1 AGE RELATED CHANGES IN M OTOR RELATED NEURAL ACTIVITY AS ASSESSED BY STRUCTURAL EQUATI ON MODELING By XIAO ZHOU A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMEN TS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2012
2 2012 Xiao Zhou
3 To my parents for all of their love and support
4 ACKNOWLEDGMENTS I would like to ac knowled ge the following individuals for their support in the completi on of this document and project. Towards the substance and science of the current work, I would like to thank my committee members: Dr. Keith White, Dr. Andreas Keil and Dr. Bruce Crosson. For technical assistance, I would like to thank Dr. Keith McGregor and Dr. Jason Craggs. I would further like to recognize Jingyi Huang for useful discussion. Finally, I would like to thank my family and friends for their inspiration and assistance throughout the completion of this project.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ 4 LIST OF TABLES ................................ ................................ ................................ ........... 6 LIST OF FIGURES ................................ ................................ ................................ ........ 7 ABSTRACT ................................ ................................ ................................ .................... 8 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ ..... 9 Aging Related Motor Decline ................................ ................................ ................... 9 Aging Brain ................................ ................................ ................................ ............ 10 Functional Magnetic Resonance Imaging and Structural Equation Modeling ......... 11 2 METHODS ................................ ................................ ................................ ............ 13 Participants ................................ ................................ ................................ ............ 13 Experimental Design ................................ ................................ ............................. 13 Data Acquisition ................................ ................................ ................................ .... 14 Data Preprocessing ................................ ................................ ............................... 15 Structural Equation Modeling Analysis ................................ ................................ ... 17 3 RESULTS ................................ ................................ ................................ .............. 24 Network Models ................................ ................................ ................................ ..... 24 Path Analysis ................................ ................................ ................................ ......... 25 4 DISCUSSION ................................ ................................ ................................ ........ 33 LIST OF REFERENCES ................................ ................................ .............................. 38 BIOGRAPHICAL SKETCH ................................ ................................ ........................... 42
6 LIST OF TABLES Table page 2 1 Participant Characteristics ................................ ................................ ................. 21 3 1 Path analysis models and goodness of fit indices ................................ .............. 29 3 2 Regression mat rix of estimated ROI direct influences for younger group ........... 29 3 3 Regression matrix of estimated ROI mediated influences for younger group ..... 29 3 4 Regression matrix of estimated ROI total influences for younger group ............ 30 3 5 Regression matrix of estimated ROI direct influences for older group ............... 30 3 6 Regression matrix of estimated ROI mediated influences for older group .......... 30 3 7 Regression matrix of estimated ROI total influences for older group ................. 31
7 LIST OF FIGURES Figure page 2 1 Regions of interest ................................ ................................ ............................ 22 2 2 Simplified path of tested network model. ................................ ........................... 23 3 1 Group representation of moto r related neural network models .......................... 32
8 Abstract of Thesis Presented to the Graduate School of the University of Florid a in Partial Fulfillment of the Requirements for the Degree of Master of Science AGE RELATED CHANGES IN M OTOR RELATED NEURAL ACTIVITY AS ASSESSED BY STRUCTURAL EQUATI ON MODELING By Xiao Zhou December 2012 Chair: Keith D. White Major: Ps ychology Aging has long been associated with decli nes in motor function part icularly in the upper extremity. Recent evidence using functional magnetic resonance imaging has shown that older adults tend to recruit more bilateral motor cortical areas as com pared to younger adults and this recruitment may be associated with decline in motor performance. However, less work has been done relating to the reliability of these patterns of recruitment across multiple sessions and groups. Further, we are aware of f ew studies that have investigated the reliability of recruitment within motor cortical regions. Using a structural equation modeling approach, we tested the reliability of activity in motor cortical regions both within and across age groups during a simple unimanual tapping task over three assessment sessions. Our resultant path models showed that reliability of network recruitment was stronger and more lateralized to the left hemisphere in younger adults. Despite some methodological limitations using the p resent approach, we believe these findings add support to the growing literature indicating increased variability of motor recruitment yields performance decl ines in aging populations.
9 CHAPTER 1 INTRODUCTION Aging Related Motor Decline During the past cen tury, the population in the United States under age 65 tripled. At the same time, the population aged 65 or over increased by a factor of 11. This age group is projected to grow by an average of 2.8 percent annually from 2010 to 2030, generation enters their elderly years. By the middle of this century, the population aged 65 or over is estimated increase to 88.5 million, resulting in a novel population dynamic in US history. That is the proportion of elderly adults will outnumber young by the year 2050 ( projected from 2010 US Census data) The rapid growth of the elderly population will likely also yield a proportional increase in health problems associated with advanced age. A primary medical concern with this age cohort is aging relat ed declines in motor ability For example, aging adults face more difficulty when th ey are trying to drive a vehicle (Anstey et al. 2005), clean the garden, or even walk to exercise (Buchman et al. 2009) Motor handicaps, or motor performance deficits, inc lude slowing of movement (Seidler et al. 2002) increased variability of movement (Christou & Carlton 2001) coordination difficulty (Ketcham et al. 2 002 ) and difficulties with balance and gait (Buchner et al. 1997) These handicaps constitute a major sou rce of injury and morbidity in older adults (Seidler et al. 2010 ) For instance, 20 to 30 percent of the older adults who fell suffer moderate to severe injuries that limit their mobility and independence, and reduce the quality of the ir life (Alexander et al. 1992 )
10 Aging Brain The mechanisms underlying impaired motor performance in old age are complex, which involve the central and peripheral nervous systems declines as well as degeneration in muscle tissues (Seidler et al. 2010 ) However, changes in the motor system of the brain are suggested to be one of the main contributing factors. That is, aging related decreases in muscle size and function are insufficient in the explanation of the level of movement and coordination deficits shown in older adults ( Enoka et al. 2003 ). Recent research has shown that while older adults do show greater variability in motor unit twitch force than younger adults, training of the first dorsal interosseus muscle reduced their force variability without changing their motor u nit twitch force (Christou 2011) This suggest s that force variability may be more relevant to the central neural network function than to the motor unit size. Numerous studies have reported that older adults exhibit different patterns of brain recruitmen t during motor tasks in comparison to younger adults. For instance, McGregor et al. ( 2009 ) show ed that during a single hand movement, the contralateral cortex in older and younger adults both tend to increase its activity proportionally with increasing mot or unit involvement and task difficulty. However, the involvement of the ipsilateral primary motor cortex differs between age groups. While younger adults tend to show decreases in activity as compared to baseline, older adults tend to show increases in ac tivity in this area. Additionally, Wu and Hallett ( 2005 ) reported that older adults recruited additional anterior medial frontal areas during a tapping task as compared to younger adults. Motor network recruitment has also been investigated with fMRI using causal and correlational modeling to denote aging related changes in the brain. It is suggested that older adults recruit different neural networks during motor task
11 performance relative to younger adults. However, it is yet uncertain as to whether this a lteration of network recruitment is compensatory allowing for improved motor performance or is task detrimental. Functional Magnetic Resonance Imaging and Structural Equation Modeling To explore the neural network activated during the performance of motor tasks, functional magnetic resonance i maging (fMRI) can be a very useful technique. When specific brain regions are activated to drive the peripheral muscles to perform intended motor functions, the increased blood oxygen demand will cause an overcompensa tion of oxygenated blood to be sent to the active area. The changes in the blood oxygenated hemoglobin level are me asured by changes in magnetic susceptibility during specialized sequences in an MRI scanner. As such, neural correlates underlying specific m otor tasks are revealed with high spatial resolution MRI apparatus. To identify the neural networks involved in particular motor tasks, we need not only explore the active brain regions, but also determine the interaction of those brain regions, which driv es the performance of the task. Structural Equation Modeling (SEM) as a network analysis technique can provide information about the interrelationships between or among distinct network components. SEM has been employed in the analysis of fMRI data in nume rous studies (e.g. Taniwaki et al. 2007; Walsh et al. 2008) and can reveal robust neural network differences both betw een and within subjects Furthermore, SEM allows both confirmatory and exploratory analyse s (McIntosh & Gonzalez Lima 1994) which allows an investigator to test pre set models in addition to those that are data driven. The present study investigates the reliability of both confirmatory and exploratory path analyses using structural equation modeling. W e hypothesize : 1) Younger adults
12 recru it a lateralized motor neural networ k involving primary motor cortex (M1), supplementary motor area (SMA) presupplementary motor area (p S M A) and ventral supplementary motor area (vSMA) during motor task performance, and this network will be reliable acro ss sessions. 2) O lder adults rec ruit a differing, more bilateral motor neural network particularly in strength of paths from left to right primary M1. Older adults exhibit additional recruitment of brain regions compared to young adults 3 ) The network re cruited by older adults shows more variability in terms of network components and their path coefficients across sessions
13 CHAPTER 2 METHODS Participants Fourteen right handed participants (7 younger [ages 19 31], 7 older [ages 64 76]) enrolled in this study. All of them were reported healthy at the time of study participation, and had no history of any neurological or psychiatric disorders. All participants were prescreened by self report for claustrophobia, pregnancy, contraindicated medications or met al implants. Participant characteristics are listed in Table 2 1. All participants were recruited and provided written consent after the experimental procedures were fully explained to t hem in accordance with the Institutio nal Review Board of the Universit y of Florida (IRB# 474 1997; IRB# 462 2006). Experimental Design A block design, right hand motor task was used for mapping of left primary motor cortex. Blocks consisted of seven images (28 seconds) for both rest and active conditions. Six cycles (altern ating between 7 rest images and 7 active images) comprised each functional run (5 minutes 36 seconds). Tasks were cued by the change word were presented using E Prime software (PST Soft, Pittsburgh, PA). Motor responses were made on an RP04U button response unit manufactured by MagConcept (Sunnyvale, CA). An additional two images (discard acquisitions) were acquired prior to stimulus present ation in each run to allow for signal equilibration. Between the two runs of a given session, participants again verified their understanding of the task via verbal report.
14 The motor task was a block presentation of a repeated button squeeze using opposit ion of the index finger and thumb at 2Hz, paced by flashing of the visual stimulus. Participants were trained on the task inside the scanner prior to data button resp onse unit prior to acquisition and instructed on target force output. The (across the abdomen). The participant was asked use the index finger thumb squeeze to depress a button for each trial press with only as much force as necessary to produce approximately active) hand was placed in a prone, resting position along the side of the body. Participa nts were visually monitored for movements of this hand during active task blocks. Data Acquisition Magnetic resonance images were acquired on a 3 Tesla Achieva Whole Body Scanner (Philips) using an 8 channel SENSE radio frequency head coil. Head motion wa s minimized using foam padding. Before functional imaging sequences, structural images were also acquired (160 1.0 mm thick sagittal slices, using a 3D T1 weighted sequence: time of echo (TE) = 8.057 ms; time of repetition (TR) = 3.685 ms; flip angle (FA ) = 8). Whole brain high resolution gradient echo echo planar functional images (EPI) were acquired using 57 2 mm thick axial slices and the following parameters: TE = 30 ms; TR = 4000 ms; FA = 87; Matrix = 96x96 (2mm x 2mm voxels); FOV = 192x192x114mm; SENSE factor = 1.5. Stimuli were presented on a first surface mirror presentation system situated at the rear bore aperture of the magnet.
15 To control for the physiological and cognitive changes between scanning sessi ons over time ( see Bennett & Miller 201 0), the scanning was conducted on the same weekday in three consecutive weeks. To minimize positional deviation between scanning sessions, a laser targeting system was used to assist in the consistent alignment of head position within the head coil. To min imize yaw, the head position and coil arrangement were set during each session so that fixed reference marks on the the nose of the participant was aligned with a st atic reference marker on head coil. Additionally, head rest padding within the coil was kept consistent across all of the sessions. Data Preprocessing Functional images were processed and overlaid onto structural images using the Analysis of Functional Ne uroim aging (AFNI) program (Cox 1996) installed o n a Linux Fedora 12 operating system on a Dell workstation. The 3dTcat program in AFNI was used to exclude the two initial magnetization equilibration images (discard acquisition) in each functional run from subsequent processing. To reduce the head motion effects, time series images from each run we re spatially aligned in 3 dimens ional space to the first functional image using the 3dvolreg program in AFNI, as it is acquired immediately after the acquisition o f T1 any time series contains a significant number of images with gross artifacts, as visually inspected. This did not occur in any of the sampled datasets. Individual time series images in each run were then orthogonalized for linear trends and concatenated into a
16 A 5 mm full width half maximum (FWHM) Gaussian filter was applied to the functional time series images of each individual to ameliorate variability in structural and functional anatomy. Structural and functional images were interpolated to volumes with 1 mm 3 voxels; after which both the structural and functional time series images were warped to stereotaxic space of Talairach and Tournoux @auto_tlrc python script. We hypothesized that younger adults show a more focal and lateralized network recrui tment than older adults during the motor task. Specifically, we expected that younger adults would recruit fewer and m ore left lateralized brain regions as compared to older adults. To test this, we compared between younger and older adults the network components and the interregional influences within the network. Based on previous findings on the neural network recruitm ent during motor tasks ( Boudrias e t al. 2012; Grefkes et al. 2008 ) eight regions of interest (ROIs) were proposed to be involved in the finger tapping process: left and right primary motor cortex ( LM1 / RM1 ), left and right supplementary motor cortex ( LSMA / RSMA ), left and right preSMA ( LpSMA / RpSMA ), and left and right ventral SMA ( LvSMA / RvSMA ). These ROIs were manually drawn as high resolution masks in Talairach space function, and were applied to the functional time series images o f each individual. Two raters independently validated each ROI for accuracy to anatomy. Regions of interest are illustrated in Figure 2 1. For the network connectivity analysis, the z normalized signal time series from all the voxels within each of the eig ht ROIs were averaged and extracted as a single data point for each volume of functional data. The data points were then concatenated into a
17 group matrix. This procedure was repeated for both the younger and the older age groups across all three sessions. Structural Equation Modeling Analysis In the context of network analysis we conducted, it is appropriate to clarify some of the terms and their relation to the terminology used to describe our results. Functional connectivity refers to the correlations of neural activity between different brain regions in the context of a specific cognitive or behavioral function. A functional connectivity between two brain regions shows a statistically significant correlation of neural activity between spatially remote ne urophysiolo gical events (Friston & Buechel 2004), regardless of the correlation. By contrast, an effective connectivity between two brain regions gives information regarding the influence of one brain region on the other one (Friston & Buechel 2004). One o f the advantages of structural equation modeling as a tool for effective connectivity analysis is that it can reveal the direct, indirect, and total effects of one brain region over another. A direct effect is the influence of one region directly over anot her without any mediation of other intervening regions An indirect effect is the influence of one region on another through mediation of one or more intervening regions. The direct and indirect effects make up the total effect of one region over another. Lima 1994) can show how the influence of one region on another could be mediated by indirect influences of other regions and whether these influences change with varying experimental conditions. The fMRI da ta from younger and older groups were used to estimate the models of effective connectivity among the eight brain ROIs involved in finger tapping. During the path analysis, for all models the left primary motor cortex ( LM1 ) was classified as
18 exogenous vari ables and all the other ROIs were treated as endogenous variables. The statistical models were in accordance with an anatomical mode l of hand function (Ward et al. 2008). We used LISREL 8.7 (Scientific Software International, Inc.) to estimate the inter re gional connectivity among the eight ROIs across the two age groups. Network model parameters indicating the inter regional connectivity were calculated (Craggs et al. 2009; Craggs et al. 2007) Any subsequence modifications to the initially proposed model were needed for identification of a model best fit data. Goodness of fit statistics indicates how well a statistical model fits the observed data. LISREL can provide several fit indices for the identified models. We chose the Non normed Fit Index (NNFI), t he Root Mean Square Error of Approximation (RMSEA), and the Standardized Root Mean Square Residual (SRMR) in our present study. The Non normed Fit Index describes how well a proposed model fits the null model as calculated using variable covariances. The index is s et to one even if the obtained value is greater than one. A value above .95 is considered to be a good fitting model, between .90 and .95 is considered marginal, and below .90 is considered as unacceptable. The Root Mean Square Error of Approxim ation is an absolute measure of fit based on the non centrality parameter. Like the NNFI, it has a penalty for complexity as the ratio of chi square to its degrees of freedom. A RMSEA value of .05 or less is considered as a good fitting model while .10 is considered as the cutoff for poor fitting. A 90% confidence interval for RMSEA was also calculated and the ideal lower bound is no greater than .05 and the upper bound is less than .1.
19 The Standardized Root Mean Square Residual, another absolute measure of fit, is defined as the standardized difference between the observed covariance and the predicted covariance. A SRMR value of 0 indicates perfect fit and .08 or less is generally considered good fitting. We conducted a preliminary analysis trying to resolv e reasonable network models that have been pro posed in previous work (Boudrias et al. 2012; Grefkes et al. 2008 ). We found little model convergence across subjects in the younger group when incorporating different regions of interest and literalities as en dogenous variables (effector variables). In fact, the only model that showed reliable convergence and goodness of fit used a single effector variable, the left M1. There is currently great controversy as to the laterality of unimanual movements in terms of their effector motor cortex (left or right). While some groups contend that unimanual activity is bilateral in the motor corte x in origin (Haken et al. 1986; Verstynen et al., 2007), others believe that the left motor cortex (particularly during unimanual movements) takes primary controlling role (Gachibeh et al., 2007; Hanna Pladdy et al., 2002). As the left primary motor cortex does enjoy significant support as the dominant effector in right hand, we tested the model of effective connectivity driven by L M1 in our session and group comparisons (see Figure 2 2 ). As the current study sought to investigate reliability of model fit within and across age groups, w e used a regression approach to test the intersession reliability of the network model. The connect ivity models used in our analysis consisted of path coefficients between the averaged values of the time series in each motor related regions of interest. Path coefficients indicating a connection from the driving variable, in
20 this case the LM1 is referre d to as gamma and the other path coefficients are referred to as beta. As stated above, t he model derived from the first session was fit to the subsequent two sessions.
21 Table 2 1. Participant Characteristics Younger (Mean: 24.6 years SD: 4.9) Older (Mea n: 70 years SD: 5.0) Participant Age Gender Participant Age Gender Y1 19 F O1 64 F Y2 23 F O2 67 F Y3 31 F O3 75 F Y4 26 M O4 64 M Y5 22 F O5 72 F Y6 31 M O6 72 F Y7 20 F O7 76 M
22 Figure 2 1. Regions of interest. Presented are regions of inte rest for the network analysis. Altering colors represent distinct bilateral regions including primary motor cortex (M1) pink & blue; Supplementary motor cortex (SMA) green and red; Pre supplementary motor cortex (pSMA) light and dark orange; ventral S MA (vSMA) light and dark yellow.
23 Figure 2 2. Simplified path of tested network model. Boxes indicate effector (latent) variables and ovals represent dependent (exogenous) variable
24 CHAPTER 3 RESULTS Network Models Structural Equation Modeling was u sed to conduct the network analysis to compare the network connectivity between younger and older adults. As noted earlier, t he base model s we initially hypothesized showed poor fit with the data As such, the exploratory approach showed that the major eff ector cortical area appeared to be the left motor cortex. As this was a reasonable assumption based on the unimanual, externally guided finger tapping task we found a model for each of the younger and older groups, which both show good fit with the data. The fit indices for the two models are shown in Table 3 1. The network models for the two age groups are shown in Figure 3 1. The model for the younger group shows direct influences of LM1 on LpSMA LSMA LvSMA RM1 RpSMA and RvSMA direct influences of LpSMA on LSMA and RpSMA direct influences of LSMA on RpSMA and RSMA direct influences of LvSMA on LpSMA and RpSMA direct influence of RM1 on LpSMA direct influences of RpSMA on RSMA and RvSMA direct influences of RSMA on LpSMA and RM1 and direct infl uence of LvSMA on RvSMA This model has a NNFI of 0.99, showing good fit with the data. The RMSEA of this model is 0.041, and the 90% confidence interval for RMSEA is (0.031, 0.051), both showing good fit as the RMSEA is less than 0.05, the lower bound of the confidence interval is less than 0.05 and the upper bound is less than 0.1. The SRMR of this model is 0.016, which also shows the good fit feature of the model. The model for the group of older adults reveals the direct influences of LM1 on LpSMA LSM A LvSMA RM1 RSMA and RvSMA direct influences of LpSMA on
25 LSMA LvSMA and RpSMA direct influences of LSMA on RpSMA and RSMA direct influences of LvSMA on LpSMA and RpSMA direct influence of RM1 on LpSMA direct influences of RpSMA on RSMA and RvSMA direct influence of RSMA on RM1 and direct influences of RvSMA on LvSMA and RM1 The NNFI for the network connectivity model for the older group is 1, showing quite good fit with the data. It has a RMSEA of 0.032, and a 90% confidence interval of (0.022 0.044), both evidencing the good fit of the model. The SRMR value of 0.011 also supports that this model fits well with the data. Path Analysis The path coefficients show the strength of the influence of on e region over another. Tables 3 2, 3 3, and 3 4 show the path coefficients of the direct, indirect, and total connectivity, respectively, between the motor related regions of interest in the group of younger adults. Tables 3 5, 3 6, and 3 7 show the path coefficients in the network connectivity model f or the older group. Path coefficients indicating a connection from the driving variable, in this case the LM1, is referred to as gamma, and the other path coefficients are referred to as beta. In the model for younger adults, the path coefficients of LM1 on LpSMA (gamma=0.39), LSMA (gamma=0.59), and LvSMA (gamma=0.18) are larger than those of LM1 on RM1 (gamma= 0.05), RpSMA (gamma= 0.06), and RvSMA (gamma=0.05), while LM1 does not have significant influence on RvSMA This demonstrates that during simple fi nger tapping movements, the left primary motor cortex has stronger interconnectivity with the motor related regions in the left hemisphere than those in the right hemisphere. However when considering the indirect influences of the motor related regions ove r each other, the larger path coefficients of LM1 on the motor related regions in the right
26 hemisphere than those in the left hemisphere reveals stronger indirect influences of LM1 on the motor related regions in the right hemisphere. Additionally, the neg ative direct path coefficient of LM1 on RM1 agrees with the negative indirect path coefficients of RM1 on all the other motor related regions of interest. This clearly supports the inhibitory influences of the motor related network on the right primary mot or cortex. Furthermore, the large direct path coefficients of the connectivity between LpSMA and RpSMA (beta=0.69), LSMA and RSMA (beta=0.78), LM1 and LSMA (beta=0.59), RSMA and RM1 (beta=0.77) reveal the strong coupling between the neural activation betwe en these regions. With the addition of positive indirect path coefficients, the total path coefficients of the connectivity between these regions are further increased, leading to an enhanced coupling between the neural activation between these regions. Th e indirect path coefficients of the connectivity between LM1 and RSMA (gamma=0.67) and between LSMA and RM1 (beta=0.66) also add to the prominent correlation between the activity in these regions. In the model for the group of older adults, the path coeffi cients of the influences of LM1 on LpSMA (gamma=0.36), LSMA (gamma=0.59), LvSMA (gamma=0.55) are also larger than those of LM1 on RM1 (gamma=0.27), RSMA (gamma=0.09), RvSMA (gamma=0.06), while the path coefficient between LM1 and RpSMA is not significant. Like in the younger group, this also supports that the left primary motor cortex has stronger interconnectivity with the motor related regions of interest in the left hemisphere than those in the right hemisphere. In accordance with the younger group, the indirect path coefficients of the influences of LM1 on the motor related regions in the right hemisphere are larger than those in the left hemisphere; however the path
27 coefficients are more variable. Moreover, as compared to the model for the younger group the direct path coefficient of the influence of LM1 on RM1 is positive (gamma=0.27) in the model for the older group, and most of the indirect path coefficients of the connectivity between RM1 and other motor related regions of interest are positive. Thu s the inhibitory influences of the motor related regions on the right primary motor cortex seen in the younger adults are relatively less prominent or absent in older adults. Similarly to the model for younger group, the large direct path coefficients of t he connectivity between LpSMA and RpSMA (beta=0.76), LSMA and RSMA (beta=0.67), LM1 and LSMA (beta=0.59), RSMA and RM1 (beta=0.43) show the strong coupling between the neural activation in the motor related regions of interest. Again, with the addition of positive indirect path coefficients, the total path coefficients of the connectivity between these regions reveal an enhanced coupling between the neural activation of these regions. The indirect path coefficient of the connectivity between LM1 and RSMA is also higher (gamma=0.62) than other indirect path coefficients, however the indirect path coefficient of the connectivity between LSMA and RM1 (beta=0.30) is not as prominent as that in the model for the younger group. Moreover, the direct path coefficien t of the connectivity between RvSMA and LvSMA is high (beta=0.52), suggesting a strong coupling between the activation of these two regions during finger tapping. It is worth notice of the negative correlation between the hemodynamic responses of LpSMA and LvSMA (beta= 0.65), suggesting a strong inhibitory influences of LpSMA on LvSMA Another noticeable feature of the model for older group is that LvSMA receives negative indirect influences from LM1 (gamma= 0.28), LSMA (beta= 0.02), RM1 (beta= 0.08), RSMA (beta= 0.04), RvSMA (beta= 0.11),
28 and LvSMA itself (beta= 0.20). This suggests the inhibitory effects of most of the motor related regions of interest on the left ventral supplementary motor area. These effects are also shown in the total influences of tho se regions on LvSMA
29 Table 3 1. Path analysis models and goodness of fit indices Group df CI SRMR NNFI Younger 10 59.74 0.041 0.031 0.051 0.016 0.99 Older 9 36.94 0.032 0.022 0.044 0.011 1 is the Root Mean Square Error of Approximation (RMSEA). CI is the 90% confidence interval for RMSEA. SRMR is the Stand ardized RMSEA. NNFI is the Non normed Fit Index. Table 3 2. Regression matrix of estimated ROI direct influences for younger group Model for Younger Adults ROI Direct Influences Left Right M1 PSMA SMA VSMA M1 PSMA SMA VSMA Left M1 PSM A 0.39 0.22 0.15 0.29 SMA 0.59 0.34 VSMA 0.18 0.29 Right M1 0.05 0.77 PSMA 0.06 0.69 0.14 0.03 SMA 0.78 0.11 VSMA 0.05 0.56 For this table and Tables 3 3, 3 4, 3 5, 3 6, 3 7, the numbers are standardize d unidirectional loadings. Direction of influence is column row. Table 3 3. Regression matrix of estimated ROI mediated influences for younger group Model for Younger Adults ROI Indirect Influences Left Right M1 PSMA SMA VSMA M1 PSMA SMA VSMA Left M1 PSMA 0.18 0.10 0.16 0.02 0.01 0.06 0.10 0.07 SMA 0.20 0.03 0.05 0.08 0.06 0.02 0.06 0.02 VSMA 0.08 0.13 0.04 0.02 0.02 0.17 0.02 0.01 Right M1 0.52 0.30 0.66 0.06 0.05 0.10 0.05 0.02 PSMA 0.51 0.12 0.12 0.18 0.12 0.04 0.14 0. 04 SMA 0.67 0.38 0.07 0.08 0.06 0.02 0.07 0.02 VSMA 0.25 0.45 0.14 0.08 0.07 0.02 0.08 0.02
30 Table 3 4. Regression matrix of estimated ROI total influences for younger group Model for Younger Adults ROI Total Influences Left Right M1 PSMA SMA VSMA M1 PSMA SMA VSMA Left M1 PSMA 0.58 0.10 0.16 0.24 0.17 0.06 0.19 0.07 SMA 0.79 0.37 0.05 0.08 0.06 0.02 0.06 0.02 VSMA 0.27 0.13 0.04 0.02 0.02 0.17 0.02 0.29 Right M1 0.47 0.30 0.66 0.06 0.05 0.10 0.83 0.02 PSMA 0.45 0.81 0. 26 0.15 0.12 0.04 0.14 0.04 SMA 0.67 0.38 0.85 0.08 0.06 0.13 0.07 0.02 VSMA 0.30 0.45 0.14 0.08 0.07 0.58 0.08 0.02 Table 3 5. Regression matrix of estimated ROI direct influences for older group Model for Older Adults ROI Direct Influences Left Right M1 PSMA SMA VSMA M1 PSMA SMA VSMA Left M1 PSMA 0.36 0.44 0.19 SMA 0.59 0.37 VSMA 0.55 0.65 0.52 Right M1 0.27 0.43 0.10 PSMA 0.76 0.05 0.04 SMA 0.09 0.67 0.16 VSMA 0.06 0.28 Ta ble 3 6. Regression matrix of estimated ROI mediated influences for older group Model for Older Adults ROI Indirect Influences Left Right M1 PSMA SMA VSMA M1 PSMA SMA VSMA Left M1 PSMA 0.23 0.16 0.05 0.07 0.03 0.07 0.07 0.20 SMA 0. 22 0.06 0.02 0.13 0.06 0.02 0.03 0.07 VSMA 0.28 0.20 0.02 0.20 0.08 0.11 0.04 0.11 Right M1 0.32 0.15 0.30 0.06 0.03 0.11 0.01 0.03 PSMA 0.48 0.09 0.04 0.29 0.13 0.05 0.05 0.14 SMA 0.62 0.31 0.03 0.13 0.06 0.02 0.03 0.07 VSMA 0.13 0.19 0.0 2 0.07 0.04 0.01 0.02 0.04
31 Table 3 7. Regression matrix of estimated ROI total influences for older group Model for Older Adults ROI Total Influences Left Right M1 PSMA SMA VSMA M1 PSMA SMA VSMA Left M1 PSMA 0.59 0.16 0.05 0.36 0.16 0.07 0.07 0.20 SMA 0.80 0.31 0.02 0.13 0.06 0.02 0.03 0.07 VSMA 0.27 0.44 0.02 0.20 0.08 0.11 0.04 0.41 Right M1 0.60 0.15 0.30 0.06 0.03 0.11 0.44 0.13 PSMA 0.48 0.67 0.09 0.25 0.13 0.05 0.05 0.14 SMA 0.71 0.31 0.70 0.13 0.06 0.18 0.03 0.07 VSMA 0.20 0.19 0.02 0.07 0.04 0.29 0.02 0.04
32 Figure 3 1. Group representation of mo tor related neural network models. Figure A shows the model for younger adults and B for older adults. Green lines in Figure A stand for paths absent in older group while red lines in Figure B stand for paths absent in younger group. The blue and orange lines illustrate the different influence of LM1 on RM1 between age groups.
33 CHAPTER 4 DISCUSSION One of the main findings of this study is that while performing the s ame simple finger opposition task, the BOLD responses in the primary motor cortex ipsilateral to the movement was negatively correlated with that contralateral to the movement in younger adults, however older adults exhibited a positive correlation between the BOLD responses in the ipsilateral primary motor cortex and that in the contralateral primary motor cortices. This is consistent with the findings of McGregor et al. (2009) that younger and older adults both had positive BOLD responses in left primary motor cortex contralateral to the active hand, however older adults had positive BOLD responses in right primary motor cortex where younger adults had negative BOLD responses. One might easily relate the positive correlation between the BOLD responses in the ipsilateral and contralateral primary motor cortices in older adults to undesirable mirror movements. However this might not be the case of our study since we carefully monitored the performance of all the participants both before the scanning and insi de the scanner and did not find any overt mirror movements in their left, nondominant hand. Moreover, given the simplicity of the motor task used in our current study, it is less likely for mirror movements to occur as the frequency of mirror movements dec reases significantly with simpler movements (Armatars et al. 1996; Hoy et al., 2004). Admittedly, precise detecting of mirror movements by measuring of motor evoked potentials in the hand muscles could provide evidence for a more comprehensive conclusion. Several studies have shown that participants increase recruitment of the ipsilateral primary motor cortex when the task difficulty increases (Meister et al. 2005; Verstynen
34 ver activation of bilateral motor cortices compared to younger adults (Mattay et al. 2002). However, given the relative simplicity of the motor task involved in the current study, we believe it is less likely that over recruitment at a neural compensatory level explain the increased activity in older adults. In contrast to the idea of compensatory over activation, Langan et al. (2010) and McGregor et al. (2011) showed evidence that additional recruitment of the ipsilateral primary motor cortex may be non co mpensatory or counterproductive to the performance of unimanual movements, which is likely the case of our study. Whether the additional recruitment of ipsilateral primary motor cortex in older adults serves to compensate for the less precise structure to function relationships still remains as an interesting topic for future research. Recent studies have showed the cause of the additional activation of ipsilateral primary motor cortex in older adults may be a loss of effective interhemispheric inhibition i n older adults. Talelli et al. (2008) found that the extent of ipsilateral sensorimotor cortex activation during unimanual hand grip is associated with changes in interhemispheric inhibition in older adults. Therefore the cause of the over recruitment of i psilateral primary motor cortex in older adults is shifted from chronological aging to cortical responsiveness and related interhemispheric inhibition. This is also in accordance with one of the findings of our study that the omnibus negative correlation b etween the ipsilateral prima ry motor cortex and all the other motor related regions of interest in younger adults is absent in the older group. In addition to the over recruitment of primary motor cortex ipsilateral to the active hand in older adults, our present investigation also shows partial consistency in the
35 network recruitment between younger and older adults. Both age groups exhibit greater coupling among the motor related regions of interest in the contralateral hemisphere compared to that in the ipsilateral hemisphere. This is in agreement with previous findings (Grefkes et al. 2008; Walsh et al. 2008) involving network analysis of motor function in fMRI. In the study of Grefkes and colleagues (2008), dynamic causal modeling of fMRI data revealed an enhanced neural coupling toward the contralateral primary motor cortex and reduced connectivity in the ipsilateral hemisphere during unimanual movements, which the investigators suggested is associated with transcallosal inhibition and top down modulati on. This is consistent with findings on aging related alterations in interhemispheric inhibition (McGregor et al. 2011; Talleli et al. 2008) For instance in the study of McGregor et al. (2011), the investigators found that older adults showed lower levels of interhemispheric suppression as indicated by the overall increased activity in the ipsilateral hemisphere as compared to younger adults during performance of index finger tapping The findings from this study and our present investigation along with t hat from the study by Grefkes et al. (2008), can be considered in light of the HAROLD (Hemispheric Asymmetry Reduction in Older Adults (Cabeza 2002)) model, which describes an aging related progressive increase in bilateral recruitment of functional activi ty during tasks for which one hemisphere is normally dominant. The study by Grefkes et al. (2008) also suggested a key role of the supplementary motor area in modulating cortical motor network activity during hand movements. This is consistent with the res ult of our study that the strong connectivity within the LSMA LpSMA RpSMA RSMA circuit is maintained throughout the study in both younger and older group. Using structural equation modeling of fMRI data, Walsh
36 et al. (2008) further evidenced the important role of the SMA interhemispheric connectivity in the production of bimanual movements, and they suggested that the interhemispheric connectivity might serve to modulate the distinct intrahemispheric networks in both hemispheres during voluntary bimanual mi rror movements. Further investigation on the specific role of the SMA interhemispheric connectivity in hand movements is warranted including the study of left and right handed participants performing both dominant and non dominant hand movement tasks at va rious difficulty levels. We intended to test the reliability of motor related network recruitment during simple finger movements, however we could not find a model with good reliability across study sessions. Several reasons may underlie behind this, prima ry of which may be the small number of participants included in the present study. The seven younger and seven older adults in each group exhibited consistent network recruitment while performing the motor task within group, however the limited sample size and run duration might not provide us sufficient power to unveil a network model that is reliable across time. The sample size is an especially important issue for SEM analysis. While it is true for most studies that the larger the sample size, the more a ccurate the results, SEM typically favors even more sample points. Boucard et al. (2007) have shown that the results of SEM analysis are reliable when the sample size reaches 100. While the 140 sample points for each participant in each session in our pres ent study meets this criterion the limited number of participants might exacerbate the larger effects of inter participant variability on our reliability metrics.
37 Another limitation to our present study is that the time series in all the voxels within ea ch ROI were taken into the SEM analysis, without being restricted to a functional connectivity network. Although we aimed to examine the effective connectivity between anatomical regions, the whole ROI approach could possibly lead us to a false negative er ror. Therefore in our subsequent investigations, we plan to restrict inclusion to functionally active voxels in our analysis and may see more reliable results. In sum, the present study revealed a series of findings on aging related differences in lateral ity and reliability of network recruitment during simple hand movements despite a few technical limitations. Future investigation will be needed to discover a more restricted and reliable motor related neural network involved in hand movements.
38 LIST OF REFERENCES Alexander, B.H., Rivara, F.P., Wolf, M.E., 1992. The cost and frequency of hospitalization for fall related injuries in older adults. Am. J. Public Health 82 1020 1023. Anstey, K., Wood, J., Lord, S., Walker, J., 2005. Cognitive, sensory, and physical factors enabling driving safety in older adults. Clin. Psychol. Rev 25, 45 65. Armatas C., Summers J., Bradshaw J., 1996. Handedness and performance variability as factor influencing mir ror movement occurrence. J. Clin Exp Psychol 18, 823 835 Bennett, C.M., Miller, M.M., 2010. How reliable are the results from functional magnetic resonance imaging? Ann. N. Y. Acad. Sci. 1191, 133 155. Boucard, A., Marchand, A., Nogues, X., 2007. Reliability and validity of structural equation modeling appli ed to neuroimaging data: a simulation study. J. Neurosci. Methods. 166, 278 292. Boudrias, M., Gon alves, C. S., Penny, W. D., Park, C., Rossiter, H. E., Talelli, P., Ward, N. S., 2012. NeuroImage 59, 3398 3405 Buchman, A. S., Boyle, P. A., Wilson, R. S ., Fleischman, D. A., Leurgans, S. Bennett, D. A., 2009. Association between late life social activity and motor decline in older adults. Arch. Intern. Med. 169, 1139 46. Buchner, D. M., Cress, M. E., de Lateur B. J., Esselman, P. C., Margherita, A. J., P rice, R., Wagner, E.H., 1997. The effect of strength and endurance training on gait, balance, fall risk, and health services use in community living older adults. J Gerontol A Biol Sci Med Sci 52, M218 M224. Cabeza R., 2002. Hemispheric asymmetry r eduction in older adults: the HAROLD model. Psychol. Aging 17, 85 100. Christou, E. A., 2011. Aging and variability of voluntary contractions. Exerc Sport Sci Rev 39, 77 84. Christou, E.A., Carlton, L.G., 2001. Old adults exhibit greater motor output variability than young adults only during rapid discrete isometric contractions. J Gerontol A Biol Sci Med Sci 56 B524 B532. Cox R. W., 1996. AFNI: software for analysis and visualization of functional magnetic resonance neuroimages. Comput. Biom ed. Res. 29, 162 173.
39 Craggs J G Staud R Robinson M E Perlstein W M Price D D ., 2012 Effective connectivity among brain regions associated with slow temporal summation of C fiber evoked pain in fibromyalgia patients and healthy controls. J. Pain 13, 390 400. Craggs J G Price D ., D Verne G N Perlstein W M Robinson M M. 2007. Functional brain interactions that serve cognitive affective processing during pain and placebo analgesia. NeuroImage 38, 720 72 9 Enoka, R. M., Ch ristou, E. A., Hunter, S. K., Kornatz, K. W., Semmler, J. G. Taylor, A. M., Tracy, B. L., 2003. Mechanisms that contribute to differences in motor performance between young and old adults. J. Electromyogr. Kinesiol. 13, 1 12. Friston, K. A., & Buechel, C. 2004. Functional connectivity: e igenimage s and multivariate analyses. Ashburner Friston, & Penny (Eds.), Human Brain Function. 2nd Edition. London: Academic Press. Ghacibeh G A Mirpuri R Drago V Jeong Y Hei lman, K. M., Triggs, W. J., 2007. Ipsilateral motor activation during unimanual and bimanual motor tasks. Cli n. Neurophysiol. 118, 325 3 32 Grefkes, C., Eickhoff, S. B., Nowak, D. A., Dafotakis, M., Fink, G. R., 2008. Dynamic intra and interhemispheric interactions during unilateral and bilateral hand movements assessed with fMRI and DCM. NeuroIma ge 41, 1382 1394 Haken, H., Kelso J. A. S., B unz H., 1985. A theoretical model of phase transitions in human hand movements. Biol. Cybern. 51, 347 356. Hanna Pladdy B Mendoza J E Apo stolos G T Heilman K M. 2002. Lateralised motor control: hemispheric damage and the loss of deftness. J Neurol Neurosurg Psychiatry. 73 574 57 7. Hoy K E Fitzgerald P B Bradshaw J L Armatas C A Georgiou Karistianis N. 2004. Inv estigating the cortical origins of motor overflow. Brain Res. Brain Res. Rev. 46 315 3 27. Ketcham, C.J., Seidler, R.D., Arend W.A. Van Gemmert, George E. Stelmach. Age related kinematic differences as influenced by task difficulty, target size, and movem ent amplitude. J Gerontol B Psychol Sci Soc Sci 57 P54 P64. Langan, J., Peltier, S.J., Bo, J., Fling, B.W., Welsh, R.C., Seidler, R.D., 2010. Functional implications of age differences in motor system connectivity. Front Syst Neurosci 4, 1 11.
40 Mattay, V. S., Fera, F., Tessitore, A., Hariri, A. R., Das, S., Callicott, J. H., and Weinberger, D. R., 2002. Neurophysiological correlates of age related changes in human motor function. Neurology 58, 630 635. McGregor, K. M., Craggs, J. G., Benjamin, M. L., Crosson, B., White, K. D., 2009. Age related changes in motor control during unimanual movements. B rain Imaging Behav 3, 317 331. McGregor K M Zlatar Z Kleim E Sudhyadhom A Bauer A Phan S Seeds L Ford A Manini T M Wh ite K D Kleim J Crosson B. 2011. Physical activity and neural correlates of aging: a combined TMS/fMRI study. Behav Brain Res. 222 158 168. McIntosh, A.R., Gonzalez Lima, F., 1994. Structural equation modelling and its application to network an alysis in functional brain imaging. Hum. Brain Mapp. 2, 2 22. Meister, I., Krings, T., Foltys, H., Boroojerdi, B., Mller, M., Tpper, R., Thron, A., 2005. Effects of long term practice and task complexity in musicians and nonmusicians performing simple a nd complex motor tasks: implications for cortical motor organiz ation. Hum. Brain Mapp 25 345 352. Seidler, R.D., Alberts, J.L., Stelmach, G.E., 2002. Changes in multi joint performa nce with age. Motor Control 6 19 31. Seidler, R. D., Bernard, J. A., B urutolu, T. B., Fling, B. W., Gordon, M. T., Gwin, J. T., Kwak, Y., Lipps, D. B., 2010. Motor control and aging: links to age related brain structural, functional, and biochemical effects. Neurosci Biobehav Rev 34, 721 733. Taniwaki, T., Okayama, A., Y oshiura, T., Togao, O., Nakamura, Y., Yamasaki, T., Ogata, K., Shigeto, H., Ohyagi, Y., Kira, J., Tobimatsu, S., 2007. Age related alterations of the functional interactions within the basal ganglia and cerebellar motor loops in vivo. NeuroImage 36, 1263 1 276. Talelli, P., Ewas, A., Waddingham, W., Rothwell, J.C., Ward, N.S., 2008. Neural correlates of age related changes in cortical neurophysiology. NeuroImage 40, 1772 1781. Verstynen, T., Diedrichsen, J., Albert, N., Aparicio, P., Ivry, R. B., 2005. Ips ilateral motor cortex activity during unimanual hand movements relates to task complexity. J Neurophysio l. 93 1209 1222. Verstynen, T., S pencer, R., Stinear, C. M., K onkle, T., Diedrichsen, J., Byblow, W. D., Ivry R. B., 2007. Ipsilateral corticospinal projections do not predict congenital mirror movements: A case report Neuropsychologia 45, 844 852.
41 Walsh, R. R., Small, S. L., Chen, E. E, Solodkin, A., 2008. Network activation during bimanual movements in humans. NeuroImage 43, 540 553. Wu, T., Hall ett, M., 2005. The influence of normal human ageing on automatic movements. J Physiol. 562, 605 615.
42 BIOGRAPHICAL SKETCH Xiao Zhou was born in Xiaogan, China. He enter ed Huazhong University of Science and Technology in Wuhan, China in 2006 and obtained his Bachelor of Engineering degree in 2010, majoring in o ptoelectronic i nformation e ngineering. H e enter ed Central China Normal University in Wuhan, China in 2008 and obtained his Bachelor of Science degree in 2010, majoring in p sycho logy. Mr. Zhou went to graduate school in 2010 at the University of Flo rida entering a PhD program in b ehavioral and c ognitive n euroscience in the Department of Psychology. He currently resides in Gainesville, FL.