animates. Other types of falloff could be employed and potential ly defined through an influence falloff attribute. Table 4 2 lists all the attributes that are added to the Right Atrium concept to achieve the influence.
98 4.2.3. 3 Sculpting r elationships Within a cardiovascular visualization, curved relationships can trace the path of blood flow and also serve as the basis of more advance d features. One feature is the ability to create an animation path for particle systems that not only traces blood flow, but visually demonstrates the oxygenation of blood, and the pu lsation of blood through the veins and arteries. Using ontological relationships to represent an abstraction of three dimensional flow requires relationships to be renderable as curves. To satisfy this requirement 3D spline functionality was incorporated into the prototype Splines are smooth interpolations of piece wise linear polynomials (Bartels et al., 1983). To create relationship splines, designers add control points to a control hull of a relationship, and then translate these control points to i mmediately see the resulting curvature. This process is depicted in Figure 4 12, where the curvature of the flows to relationship between the Right Ventricle and Pulmonary Arterial Tree is defined. To create the complete visualization, all flows to relationships of the Beneken model should be sculpted. In Figure 4 13, the entire model is shown co located with the heart mesh. When implementing such a spline based system, careful consideration must be paid to the frame of the curve. The frame of the c urve is defined by the tangent, normal, and bi normal vectors of the curve at any point ( Yamaguchi, 1988 ) The frame is important because it is used to define the three orthogonal axes required to calculate any offset from the curve in 3D space A simple illustration of the frame is shown in Figure 4 14 A O ffset calculated using the frame can be used to give a relationship or to define the offset of a particle within a particle system Fra mes must stay oriented correctly across the curve to avoid any sudden jumps in the normal or bi normal. S udden jumps in the frame s
99 orientation will result in a twisting of a curve with any width, and the instantaneous jumping of offset particles from one side of the curve to another. In the prototype, curved relationships are given width by extruding a square along the underlying path, which consists of a set of solved spline points. To keep the frames from twisting (i.e., keep the square extrusion from abruptly twisting around the solved path), the technique of rotation minimizing frames (RMF) is employed. With this technique, the tangent vector is calculated as the normalized displace ment vector between adjacent solved points (Wang et al., 2008) A depiction of the tangent vectors between a series of solved points is shown in Figure 4 14B. The minimum rotation to transform one tangent vector to the next is determined, and this rotatio n transform is applied to the previous normal and bi normal vectors to calculate the next normal and bi normal vectors An example of the extrusion twisting and the RMF fix is shown in Figure 4 15. 184.108.40.206 Creating particle s ystems Particle systems are a dded to sculpted relationships to visualize the 3D path and dynamics of blood flow and show the deoxygenation of blood between the extrathoracic arteries and extrathoracic veins. Section 3.2.7 introduces the idea of leveraging the ontology structure as a n interface affordance for creating and editing particle sys tems A particle system is added to each flows to relationship P article system s emit particles from the source concept animate the particles along the curve of th e relationship, and remove the particles when they reach the destination concept. The particles within the particle system are each a billboarded (i.e., always oriented towards the virtual camera ) textured plane. The plane is t extured with the value of the particle image attribute ad ded to the relationship. Figure 4 16 shows the image cell.jpg that is assigned to the
100 particle image attribute to texture the particles. The image is gray scale and may be tinted with a color by assigning a particle color attribute to the relationship. A list of all particle attributes assigned to the 10 edges of the Beneken model is provided in Table 4 3. Particle attribute values may be bound to values of attributes assigned to both the source and destination concepts. Figure 4 17 shows the result of binding the particle color attribute of a relationship to the attribute presentation color in the source and destination concept This results in a linear interpolation of the particle color as it traverses the edge. Figure 4 18 shows the r esult of adding particle systems to all 10 relationships. What is depicted in 4 18 represents a completed module envisaged by a designer. 220.127.116.11 Defining the viewer m odule A designer often has a certain audience in mind when creating visualizations. Given this audience, it is up to the designer to choose what interactivity should be permitted in the final visualization module, and what attributes should be visible. Within the proposed framework, many attributes will be design oriented (e.g., consider the influence attributes listed in Table 4 1), and not important to an end consumer. For the case study, only the simulation variables volume pressure and resistance are set resistance attributes are made adjustable and given a range of 0.01 to 2.0. A snap shot of the designed visualization 19. The prototype vie wer provides a subset of the functionality of the prototype designer. The distinction between the two modalities in described in Section 3.2.9.
101 4.2.4 Extensions for Hypovolemic Shock With a base cardiovascular model defined, one can explore augmentations that can be used to simulate different pathophysiology or abnormal conditions. To this end a simulation and visualization of hypovolemic shock was constructed in the designer prototype by starting with the base executable simulation model defined in Se ction 18.104.22.168 Hypovolemic shock occurs when there is substantial loss of blood volume in the veins. The shock simulation model contains the following two extensions from the base model : an electrocardio gram (ECG); and a mapping of blood volume loss to heart rate and the strength of the heart beat. The ECG was added as an attribute to a Heart concept. The equation used to create a realistic ECG signal was defined by McSharry et al. (2003). Table 4 4, taken from Lawrence et al ( 2006), demonstrates the symptoms of hypovolemic shock. To add the effects of this type of shock to the simulation model, code was added to the compartmental model solver of the prototype. T his type of shock is represented by a conditional function with the percentage ranges def ining the condition. because the equations as attributes feature (discus sed in Section 3.2.2), does not s upport conditional formulations in its current state. The two influences in the hypovolemic shock case s tud y are shown in Figure 4 21 The vo subcomponents) is linked to the scale. The volume of the heart in the range (200ml, 600ml) is mapped to a scale in the range ([1.0,1.0,1.0], [1.3,1. 3,1.3] ). This results in a beating effect that can visually relay the pulse frequency and strength. As also depicted in Figure 4 21 total blood volume is mapped to the color of the skin. The blood volume
102 influence ranges from 4240 ml ( a normal amount of blood for an adult m ale) to 2500 ml. The skin color ranges from [red: 0.83, green: 0.82, blue: 0.81] (very pale) to [red: 0.86, green: 0.72, blue: 0.63] (a typical Caucasian skin tone). The result of these influence additions are shown in Figure 4 22 When the shock simulation is executed, t he equations in the model are solved and total blood volume is used to scale the ECG output (much like it is used to color the skin of the human body) This linkage results in a wea kened ECG signal as blood is drained Clearly, a medical domain expert would desire a more sophisticated and accurate linkage of physiological parameters to ECG output, but this example demonstrates how using equations as attributes can enab le executable ontology based si mulation models that c ross mathematical paradigms. 4.3 Preliminary Human Considerations A theoretical framework was described in Chapter 3. Future work that builds on this framework should be informed by user feedback because the ultimate goal is to make visualization construc tion and consumption simpler for the end user compared to current techniques. Two types of feedback were sought out: student feedback on the ontology enabled visualization consumption interface; and expert feedback on ontology enabled visualization constr uction interface. In support of both surveys, an example lesson plan was created through collabo rations with a medical educator, Dr. Juan Cendan of the University Of Central Florida College Of Medicine who provided the introductory narratives and questi ons for the lesson. The lesson plan presents two real world scenarios, and then guide s participant s through the functions of the viewer prototype loaded with the Beneken executable simulation and visualizatio n of human cardiovascular system The lesson p lan serve s
103 and also as a learning module. The complete lesson plan is presented in Appendix A. The lesson plan was encoded in a webpage and rendered in a side panel of the software prototype, as depicted in Figu re 4 23. For the purpose of the study, the prototype software was referred to as KOG (Knowledge, Ontology, Graph). 4.3.1 Student Survey Second year medical students at the University of Central Florida were presented with the option to participate in a study that compares two modules for teaching cardiovascular function. Students that participated in the study were given a pre test assessing their general second year medical knowledge, randomly given one of the two learning modules with built in lesso n plan, and then given a post test which contains the same questions as the pre test but in a different order. No time constraints were placed on participants and time spent with the module was not tracked. Study enrollment was low, so a full statistical comparison of the modules was not viable P resented in this section are the responses to the open ended survey questions of the four students that performed the study with the KOG viewer. The results are as follows: Q1: Do you find this educational software to be different from tools you have used in the past to learn physiology? All four students Q2: If you found this educational software to be different, please elaborate on the differences Student 1: It is the most interactive form o f learning and encourag es self directed exploration. There aren't very many other ways to explore all of the multiple ways you can view the heart's functions (PV loops, EKG, BP...) when you affect flow or HR. Student 2: Using this technology required a kno wledge of what the components in the loop were. Higher level of understanding of the anatomy was required. Student 3: More interactive
104 Student 4: It was very interactive. I like the possibility to see how one parameter affects the other. Instead of readin g the text I could easily see the relation. Q3: Do you believe this educational software can provide any benefit to the learning experience? If so, please elaborate. Student 1: Yes! Student 2: Yes, these tools are good to use if given appropriate scenario s Student 3: Yes, but I would have gotten more out of it if this was a required SLM [Self Learning Module] or information was to be used for testing purposes because I probably would have spent more time with it. Student 4: Yes, it is helpful. I would fo llow up the quiz questions with quick explanation though. Q4: Please provide any additional feedback you may have on the content or technology that was just presented to you. Student 3 : I liked the technology and content. However, I didn't think the pre and post test questions were reflective of content that would be learned with this presentation. Student 4: The first time I used it was a little confusing because there were so many instructions. The second scenario was much easier and faster to navigate. Once you go beyond learning how to use the software it becomes very helpful. There are several observations of the student response I think are worth further consideration. components igher level of understanding of the anatomy was required knowledge was explored by Quarles et al. (2008) with respect to anesthesia machine function. The viewer prototype could be used to run a similar study with respect to
105 anatomy and physiology. Such a study would explore any benefits or downfalls of the integrative visua lizations constructed with the proposed approach. Secondly, all four students reported the tool could benefit the learning experience, but three out of four mentioned that for the tool to be useful, it needs to be integrated into pedagogy by an instructor. This early feedback suggests that, at least in the context of medical education, these types of simulation based visualizations are likely not to be leveraged in an exploratory manner by most students. ck, it was claimed that they learned the interface aft er performing just one scenario, t he second scenario was much the labeled ontology served as an interfa ce that could be learned quickly and permitted fast, intuitive access to desired information (in the form of attributes). From a perspective of visualization consumption, the 3D ontology graph could prove to be a highly effective interface. To make this c laim, a new study will need to be performed. This is a challenging task, however, because there are no current standards employed by educational visualizations that could be compared to the ontology based method. One potential approach is to collect a la rge sample of visualizations (possibly freely available online) that explore the same basic phenomen on and have a group of students p erform a lesson with each module, including the ontology based approach 4.3.2 Expert Survey Five experts took part in a survey that aimed to assess the prototype viewer and also the design theory employed by the proposed framework to be someone working in medical education as an instructor or engineer in a medical research lab that would potentially l everage visualizations in their work. Each expert
106 was presented with a 12 minute video demonstrating the design features of the proposed framework. In the video, the process of constructing the Beneken beating heart visualization is illustrated. The vid eo is provided as Object 4 1. After watching the video, the expert was given a link to download and install the same prototype viewer provided during the student survey (see Section 4.3.1) The time participants took with the prototype viewer was not trac ked. Three medical faculty members and two engineers from research labs took part in the survey. The template expert survey provided by Ravden et al. (1989) was used and the complete results are presented in Appendix B. In this section an overview of the responses will be given along with an analysis The overall expert response to the presentation (i.e., the video and viewer prototype) was promising. It was stated that there is a lot of versatility in the animation system and and that there are An engineer referred to the These positive responses are encouraging, but the major reason f or performing this survey is to seek out valuable criticism from the experts. Much of the critical feedback from the experts pertained to their experience with the viewer prototype, and not the design concepts presented in the video This is understandable because they were able to interact with the software prototype and had a more passive experience with the content in the video. I beli eve this is indicative of the challenge of trying to assess preliminary technology in non interactive form (e.g.,
107 paper mock ups, videos). There was, however, one important critique of the design theory presented to the experts. Two of the medical facult y noted that the design language in the video was not intuitive for non programmers. While technical programming concepts were not mentioned in the video, there were computer graphics ). This feedback specifically from the medical faculty suggests that even though all participants were able to effectively consum e the interactive visualization the process of designing may still remain predominantly an engineering practice. Findings from this survey inform a placement of the proposed framework within the current visualization construction pipeline. This placement is discussed further in Chapter 5 along with future directions for research. 4.4 Summary In this chap ter, a case study was detailed that demonstrates the theory presented in Chapter 3. Preliminary human considerations were also discussed based on the results of two surveys. The case study a simulation and visualization of the human cardiovascular syste m, demonstrates a series of steps to facilitate simulation and visualization construction by leveraging domain knowledge. A compartmental model is constructed and additional visualization semantics are attached to the model. There are other sequences of interactions carried out by the domain expert that can result in the same dynamic visualization. One could start with an existing ontology and augment this with simulation semantics and graphical objects. Conversely, the graphics (3D and 2D media) could s erve as a starting point and ontological structures, including dynamics, could be added. The approach taken will depend on the resources of the
108 domain expert and what co interaction between the three model types (ontologic, dynamic, and graphic) is envisi oned.
109 Table 4 1. The semantics required to re create the 10 compartment Ben e ken model of the human cardiovascular system. Attribute Parent Structure Description v olume Concept The blood volume in a chamber p ressure Concept The blood pressure in a chamber. f low Concept The out flow of blood in a chamber unstressed v olume Concept The unstressed volume of a chamber e lastance Concept The elastance of a chamber e lastance m in Concept The minimum elastance of a chamber of the heart e lastance m ax Concept The maximum elastance of a chamber of the heart i nertia Relationship The inertia value which should be placed between intrathoracic arteries and extrathoracic arteries. v alve Relationship The attribute that determine s if a relationship has a valve v alve open factor Relationship The factor in which to multiply flow rate if the relationship has a valve and the valve is open v alve close factor Relationship The factor in which to multiply flow rate if the relationship has a valve and the valve is closed Table 4 2 Influence attributes added to the Right Atrium concept to drive the animation of the right atrium portion of the heart mesh. Attribute Type Value influence radius Variable Scalar 0.43 influence source Basic V olume influence destination Basic S cale influence source min Variable, Scalar 0 influence source max Variable, Scalar 320 influence destination min Variable, Vector 0.7,0.7,0.7 i nfluence destination max Variable, Vector 2.2, 2.2, 2.2
110 Table 4 3 Particle system attributes added to flows to relationships Attribute Type Value p article emission Basic Flow p article image Basic Cell.jpg p article color Basic Presentation color p article emission factor Variable, Scalar 1.0 p article random offset Variable, Scalar 0.2 p article wander interval Variable, Scalar 6 p article random offset Variable, Scalar 0.2 Table 4 4. The effects of h ypovolemic shock as illustrated by Lawrence et al (2006). Blood loss amount Blood Pressure Pulse Skin Vasoconstriction Urine Output NL NL NL Flow 20% 30% 50%
111 Figure 4 1. A block diagram model created by Warner (1959) to simulate blood circulation Figure 4 2. The Beneken (1965) compartmental model for blood flow.
112 A B C D Figure 4 3 Pathology models by S Couto et al. (2006) See Figure 4 2 for a key of symbols and compartment names. A) Patent ductus arteriosus. B) Tetralogy of Fallot. C ) Coarctation of the aorta with patent foramen ovale and a small pate ncy of the ductus arteriosus. D ) Transposition of the great arteries with septal defects.
113 Figure 4 4 The architecture of the software prototype created to demonstrate the proposed theory. This architecture is an instantiation of the high level architecture presented in Chapter 3. Figure 4 5 A screenshot of the prototype. The yellow text labels the different components of the user interface selecte
114 Figure 4 6 A snap shot of building the Beneken model. The designer adds a relationsh ip between the Pulmonary Arterial Tree c oncept and the Pulmonary Venous Tree c oncept to denote blood flow. Figure 4 7. The complete Beneken model created with the software prototype. The Left Ventricle is selected so a translation widget appears roote d at this concept. All attributes of the Left Ventricle are also shown in the Attribute Editor.
115 Figure 4 8. The Beneken model executing in the software prototype. The attribute pressure of the Left Ventricle concept is expanded. Figure 4 9. The result of adding a Heart concept and assigning it a mesh and material attribute.
116 Figure 4 10. A snap shot of the prototype while a designer positions the Right Atrium concept to be within the heart mesh. The Heart is linked to the Right Atrium through a has a relationship, which is colored blue in the prototype. Figure 4 11. The influence of t he Right Atrium concept over the heart mesh Concepts and relationships not involving has a semantics are hidden in this view as determined by the settings in the View Settings interface (lower left)
117 Figure 4 1 2 Three snap shots of a designer formin g a curve from the flows to relationship between the concepts of Right Ventricle and Pulmonary Arterial Tree. Control points are add ed to the relationship to form a hull that defines curvature (left) The control points can be translated to change the r elationship s curvature (center, right). Figure 4 1 3 The complete Beneken model co located with the heart mesh in 3D. Concepts are color coded based on whether blood in the corresponding chamber is oxygenated (red is oxygenated, blue is deoxygenated).
118 A B Figure 4 14. The frame of a curve. A) An i llustration of the frame of a curve, including the tangent ( T ), normal ( N ), and bi normal ( B ) vectors of a solved spline. Solved spline points are connected by line segments. O nce normalized, these segments can be used as tangent vectors when calculating the frame of the curve at a given point. Given the tangent, t he rotation minimizing frame technique can be used to cal culate the normal and bi normal. A B Figure 4 15. The result of using the rotation minimizing frame technique. A) An arbitrary source vector (e.g., <0,1,0>) is used in the frame calculation to extrude a square along a curve to create thickness. An undesirable twist of the extrusion is highlighted. B) The same curve, without the twist, when the rotation minimizing frame technique is employed.
119 Figure 4 16 A 32x32 pixel image, cell.jpg, used in the creation of particles for the case study The black in the image maps to transparency. An image is assigned to a particle system through the particle image attribute. Figure 4 17. A particle system animates between the Extrathoracic Arteries and Extrathoracic Veins concepts.
120 Figure 4 1 8 The complete Beneken model co located with the heart mesh in 3D. Concepts are color coded based on whether blood in the corresponding chamber is oxygenated (red is oxygenated, blue is deoxygenated). Figure 4 19. A snap shot of the designed visualization rendered in the prototype viewer. A user dr ags a slider to manipulate the resistance between the extrathoracic arteries and extrathoracic veins.
121 Figure 4 20. A snap shot of the Beneken model co located with heart and human body 3D geometry. A flows to edge is selected and its control hull is m ade visible. Figure 4 2 1 An illustration of two influences added to the concepts Heart and Human Body within the shock visualization.
122 Figure 4 2 2 A simulation and visualization of hypovolemic shock over time. The less saturated and the heart beat weakens. Figure 4 23. The viewer prototype with lesson plan that was distributed with the student and expert surveys.
123 CHAPTER 5 CONCLUSIONS In this dissertation a new interaction theory was presented that connects simulation model building and visualization construction activities at the user interface level. The theory is built around an interactive 3D graph structure: a visualization of a n augmented ontology. The motivating claim for this work is : a three dimen sional visualization of the domain ontology can serve as a central user interface structure to connect simulation modeling and visualization construction activities and allows domain specific semantics to guide the interactive modeling process. Such an in terface may improve the simulation and visua lization construction pipeline by allowing simulations and visualizations to be construct ed in a more efficient manner compared to current techniques. The interface could also open simulation and visualization d esign and consumption to a wider audience. The ontology visualization with concepts, relationships, and attributes, can encode domain specific semantics in a format that is machine processable and also understandable by the domain experts. In the defined interface, t he visualization of an ontology as a 3D graph serves to anchor other visualization elements such as meshes and variable plots. Relationship visualizations can be sculpted to trace meaningful 3D paths within the visualization environment for pa rticle systems and other animated objects. To demonstrate the methodology, a software prototype was created. I n turn, a n executable simulation and visualization of the human cardiovascular system was created with the prototype. I believe this case study p resented in Chapter 4, along with the structural and semantic aff ordances presented in Chapter 3 make a strong case for the proposed interface an d justify future efforts toward expanding this work.
124 Research and development of the on tology based interaction theory and preliminary fee dback has led to a projection of how such a framework could be utilized in present day visualization construction efforts. Currently, if a domain expert wishes to create a visualization of a dynamic proce ss, they must work with engineers to build the visualization and create any executable simulation models. This process is d epicted in Figure 5 1 A A domain expert presents a design (represented by a paper sketch in the figure) to a visualization expert. The visualization expert builds the visualization by synthesizing a set graphical resource with executable interaction and simulation models. The visualization expert may need to employ an engineer with simulation expertise to fully implement the simulati on This additional expertise is represented under the generic label of in Figure 5 1 and manifests i n many forms (e.g., such as through pre coded modules found online, and through collaborations with a simulation specialist ) Also w o rth noting is that what is being one, two, or all three of the required skill sets for visualization construction. By l everaging the framework proposed in this diss ertation, the construction process could be improved by lessening the effort required by the computer engineering exp ert and afford the domain expert more creative control over the final visualization module. This improved process is presented in Figure 5 1B. The computer engineer needs to create a customized simulation solver for the domain of interest and pass it to the visualization expert. The visualization exper t then can build the visualization with the ontology centered tool an d to the domain expert. The domain expert can finely tune the influences between simulation and visualization
125 and any required simulation coefficients. The tuning by the domain expert is 1B. The end goal of this work is depicted in Figure 5 1C, where a domain expert constructs an interactive, integrative visualizat ion independently In this process, the domain expert has complete expressive freedom and does not require the resources of the visualization and computer engineering experts. Future work to expand on the theory and framework presented in this dissertation falls into two categories: human computer interaction studies, and technical enhancements. Studies could be run that expand on the preliminary studies presented in Section 4.3 Such a study is currently underway at the University of Central Florida College M edicine This study replicates the procedure of the student survey that was carried out during this work, bu t is being administer ed to more participants The study also includes questions that will be used to compare the ontology based visualization viewer to the learning modules currently used by the C ollege of Medicine instructors. The goal of this study is to identify which characteristics of integrative visualizations are useful for various educational purposes. With respect to technical enhancements, the prototype can be developed further to allow for general simulation model building within a particular m odeling paradigm (e.g., compartmental modeling). This was not done during this dissertation because of the development effort required to create a general purpose solver. However, w ith much of the other required development ( e.g., coding of the rendering and interaction system) complete this may be a viable option for future work. A general version of the ontology based designer prototype could be deployed to experts and the expert survey from
126 Section 4.3.1 could then be carried out again, replacing the video demonstration that was used with a fully functional designer prototype. Collectively, t his proposed future work could help further define the strengths and limitations of the approach. Thus far the strengths have been identified as : the domain sem antics based interface ; and simulation modeling and visualization construction activities taking place in the same 3D interaction space. L imitations have been identified as : the approach may not sc ale well to large visualization scenes with many concepts and simulation rules ; and certain required computer graphics concepts may still be prohibitively advanced to be employed by domain experts. By considering the strengths and limitations identified thus far and by future work, a designer should be able to m ake an informed choice as to whether or not the ontology based visualization tool should be used to b uild the visualization i n mind.
127 A B C Figure 5 1. Various engineering processes for constructing simulation based 3D interactive visualizations. A) The approach with current tools. B) A compressed approach that could be used with the current version of the ontology based framework. C) T he ultimate goal a nd reason to continue this work.
128 APPENDIX A EXAMPLE LESSON In this appendix, a lesson plan is presented that was created in collaboration with a medical educator. The target audience for this lesson plan is second year medical students. The lesson was displayed in a side panel of the software prototype and studen ts were asked to complete the lesson as part of a survey. The lesson is formatted here in a manner similar to how it was rendered within the prototype, with figures in line. Welcome Thank you for taking part in this survey. This sidebar contains a sample lesson created by a physiology instructor. The lesson leverages the KOG visualization and serves as an introductory tutorial to the KOG Viewer user interface. Please go through the lesson. You do not need to answer the questions in the lesson, but y ou may do so if you would like. Sample Lesson A 24 year old 3rd year medical student enters the operating room for the first time and passes out. The mechanism of syncope is due to stimulus of the nucleus tractus solitarius of the brainstem resulting in simultaneous enhancement of parasympathetic nervous system tone and withdrawal of sympathetic nervous system tone. You will examine the two subcomponents of this response in this simulation exercise.
129 1 On one end of the spectrum is the cardioinhibitory response, characterized by a drop in heart rate (negative chronotropic effect) and in contractility (negative inotropic effect) leading to a decrease in cardiac output that is significant enough to result in a loss of consciousness. It is thought that thi s response results primarily from enhancement in parasympathetic tone. 1.1 Start the simulation 1.1.1 Press the Play button in the Simulation Control Dock to begin the simulation. Notice the heart should start beating. You can change the perspective of t he heart by using the camera controls (click in the scene and then hold ALT to see controls). Press CTRL + R at any time to reset the camera position. 1.1.2 Expand the View Settings Dock. 1.1.3 In the View Settings Dock, check "Knowledge" to show anatomical concepts and "Flows To" to show blood flow relationships.
130 1.2 Display the baseline features 1.2.1 Select the Heart node (click it with the left mouse button and it should turn yellow). 1.2.2 Expand the Attribute Dock. 1.2.3 Drag and drop the ECG field into the scene. NOTE: Plots can be removed by pressing the 'x' in the upper right corner of the plot. Plots can be moved by dragging the plot with the right mouse button
131 Press and hold the right mouse button to move plots. 1.2.4 Select th e Human Body node and drag the Blood Pressure (BP) plot into the scene. 1.2.5 Select the Left Ventricle (LV) and drag the 'pressure' plot into the scene. 1.2.6 Now drag volume of the Left Ventricle (LV) onto the x axis of the pressure plot. This creates the P V loop. Remember that you can always remove the plot and start over if a mistake is made.
132 1.2.7 Now the simulation should be executing (the heart should be beating) and you should have three plots in your scene. Remember you can control the perspec tive on the heart by moving the camera (hold ALT to see the camera controls) and you can move the plots by right clicking anywhere on the plot and dragging. 1.3 You are now ready to answer Question 1. Question 1 requires you to change the heart rate. To do so you will need to use the slider for heart rate that appears in the Attribute Editor Dock when the Heart node is selected. Please answer Question 1 now.
133 Question 1 Interact with the visualization using the variable plots and sliders to answer the following questions. Be sure to "submit" your responses before moving on. 1A. What happens to the B lood Pressure when the HR=150? 1B. What happens to the BP when the HR =50? 1C. As the heart rate increases, what happe ns to the amount of blood pumped, per stroke, by the left ventricle? 1D. As heart rate increases, which component of the ventricular cycle is most affected: Ve ntricular filling or emptying? 2 On the other end of the spectrum is the vasodepressor response, caused by a drop in blood pressure (to as low as 80/20) without much change in heart rate. This phenomenon occurs due to vasodilation, probably as a result of withdrawal of sympathetic nervous system tone.
134 2.1 Start or Reset the simulation. 2.1 .1 If you are continuing this exercise from Question 1 then you can simply reset the simulation (see image below for reset button). The variable plots will stay in the scene but simulation parameters will be reset. 2.1.2 Make sure "Knowledge" and "Flows To" are both set to visible (see 1.1.2) in the View Settings Dock. 2.1.3 If you are starting this part of Scenario 1 with a freshly opened copy of KOG (your plots from part 1 are no longer visible), then you will need to add the following variable plots t o the scene: 'BP' from the Human node, 'ECG' from the Heart node, and create the P V loop from the Left Ventricle (LV) node (see 1.2.5 and 1.2.6). 2.1.4 Press Play to begin the simulation. 2.1.5 At any time you may adjust the simulation speed using the slider in the Simulation Control dock. The range is from 0.01 (one hundredth the speed of real time) to 1.0 (real time).
135 2.1.6 You should now see a beating heart and with three variable plots in the scene. Remember you can move the camera (press ALT to s ee controls) and move the plots (Right click and drag the plots) at any time. 2.2 You are now almost ready to begin Question 2. To answer Question 2, you will need to change peripheral resistance (the resistance between the Extrahoracic Arteries and the E xtrathoracic Veins). 2.2.1 To access peripheral resistance, select the edge between Extrahoracic Arteries and the Extrathoracic Veins. 2.2.2 To modify peripheral resistance, use the slider for resistance in the Attribute Editor Dock. 2.3 You are now ready to answer Question 2. Please do so now.
136 Question 2 2A. What happens to the Blood Pressure when peripheral resistance is very low (0.5) ? 2B. What happens to the BP when the resistance is elevated (2.0)? 2C. Return to the low resistance state (0.5) and compound that with a low heart rate (50). What happens to the blood pressure now? 2D. Now, increase the resistance to 2.0 but keep the heart rate at 50. What happens to the blood pressure? Which component appears to have a more substantial effect? 2E. Feel free to investigate the interplay between heart rate, peripheral resistance and blood pressure. You can report other relationships to us in the text box below. Final Thoughts The majority of people with vasovagal syncope have a mixed response somewhere between or as a combination of these two ends of the spectrum. Exploratory You may continue to explore the visualization. Every 'flows to' edge has a resistance variable that can be modified.
137 AP PENDIX B EXPERT SURVEY RESULT S Results from an expert survey about a case study on the proposed framework are presented in this appendix. Three medical faculty and two engineers took part in the survey. What are the best aspects of the system? Medical Faculty 1: Best aspects are that there is potentia l here. It is also a relatively simple model which can be good or bad. Medical Faculty 2: It allows you to modify a parameter and see the interplay with other factors in the loop Medical Faculty 3: There is a lot of power in the physiological model and m any "experiments" that can be done. There is obviously a lot of versatility in the animation system and infinite uses by the look of it. The ability to visualize to see real time changes and make observations is very important. Engineer 1: Good visual representation of heart. Very intuitive. Engineer 2: B eing able to manipulate variables, thinking about what I want to see What are the worst aspects of the system? Medical Faculty 1: Lacks the ability to stop at certain points for students to record average pressures, or to calculate variables when things have changed. Medical Faculty 2: Cannot display timing markers on graphs to coordinate time on one loop vs another (or with the EC G ) Medical Faculty 3: Not completely intuitive how to include charts in the animation needs the instruction sheets Engineer 1: Artery to vein that we adjusted pressure in could use a visual representation like a vessel that gets bigger and smaller depe nding on resistance. Engineer 2: Spending time getting numerical values from the graph. Although I could have estimated much faster.
138 Are there any parts of the system which you found confusing or difficult to fully understand? Medical Faculty 1: The demo video is in "programming" language and the actual development is not useful to a basic scientist. The terms do not mean much to me. Medical Faculty 2: Would be helpful to have n umbers from graphs/loo p display at side Medical Faculty 3: Not really some of the terminology is not intuitive for someone who is not a programmer but this was not a problem from the user perspective when doing the exercise. Engineer 1: V isually it is easier for me to see the name of the attribute next to the axis it is plotting vs. seeing them both below the x axis. Not that big of a deal Engineer 2: I t s been awhile since I've seen p ressure volume loops. I had to G oogle that just to be certain. B ut that is the beauty of this kind of learning, I could read about it and mess ar ound with it at the same time. Were there any aspects of the system which you found particularly irritating although they did not cause major problems? Medical Faculty 1: Not being able to stop to collect number data or to overlay PV loops to compare before or after a condition/treatment. Medical Faculty 2: Cannot start knowledge section without turning off ALT control box. Print is small. Do not see an option to enlarg e. Medical Faculty 3: It may be better to have a separate panel of icons to drag graphs into the animation Engineer 1: No ability for Y axis look inversion makes it very difficult for me to navigate. (I always use inverted look). Engineer 2: I made some m I'd like to see default values marked. I was changing the graph scales, and there was some funny behavior; I could not use the number pad. units of s What were the most common mistakes you made when using the system? Medical Faculty 2: R emembering function of mouse and key settings
139 Medical Faculty 3: I had to check back on the instructions once or twice to figure out how to alter resistance if I was just working in another area like heart rate but this was OK after using the system for a few minutes. An orientation of 5 minutes with students would be enough to navigate this. Engineer 1: O nly mistake was inverting the plot for x and y axis on the pressure/volume curve. Engineer 2: Not enough mistakes to have a most common one. I built the PV loop the other way around once so the axes were reversed. If this were a public, live teaching tool, that should be allowed but not without a note that appears for "common settings" Similar to a note for default settings in the user adjustable variables. What changes would you make to the system to make it better point of view? Medical Faculty 1: If a basic scientist were to use it, it should be in terms of how we can modulate. Not sure if it's meant for a programmer to maintain or a basic scientist. Medical Faculty 3: Have an interface to select variables like pressure, volume, flow without having them hidden one step away inside another panel? Engineer 1: Y axis look inversion ability. Y value units next to the Y axis. Dropping Y value next to the Y axis. Engineer 2: C lick on any point on the gr aph, see the (y) value w ith units. Do you currently leverage 3D visualizations in your work? Why or why not? Medical Faculty 1: Currently, no as physiology is more concept based and less dependent perhaps on 3 D. It could be more beneficial in other sy stems such as the kidney. Medical Faculty 2: Yes. Medical Faculty 3: Yes we use them more for anatomical work at present but incorporating physiology is excellent student need to integrate structural relations with function. Nice work Engineer 1: Ye s. M ixed reality simulation. Engineer 2: Y es, it s the only way to fly
140 L IST OF REFERENCES Alani H (2003). TGVizTab: An ontology visualization extension for Protg. Proceedings of the 2003 Workshop on Visualization Information in Knowledge Engineerin g. ACM Press: New York, NY, pp. 434 435. Ald ridge R M (2009). Using models to teach congenital heart defects: a hands on approach. Dimensions of Critical Care Nursing 28 (3):116 122. Artignan G and Hascot M (2011). From Coding to Automatic Generation of Legends in Visual Analytics Lecture Notes in Business Information Processing 73 (4): 404 417. Bates DW, Cullen DJ, Laird N, Petersen LA, Small S D Servi D, Laffel G, Sweitzer BJ, Shea B F, Hallisey R et al (1995). Incidence of a dverse d rug e vents and p oten tial a dverse d rug e vents. American Medical Association 274 : 29 34. Bartels R H, Beatty J C and Barsky A (1987). An introduction to splines for use in computer graphics and geometric modeling Kaufmann Pu blishers: Los Altos, CA, USA Beermann J, Tetzlaff R, Bruckner T, Schebinger M, Mller Stich BP Gutt CN, Meinzer HP, Kadmon M and Fischer L (2010). Three dimensional visualization improves understanding of surgical liver anatomy. Medical Education 44 (9): 936 940. Bell D and Ludwig S A (2005). Grid service discovery in the financial markets sector. Journal of Computing and Information Technology 13 (4): 265 270 Bell D, Mu stafee N, de Cesare S, Taylor SJE, Lycett M and Fishwick P A (2008). Ontology engineering for simulation component r euse. International Journal of Enterprise Information Systems 4 (4): 47 61. Beneken JE W (1965). A mathematical approach to cardiovascular function: the uncontrolled human system PhD t hesis, Medisch Fysisch Institut TNO. Benjamin P and Graul M (2006). A f ramework for adaptive modeling and ontology driven simulation. Proceedings of the SPIE, Enabling Technologies for Simulation Science X, 6227 : 622705.1 622705.11. Benjamin P, Patki M and Mayer R (2006). Using ontologies for simulation modeling. In: Perro ne LF, Lawson B Liu J and Wieland FP (eds). Proceedings of the 2006 Winter Simulation Conference IEEE Computer Society : Washington, DC, USA, pp 1151 1159. Benzaken V Fekete J, Hemery P, Khemiri W and Manolescu I (2011) Edi t Flow: Data intensive interactive workflows for visual analytics IEEE 27th International
141 Conference on Data Engineering IEEE Computer Society: Washington, DC, USA pp 780 791 Berners Lee T, Hendler J and Lassila O (2001). The Semantic Web. Scientific American 284 (5): 34 43. Bhatt M, Rahayu W and Sterling G (2004). sedOnto: A Web enabled ontology for synthetic environment representation based on the SEDRIS specification. Fall Simulation Interoperability Workshop. Fall Simu lation Interoperability Workshop IEEE Computer Society Press: Piscataway, NJ, USA Bhatt M, Rahayu W and Sterling G (2005). Synthetic environment representational semantics using the web ontology language. Proceedings of the Sixth International Conference on Intelligent Data Engineering and Automated Learning Sp ringer: Berlin, pp 9 16. Blender3D (2012). Blender H ome. http://www.blender.org accessed 1 September 2012. Blomqvista E and hgren A (2008). Constructing an enterprise ontol ogy for an automotive supplier. Engineering Applica tions of Artificial Intelligence 21 (3): 386 397. Bosca A and Bonino D (2006), OntoSphere3D: A m ultidimensional v isualization t ool for o ntologies. Proceedings of the 17th International Conference on Database and Expert Systems Applications IEEE Computer S ociety: Washington, DC, USA, pp 339 343. Buntain, C ( 2008 ) 3 D ontology visualization in semantic search Proceedings of the 46th ACM Southeast Conference ACM Press : New York NY, USA, pp 204 208. Carroll J, Dickinson I, Dollin C, Reynolds D, Seaborn e A and Wilkinson K (2004). Jena implementing the semantic web recommendations. Proceedings of the Thirteenth International World Wide Web Conference A CM Press: New York, NY, USA, pp 434 435. Chand rasekaran S, Silver G, Miller JA, Cardoso J and Sheth A P (20 02). Web service technologies and their synergy with simulation. Proceedings of the 2002 Winter Simulation Conference In: Ycesan E, Chen CH, Snowdon JL and Charnes JM (eds). IEEE Computer Society : Washington, DC, USA, pp 606 615. Chang K, Sung Y and Che n I (2002). The effect of concept mapping to enhance text comprehension and summarization. Experimental Education 71 : 5 23.
142 C ooper JB, Newbower RS, Long C D and McPeek B (2002). Preventable anesthesia mishaps: a study of human factors. Quality and Safety in Health Care 11 : 277 282. da Silva P (2001). User i nterface d eclarative m odels and d evelopment environments: a s urvey. In: Palanque P and Patern F (eds). Proceedings of the 7th I nternational C onference on Design, S pecification, and V erification of I nteractive S ystems Springer Verlag: Berlin, pp 207 226 Daconta MC, Obrst LJ and Smith K T (2003). The Semantic Web: A Guide to the Future of XML, Web Services and Knowledge Management Wiley Publishing: Indianapolis, IN, USA. Derakhshani D (2010). Intro ducing Maya 2012 Wiley Publishing, Inc: Indianapolis, IN, USA. Diaz J, Petit J and Serna M (2002). A s urvey of graph layout p roblems ACM Computer Surveys 34 (3): 313 356. Eades P (1984). A heuristic for graph drawing In: Stanton R G (ed ). Congressus numerantium Utilitas Mathematica: Winnipeg, pp 149 160. Ebbesen J, Buajordet I, Erikssen J, Brrs O, Hilberg T, Svaar H and Sandvik L (2001). Drug r elated d eaths in a d epartment of i nternal m edicine. Archives of Internal Medicine 161 (19): 2317 2323. Einsfeld K, Ebert A and Wlle J (2007). HANNAH: a v ivid and f lexible 3D i nformation v isualization f ramework In: Banissi E Burkhard RA, Grinstein G Cvek U, Trutschl M, Stuart L, Wyled TG, Andrienko G, Dykes J, Jern M, Groth D and Ursyn A (eds). 11th Inte rnational Conference on Information Visualization IEEE Computer Society: Washington, DC, USA, pp 720 725 Euler L (1775). Principia pro motu sanguinis per arterias determinando. Opera posthuma mathematica et physica 2 :814 823. Ezzell Z, Fishwick P A, Lok B, Pitkin A and Lampotang S (201 1 ). An ontology enabled user interface for simulation model construction and visualization Journal of Simulation 5 : 147 156. Fekete JD (2004). The InfoVis Toolkit. In: Ward MO and Munzner T (eds). Proceedings of the 10 th IEEE Symposium on Information Visualization IEEE Computer Society: Washington, DC, USA 167 174. Fischler IS, Kaschub CE, Lizdas D E and Lampotang S (2008). Understanding of anesthesia machine function is enhanced with a transparent reality simulation. Simulation in Healthcare 3 (1): 26 32.
143 Fishwick P A (1995). Simulation Model Design and Execution: Building D igital W orlds Prenti ce Hall: Upper Saddle River, NJ, USA. Fishwick PA (2004). Toward an integrative multimodeling interface: a human computer interface approach to interrelating model structures. Journal of Simulation 80 : 421 432. Forbes AG, H llerer T and Legrady G (2010) behaviorism : a f ramework for d ynamic d ata v isualization Transactions on Visualization and Computer Graphics 16 (6): 1164 1171. Ganser ER and North SC (2000) An open gra ph visualization system and its applications to software engineering. Software Practice and Experience 30 : 1203 1233. The Gene Ontology Consortium (2000). Gene Ontology: tool for the unification of biology. Nature Genetics 25 : 25 29. The Gene Ontology Consortium (2010). The Gene Ontology in 2010: extensions and refinements. Nucleic Acids Research 38 (1): 331 335. Gibson JJ (1979). The Ecological Approach to Visual Perception. Houghton Mifflin : Boston. Gilson O, Silva N, Grant PQ and Chen M (2008). From w eb d ata to v isualization via o ntology m apping. In: Vilanova A, Telea A, Scheuer mann G and Mller T (eds). Proceedings of the 200 IEEE VGTC Symposium on Visualization Blackwell Publishing Oxfor d, UK, pp 959 966. Gomez Perez A and Corcho O (2002). Ontology l anguages for t he s emantic w eb. Intelligent Systems 17 (1): 54 60. Gruber T R (1993). A Translation approach to portable ontology specifications. Knowledge Acquisition 5 : 199 220. Guo D, Chen J, MacEachren AM and Liao K (2006). A Visualization s ystem for space t ime and m ultivariate p atterns (VIS STAMP) IEEE Transactions on Visualization and Computer Graphics 12 (6): 1461 1474. Haarslev V and Moller R (1988). Visual ization of experimental systems. 1988 IEEE Workshop on Visual Languages pp 175 1 82. Hastings E, Guha R and Stanley KO (2007). NEAT p articles: d esign, r epresentation, and a nimation of p article s ystem e ffects Proceedings of the 2007 IEEE
144 Symposium on Computational Intelligence and Games. IEEE Computer Society: Washing ton, DC, USA pp 154 160. Heer J, Card SK and Landay JA (2005). prefuse: a toolkit for interactive information visualization. In: van der Veer GC and Gale C (eds). Proceedings of the SIGCHI conference on Human factors in computing systems A CM, New York, NY, USA, pp 421 430. Hopkins JF and Fishwick P A (2003). The rube framework for software modeling and customized 3 D visualization. Visual Language and Computing 14 (1): 97 117. Huh MY and Song K (2002). DAVIS: A Java based data visualization s ystem. Computational Statistics 17 (3): 411 423 Jackson S L, Krajcik J and Soloway E (1998). The design of guided learner adaptable scaffolding in interactive learning environments. In: Atwood ME, Karat MC, Lund A, Coutaz J and Karat J (eds). Proceedings of the 1 998 Conference on Human Factors in Computing Systems IEEE Computer Society: Washington, DC, USA, pp 187 194. Jrgensen S E, Fath B, Bastianoni S, Marques J, Muller F, N ielsen S N, Patten B, Tiezzi E and Ulanowicz R (2007). A New Ecology: Systems Perspecti ve Elsevier: Oxford, UK. Kalogerakis E, Christodoulakis S and Moumoutzis N (2006). Coupling ontologies with graphics content for k nowledge driven visualization. In: Simon A, Welch G and Bolas M (eds). Proceedings of the 2006 IEEE conference on Virtual R eality IEEE Computer Society: Washington, DC, USA, pp 43 50. Krahl D (2002). The E xtend s imulation e nvironment. In: Yucesan E, Chen CH, Snowdon JL and Chames JM ( eds ) Proceedings of the 2002 Winter Simulation Conference IEEE Computer Society : Washington, DC, USA, pp 205 2 1 3. Katifori A, Halatsis C, Lepouras G, Vassilakis C and Giannopoulou E (2007). Ontology vis ualization methods a survey. ACM Computing Surveys 39 (4). Klipp E, Liebermeister W, Wierling C, Kowlad A, Lehrach H and Herwig R (2009). Systems Biology: A Textbook Wiley Publishing: Weinheim, Germany. Kohn L T, Corrigan J M and Donaldson M S (2000). To Err Is Human: Building a Safer Health System National Academy Press: Washington, DC, USA. Kolodnytsky M and K ovalchuk A (2001). Interactive software tool for d at a v isualization In: Loncaric S and Babic H (eds). Proceedings of the 2nd International Symposium on Image and Signal Processing and Analysis IEEE Computer Society Press: Piscataway, NJ, USA, pp 91 93.
145 Knublauch H, Fergerson RW, Noy N F and Musen M A (2004). The Protg OWL plugin: a n open development environment for semantic we b applications. In: McIlraith S A, Plexousakis D and Harmelen F (eds). Proceedings of the 2004 International Semantic Web Conference Spring er: Berlin, pp 229 243. Kuebler W M, Mertens M and Pries A R (2007). A two component simulation model to teach respiratory mechanics. Advances in Physiology Education 31 : 218 222. Lacy L W (2006 ). Interchanging discrete event simulation process interactio n models using the web ontology language OWL PhD t hesis, University of Central Florida. Lacy L and Gerber W (2004). Potential modeling and simulation applications of the web ontology language OWL. In: Ingalls RG, Rossetti MD, Smith JS and Peters BA (eds). Proceedings of the 2004 Winter Simulation Conference. IEEE Computer Society Press: Piscataway, NJ, USA, pp 265 270. Lawr ence PF, Bell RM and Dayton M T (2006). Essentials of General Surgery Lippincott Wi lliams & Wilkins: Philadelphia, PA, USA. Le ape L, Berwick D, Clancy C, Conway J, Gluck P, Guest J, Lawrence D, Morath J, healthcare: a safety imperative. Quality and Safety in Health Care 18 : 424 428. Lee EA and Seshia SA (2011). Introduction To Embedded Systems LeeSeshia.org Liang V and Paredis C J (2003). A port ontology for automated model composition. In: Chick S, Snchez P J, Ferrin D and Morrice D J (eds). Proceedings of the 2003 Winter Simulati on Conference IEEE Computer Society Press: Piscataway, NJ, USA pp 613 622. Lin M C and Canny J F (1991). A fast algorithm for incremental distance calculation. Proceedings of International Conference on Robotics and Automation. IEEE Computer Society: Washington, DC, USA, pp 1008 1014. Loresnsen WE and Yamrom B (1989). Object oriented computer animation. Proceedings of the IEEE 1989 National Aerospace and Electronics Conference IEEE Computer Society: Washington, DC, USA, pp 588 595. Lucas B, Abram GD, Collin NS, Epstein DA, Gresh DL and McAuliffe KP (199 2 ). An architecture for a scientific v isualization. In: Kaufman A and Nielson G (eds). Proceedings of the 3rd Conference on Visualization IEEE Computer Society Press: Los Alamitos, CA, USA, pp 107 114.
146 Mason WT and Strike P W (2003). See one, do one, teach one -is this still how it works? A comparison of the medical and nursing professions in the teaching of practical procedures. Medical Teacher 25 (6): 664 666. Matkovic K, Hauser H, Sainitzer R and Gro ller M E (2002). Process visualization with levels of detail. Proceedings of the 2002 IEEE Symposium on Information Visualization. IEEE Computer Society: Washington, DC, USA, pp 67 70. McSharry PE Clifford GD, Tarassenko L and Smith LA ( 2003 ) A dynam ical model for generating synthetic electrocardiogram signals Transactions on Biomedical Engineering 50 (3): 289 294. Mitra P, Noy N and Jaiswal A (2005). Ontology mapping discovery with uncertainty. In : Gil Y Motta E Richard Benjamins VR and Musen MA (eds). Proceedings of the 4th Internationa l Semantic Web Conference Springer: London pp 537 547. Miller J A, Baramidze GT, Sheth AP and Fishwick P A (2004). Investigating ontologies for simulation modeling. Proceedings of the 2004 Symposium on Simulation IEEE Computer Society: Washington, DC, USA, pp 55 71. Miller E (1998). An introduction to the resource description framework. Bulletin of the American Society for Inf ormation Science and Technology 25 (1): 15 19. Miller A, Seila A F and Xiang X (1999). The JSIM w eb based simulation environment. Fu ture Generation Computer System 17 : 119 133. Modell H I (2000). How to help students understand physiology? Emphasize general models. Advances in Physiology Education 23 : 101 107. National Instruments (2008). Create advanced user interfaces with OpenGL based 3D v isualization [White Paper]. http://www.ni.com/white paper/3170/en#toc2 accessed 1 September 2012. Noy N F, Sintek M, D ecker S, Crubzy M, Fergerson RW and Musen MA (2001). Creating semantic w eb contents with Protg 2000. IEEE Intelligent Systems 16 (2): 60 71. OGRE (2012). OGRE Open Source 3D Graphics Engine. http://www. ogre3d .org accessed 1 September 2012. Oliver R and Herrington J (2003). Exploring technology mediated learning from a p edagogical perspective. Interactive Learning Environments 11 (2): 111 126. Page EH ( 1994 ) Si mulation modeling methodology: p rinciples and etiology of decision support PhD thesis, Virginia Tech
147 Papert S and Harel I (1991). Constructionism Ablex Publishing: Norwood, NJ. Park M (2005). Ontology based customizable 3D modeling for simulation PhD t hesis, University of Florida. Park M and Fishwick P A (2005). Integrating dynamic and geometric model components through ontology based inference. Journal of Simulation 81 (12): 195 813. Paulheim H and Probst F (2010). Ontology enhanced user interfaces: a s urvey. International Journal on Semantic Web and Information Systems 6 (2): 39 59. Petridis P, Dunwell I, Panzoli D, Arnab S Protopsaltis A, Hendrix M and de Freita S (2012). Game engines selection f ramework for high fidelity serious a pplications International Journal of Interactive Worlds 2012 : 1 19. Phanouriou C (2000). UIML: A device independent user interface markup language. PhD t hes is. Virginia Polytechnic Institute and State University. Pietriga E (2006). Semantic web data visualization with graph style sheets. In: Proceedings of the 2006 ACM Symposium on Software Visualization. ACM: New York, NY, USA, pp 177 178. Pop A, Fritzson P and Johansson O ( 2004 ) An integrated framework for model driven design and development using Modelica. In : Elmegaard J, Sporring K, Srensen and EK (eds). Proceedings of the 45th Conference on Simulation and Modeling of the Scandinavian Simulation Soc iety pp 23 24. Prinz A, Bolz M and Findl O (2010). Advantage of three dimensional animated teaching over traditional surgical videos for teaching ophthalmic surgery: a randomized study. Ophthalmology 89 (11): 1495 1499. Puerta A R (1997). A model based i nterface development environment. IEEE Software 14 (4): 40 47. Quarles J (2009). The design and evaluation of a mixed reality approach to interactively blend dynamic models with corresponding physical phenomena PhD thesis, University of Florida. Quarles J, Lampotang S, Fischler I and Fishwick P (2008). A Mixed reality approach for merging abstract and concrete k nowledge. Proceedings of the 2008 IEEE Virtual Reality Conference. IEEE Computer Society: Washington, DC, USA, pp 27 34. Ravden SJ Graham I and Johnson GI (1989). Evaluating usability of human computer interfaces: A practical method Wiley Publishing: Chichester, England.
148 Rodriguez Paz J M, Kennedy M, Salas E, Wu AW, Sexton JB, Hunt EA and Pronovost P paradigm. Quality and Safety in Health Care 18 :63 68. Roehler L and C antlon D (1997). Scaffolding: a powerful tool in social constructivist classrooms. In Hogan K and Pressley M (eds). Scaffolding student learn ing: Instructional approaches and issues. Brookline: Cambridge, MA. Rothe CF and Gertsing J M (2002). Cardiovascular interactions: an interactive tutorial and mathematical model. Advances in Physiology Education 26 : 98 109. Rosse C and Mejino JL V (2007). The foundational model of anatomy ontology. In: Burg er A, Davidson D and Baldock R (eds). Anatomy Ontologies for Bioinformatics: Principles and Practice Springer: London. S Couto CD, van Meurs WL, Goodwin J A and Andriessen P (2006). A model for educati onal simulation of neonatal cardiovascular pathophysiology. Simulation in Healthcare 1 : 4 12. Santhanam AP, Imielinska C, Davenport P, Kupelian P and Rolland J P (2008). Modeling real time 3 D lung deformations for medical v isualization IEEE Transactions on Information Technology in Biomedicine 12 (2): 257 270. Schroeder WJ, Lorensen WE, Montanaro GD and Volpe C R (1992). VISAGE: An Object Oriented Scientific Visualization System In: Kaufman A and Nielson G (eds). Proceedings of the 3rd Conference on Vis ualization. IEEE Computer Society Press: Los Alamitos, CA, USA, pp 107 114. Schroeder WJ, Martin KM, Lorensen WE (1996). The design and implementation of an object o r iented t oolkit for 3D graphics and v isualization In: Yagel R and Nielson GM (eds). Proc eedings of the 7th conference on Visualization IEEE Computer Society Press : Los Alamitos, CA, USA, pp 93 100. SEDRIS (2012). SEDRIS Technologies: The Source for Environmental Data Representation & Interchange. http://www.sedris.org accessed 15 September 2012. Senay H and Ignatius E (1994). A knowledge based system for visualization design Computer Graphics and Applications 1 4 (6): 36 47. Shneiderman B and Plaisant C (2005). Designing the User Interface: Strategies for Effective Human Computer Interaction. Pearson Education, Inc: NJ, USA. Shim H and Fishwick P A (2008). Visualization and interaction design for ecosystems. In: Jorgensen S E and Elsevier B V (eds). Encyclopedia of Ecology Elsevier: Amsterdam, T he Netherlands, pp 3685 3693.
149 Sil n C, Wirell S, Kvist J, Nylander E and Smedby (2008). Advanced 3D v isualization in student centered medical education. Medical Teacher 30 (5): 115 124. Silver GA, Lacy L W and Miller J A (2006). Ontology based representations of simulation models followi ng the process interaction world view. In: Perrone LF, Lawson B, Liu J and Wieland FP (eds). Proceedings of the 2006 Winter Simulation Conference IEEE Press : Washington, DC, USA, pp 1168 1176. Silver G A, Hassan OA, and Miller J A (2007). From domain ontologies to modeling ontologies to executable simulation models. In: Henderson SG Biller B, Hsieh MH, Shortle J, Tew JD and Barton RR. Proceedings of the 2007 Winter Simulation Conference IEEE Press : Washington, DC, USA, pp 1108 1117. Simulink (2012). Simulink: Simulation and Model Based Design. http://http://www.mathworks.com/products/simulink/ accessed 1 September 2012. Sing J an d Sivaswamy J (2010). Creating user i nterface for interactive s imulations. In: Biswas G, Carr D, Chee YS and Hwang WY (ed s). IEEE 3rd International Conference on Digital Game and Intelligent Toy Enhance Learning IEEE Computer Society: Washington, DC, USA, pp 38 45. Tang L, Chen C Zou J, Lin Y, Lin D and Li J (2011). OntoPlant: An integrated virtual plant software package for different scale applications Proceedings of the 2011 IEEE International Conference on Spatial Data Mining and Geographical Knowledge Services IEEE Computer Society: Washington, DC, USA, pp 308 314. Thiriet P, Batier C, Rastello O Sylvestre v E, Ree th NV, Guillot A, Collet C and Hoyek NE (2011). 3D h uman a natomy l earning d emonstration of 3D Tools used in t eaching: 3D v ideos, p odcasts, PDF In: Verbraeck A, Helfert M Cordeiro J and Shishkov B (eds) Proceedings of the 3rd International Conference on Computer Supported Education SciTePress, pp.408 411 Toth E (2000). Representational scaffolding during scientific inquiry: interpretive and expressive use of inscriptions in classroom learning. I n: Gleitman L R, Joshi A K (eds ) Proceedings of the tw enty second annual conference of the cognitive science societ y Erlbaum: Mahvah, NJ USA, pp 953 958. Tolk A (2006). What comes after the Semantic Web PADS i mplications for the Dynamic Web. Proceedings of the 20th Workshop on Principles of Advanced and Distributed Simulation. IEEE Computer Society: Washington, DC, USA, pp 55 62.
150 Tolk A and Tu rnitsa C D (2007). Conceptual m odeling of information exchange requirements based on ontological means. In: Henderson SG Biller B, Hsieh MH, Shortle J, Tew JD and Barton RR. Proceedings of the 2007 Winter Simulation Conference. IEEE Computer Society: Washington, DC, USA, pp 1100 1107. TouchGraph (2012). TouchGraph Navigator 2. http://www.touchgrap h.com, accessed 15 September 2012 Treinish LA, Butler DM, Senay H, Grinstein GG, Bryson ST (1992). Grand challenge problems in visualization s oftware. In: Kaufman A and Nielson G (eds). Proceedings of the 3rd Conference on Visualization IEEE Computer Society Press: Los Alamitos, CA, USA, pp 366 371. Troy M and Molle r T (2004). Human F actors in Visualization Research. IEEE Transactions on Visualization and Computer Graphics 10 (1): 72 84 Upson C, Faulhaber TA K amins D, Laidlaw D, Schlegel D, Vroom J, Gurwitz R and van Dam A (1989). The application visualizatio n system: a computational environment for scientific visualization IEEE Computer Graphics and Applications 9 (4): 30 42. Vasconcelos V (2011). Blender 2.5 Character Animation Cookbook Packet Publishing : Birmingham, UK. Vigas FB Wattenberg M van Ham F Kriss J McKeon M (2007) Many Eyes: A Site for Visualization at Internet Scale. IEEE Transactions on Visualization and Computer Graphics 13 (6): 1121 1128. von Landesberger T, Kuijper A, Schreck T, Kohlhammer J, van Wijk JJ, Fekete JD and Fellner DW (2 011). Visual analysis of large graphs: state of the art and future r esearc h c hallenges. Computer Graphics Forum 3 0 (6): 1719 1749. Vygotsky LS (1978). Mind and society: t he development of higher mental processes. Harvard University Press: Cambridge, MA. Wang K, Zhang L, Gai C and Zuo W (2011). Illustrative visualization of segmented human cardiac anatomy ba sed on context preserving model. Proceedings of the 2011 Conference on Computing in Cardiology IEEE Press: Washington DC, USA, pp 485 489. Warner H R (1959). The use of an analog computer for analysis of control mechanisms in the circulation. Proceedings of the Institute of Radio Engineers pp 1913 1916. Wang W, Juttler B, Zheng D an d Liu Y (2008). Computation of r otation m inimizing f rames. ACM Transactions on Graphics 27 (1): 2:1 2:18.
151 Weber EH (1850) Uber die Anwendung der Wellenlehre auf die Lehre vom Kreislaufe des Blutes und ins besondere auf die Pulslehre. Ber Verh Kgl Sachs Ges Wiss. Math Phys Kl Weinberg G M (2001). An introduction to ge neral systems thinking. Dorset House: New York. Woods W A (1975). What's in a l ink: f oundations for s emantic n etworks. In: Bobrow D and Collins A (eds.). Representation and Understanding: Studies in Cognitive Science Academic Press: New York, NY, USA. Yamaguchi F (1988). Curves and s urfaces in c omputer a ided g eometric d esign Springer: Berlin Yu X and Ganz A (2011). MiRTE: Mixed r eality t riage and e vacuation game for m ass c asualty information systems design, testing and training Proceedings of the 2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society pp 1899 2002. Zeigler B P, Praehofer H and Kim TG (2000) T heory of Modeling and Simulation: Integrating Discrete Event and Continuous Complex Dynamic Sys tems Academic Press: San Diego, CA.
152 BIOGRAPHICAL SKETCH Zach Ezzell was born in 1983 in Crystal River, Florida to Peter and Patricia Ezzell. Zach grew up in Crystal River, graduated from Crystal River High School in 2001 and then began his studies at the University of Florida As an undergraduate, Zach studied digital a rts and was a University Scholar This allowed him to be introduced to the world of research. Zach continued at the University of Florida to pursue his Ph.D. under the advisement of D r. Paul Fishwick and was awarded a four year UF Alumni Fellowship. Zach worked on a variety of projects during his time as a graduate student, including projects funded by the Department of Defense NASA, and the National Institute s of Health. His Ph.D. work focused on defining a better user interface for constructing 3D visualizations based on simulation models His work has appeared in numerous journals and conferences including the Journal of Simulation and the Winter Simulation Conference. After gra duation, Zach plans to co found a company with Dr. Fishwick to create mobile technology to support public health investigators.