Human in-vivo cardiac phosphorus NMR spectroscopy at 3.0 Tesla


Material Information

Human in-vivo cardiac phosphorus NMR spectroscopy at 3.0 Tesla
Physical Description:
xvii, 282 leaves : ill. ; 29 cm.
Bruner, Angela Properzio
Publication Date:


Subjects / Keywords:
Nuclear magnetic resonance spectroscopy   ( lcsh )
Nuclear and Radiological Engineering thesis, Ph.D   ( lcsh )
Dissertations, Academic -- Nuclear and Radiological Engineering -- UF   ( lcsh )
bibliography   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph.D.)--University of Florida, 1999.
Includes bibliographical references (leaves 260-280).
Statement of Responsibility:
by Angela Properzio Bruner.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 030471162
oclc - 42450667
lcc - LD1780 1999 .B894
System ID:

This item is only available as the following downloads:

Full Text







Copyright 1999


Angela Properzio Bruner

This work is dedicated to my loving husband, Thom Bruner, and my
parents, Sharon and Bill Properzio, without whom I would not have had
the loving support I needed to complete this work.


I would like to thank the following individuals for their

assistance and for making this work possible. The greatest thanks go to

my advisor and mentor, Dr. Kate Scott, who encouraged and advised me

through this project. Thanks also goes to Dr. Hee-Won Kim, who taught

me the basics of the ISIS pulse sequence, shared experiences in working

on the General Electric (GE) Signa", and offered valuable pulse

programming support. Credit must be given to David Peterson and Bryan

Wolverton, under the direction of Dr. Fitzsimmons, who built the coils

that were used in this study. Thanks also go to the combined efforts of

Dr. Scott, Dr. Fitzsimmons, Dr. Ballinger, Shands at UF, the VA

Hospital, and the Brain Institute for their efforts in getting a 3.0 T

whole body magnet to the University of Florida. Thanks are also well

deserved for Jim Scott who taught me the chemistry for the phantom


After starting this project, I was lucky to begin collaborative

work with a number of individuals in cardiology both here at the

University of Florida (UF) and at the University of Alabama at

Birmingham (UAB). Great appreciation goes to Dr. Carl Pepine, currently

the division chief of cardiology, Shands at UF, who fully supported my

efforts and encouraged greater work. Also, countless thanks and

appreciation are well deserved for Alice Boyette, the Women's Ischemic

Syndrome Evaluation (WISE) research cardiology technologist. Alice not


only attended all the WISE meetings and scheduled the cardiac patient

studies on the 1.5 and 3.0 T, but also provided patient handholding and

technical support during the studies. In addition, WISE research nurse

Eileen Handberg-Thurmond was very supportive and made sure that all

financial issues were under control and the necessary equipment

purchased. It was through the efforts of cardiology at UF that the

Dinamap" vital signs monitor was purchased and that all patient studies

were financially compensated. I was also fortunate to have my protocol

and my study results analyzed by a group at UAB who already had a

published history in doing cardiac phosphorus spectroscopy as well as

some experience with higher Tesla whole body systems (a 4.1 Tesla). The

expertise, careful analysis, and approval of my data by Dr. Steven

Buchthal, Dr. Jan den Hollander, and Dr. Gerald Pohost provided

excellent feedback that I was on the right track and succeeding in my


I would especially like to thank all of the members of my

committee, Dr. Katherine Scott (Chair), Dr. Jeffrey Fitzsimmons, Dr. J

Ray Ballinger, Dr. Richard Briggs, Dr. Christine Stopka and Dr. David

Hintenlang. These individuals took time out of their busy scheduled to

review my work and support my efforts.

I would also like to sincerely thank those who believed in me and

helped me achieve my doctorate through their continuous moral support.

This includes a long list of family members (Thom Bruner, Sharon

Properzio, Bill Properzio, Tom Bruner Sr, Jess Bruner, Dee Dee Haun,

Carter Haun) and friends (Manuel Arreola, Libby Brateman, Lynn Rill,

Michelle Werner, Cathy Carruthers, Mark Knudsen, Beth Knudsen, Sheila




ACKNOWLEDGMENTS ................... ................... .................iv

LIST OF TABLES. ... .. ....................... ........................ix

LIST OF FIGURES ........... ...... ....................................xi

ABSTRACT ....................................... ....................... xvi


1 INTRODUCTION ...................................... .................1

Statement of Problem ................................................. 2
Spectroscopy's Contribution to Diagnosing Myocardial
Ischemia ...................................................... 3
Research Hypotheses ................................................. 4
Specific Objectives .................................... ............. 5
Assumptions.......................................................... 6
Scope of the Project .................................................7
Significance .................... ..................................... 8

2 REVIEW OF LITERATURE ................. .............................. 10

P-31 Spectroscopy Pulse Sequence Options for the Heart............. 13
Slice Localization Techniques ............. ....................... 14
Multi-Voxel Localization Techniques..............................15
Single Voxel Localization Techniques ............................17
Cardiac P-31 Spectroscopy Results in the Literature ................22
Animal studies ..................................... .............. 22
Human Studies................................... .................24
Post-Processing Calculations and Corrections.......................48
Skeletal Muscle Contamination .................................49
Blood Contamination ..............................................49
Relaxation Corrections............................................ 52
Calculation of pH...................................... ..........56


Phantom Design................... .................................. 61
Gate-able and Depth Phantoms ................................... 61
Slice Profile Phantom .. ........... ..............................64
GE Phosphoric Acid Phantom ......................................64
Radio Frequency Coil Design..................................... ... 65
3.0 Tesla Square Proton Coil Paired with Quadrature
Phosphorus Coil ...............................................65
3.0 Tesla Phosphorus Single-Turn Coil ...........................66


3.0 Tesla Coil Comparisons. ......................................67
1.5 Tesla Coils.................................................. 69
Coil Ideas for Future Cardiac Spectroscopy Studies ..............70
Imaging .................. ............................. ......... .... 70
Spectroscopy ...... ..... .......................... .. .....................71
Localized Proton Spectroscopy. .................................... 72
Localized Phosphorus Spectroscopy................... .............73
Phosphorus STEAMCSI and PRESSCSI............................... 74
Phosphorus ECHOCSI ............................................... 74
Phosphorus SPINECHO ...................................... ........ 75
Phosphorus ISISCSI ............................................... 76
Phosphorus FIDCSI ................................................78
1.5 Tesla to 3.0 Tesla Phosphorus Spectroscopy Comparisons...... 82
Phantom Results ................................. ... ............... 84


Imaging.................... ................................. ........ 113
Spin-Echo Imaging...............................................113
Gradient-Echo Imaging ............ .... .........................114
Human Positioning..................................................... 114
Gating .. ................ ................. ........ ........... 116
Shimming with Localized Proton Spectroscopy.......................119
Cardiac Phosphorus Spectroscopy In-Vivo Acquisition ...............120
Localized P-31 Multivoxel CSI ...................................120
Localized P-31 ISISCSI. .........................................121
Slice Localized P-31 FIDCSI (DRESS) ............................122
Single Turn versus Quadrature Surface P-31 Coil at 3.0 T.......123
Human Test Participants..................................... ........123
In-Magnet Exercise ................................................. 128
Spectroscopy Post-Processing..................................... 129
Post-Processing Software ................... ................. ...130
Skeletal Muscle ................ .................... ............136
Blood Contamination ......................... ....................136
Ti Relaxation Corrections ........................................ 137
Calculations of pH ..............................................139
Analysis .......... ............................................... 140

5 HUMAN DATA REPRODUCIBILITY. ................. ...................... 156

T, Relaxation Corrections ................... .......................157
Overall Reproducibility of the Oblique DRESS Method.................159
Adequacy of the Hydraulic, In-Magnet, Handgrip Exerciser.......... 165
Reproducibility of the Hydraulic Handgrip ......................... 167
No Drop in [PCr]/[ATP] During Exercise with Reference Volunteer...168
Drop in [PCr]/[ATP] Seen with the Handgrip Exerciser with
Ischemia ................ ......................... ............... 170
Myocardial pH Measured at 3.0 T. ....................................171

6 SUMMARY AND CONCLUSIONS ........................................... 174

Implications for Future Research .................................. 176


A IRBS AND SCREENING FORMS .......................................... 180


B FIGURE ACQUISITION PARAMETERS ..... ................................. 193

C HYDRAULIC HANDGRIP................................................. 224

D 3.0 T CARDIAC ACQUISITION PROTOCOL................... ...........228


F HUMAN Ti RELAXATION DATA .......................................... 256

G T, RELAXATION RATES OF DEPTH PHANTOM .............................. 259

REFERENCES ..................................... ......................... 260

BIOGRAPHICAL SKETCH............... .... ................................. 281


Table page

1. Nuclear Spin Parameters ................... ..................... .. 13

2. Research Published on Human, In-Vivo Cardiac P-31
NMR Spectroscopy of Studies on Normal-Controls or
Patients and Related Work ...................................... 26

3. Cardiac [PCr]/[ATP] Ratio of Healthy Volunteers
(Normal Controls) at Rest and During Stress .................... 36

4. Cardiac [PCr]/[ATP] Ratio of Patients with Myocardial
Infarction at Rest and During Stress ........................... 38

5. Cardiac [PCr] and [ATP] Amounts at Rest in Patients with
Myocardial Infarction, Ischemia and in Normal Controls ......... 39

6. Cardiac [PCr]/[ATP] at Rest and Stress in Patients with
Myocardial Ischemia.............................................. 43

7. Cardiac [PCr]/[ATP] at Rest and Stress in Patients with
Myocardial Ischemia with Some Type of Intervention ............. 44

8. Literature Review of In-Magnet Handgrip Exercise Response .......48

9. Published Spin-Lattice Relaxation Times of Myocardial PCr
and ATP ......................................................... 55

10. Myocardial pH in the Literature as Measured by Human,
In-vivo Phosphorus NMR Spectroscopy ............................ 57

11. Gate-able and Depth Phantom Compartment Sizes and
Concentrations ........... .... ............ .................... 62

12. A Comparison of 3.0 Tesla Coil Parameters. ..................... 67

13. Comparison of 1.5 Tesla Coil Parameters ....................... 69

14. Participants at 3.0 T for Cardiac P-31 Spectroscopy .......... 126

15. In Magnet Hydraulic Exercise Handgrip Participants ............ 127

16. Oblique DRESS Acquisition P-31 Metabolite area Values
Obtained with Different TR gating Intervals for the
Purpose of Relaxation Measurements ........................... 158

17. Summary of Resulting T, values for PCr and ATP.c ................ 159


18. WISE 1.5 T Cardiac Spectroscopy Acquisition Success Rate ....... 161

19. 3.0 T Cardiac P-31 Spectroscopy at Rest Only, 3xRR Gating ...... 162

20. 3.0 T Cardiac P-31 Results Categorized by P-31 Surface Coil .... 165

21. Heart rate (HR) and Systolic Blood Pressure (SBP) Response
from 30% of Maximum Effort Isometric Hydraulic Handgrip ....... 166

22. Handgrip exercise compared to dobutamine and treadmill
responses for known WISE studies .............................. 167

23. Handgrip 30% Maximum Effort Isometric Exercise Results of
K.S. Subject Tested Repeatedly................................. 168

24. Results from 1.5 T Cardiac P-31 Exercise Study on Reference
Normal Volunteer .............................................. 170

25. Ischemic and WISE Studies at 3.0 T Show Drop in [PCrl/[ATP]
with Handgrip Exercise ......................................... 171

26. Myocardial pH as Measured on Human Volunteers Using Oblique
DRESS Cardiac P-31 MRS on the GE 3.0 T SIGNA'" for those
Studies where the Pi Peak was Discernible due to Adequate
SNR...................... ................................ ...... 172

27. Comparison of Cardiac P-31 Post-Processing Software. ........... 252


Figure page

1. STEAM Pulse Sequence ............................................ 58

2. PRESS Pulse Sequence ........... ........................ ....... 58

3. ISIS Volumes for 8 Acquisition Voxel Localization. .............. 59

4. Gate-able-Phantom (a) photograph and (b) position in magnet,
without the liquids and with movement direction demonstrated... 87

5. Gate-able-Phantom images, (a) axial, (b) coronal and
(c) sagittal views, as imaged with 25cm square proton coil..... 87

6. The Depth-Changing-Phantom (a) photograph with top open and
(b) position in magnet, with movement direction demonstrated... 88

7. Depth-Changing-Phantom images, (a) axial, (b) coronal and
(c) sagittal views, as imaged with 25cm square proton coil..... 88

8. Axial slice image of Slice Profile Phantom and details on
how oblique DRESS slices were placed within phantom to
estimate the amount of potential contamination from outside
the localized slice ................... ............. ............ 89

9. Photographs of GE's 14.7 M P-31 Phantom from (a) front,
(b) back and relative position in magnet ...................... 89

10. Photographs of quadrature phosphorus and square proton
coil as (a) a paired set and (b) separated onto individual
platforms for 3.0 Tesla ......................................... 90

11. Schematic diagrams of (a) the 25 x 25 cm2 square proton
coil set (tuned to 127.75 MHz) used with (b) the 10 cm
phosphorus quadrature coil at 3.0 T (tuned to 51.71 MHz) ....... 90

12. Photograph of the single turn, 9.5 cm diameter, phosphorus
transceive coil tuned to 51.71 MHz (3.0 Tesla) ................ 91

13. Schematic diagram of the single-turn phosphorus transceive
9.5 cm diameter coil at 3.0 T tuned to 51.71 MHz............... 91

14. Comparisons of single-turn versus quadrature P-31 RF coils
in terms of relative signal based on 25 mm thick DRESS
acquisitions of 14.7 M phantom at 0.5 cm intervals with
TG optimized at each position. ................................. 92

15. Photograph of the single turn, 9.5 cm diameter, phosphorus
transceive coil tuned to 25.87 MHz (1.5 Tesla) with three
small vials that are used to locate the coil in the
proton images........................................ ............ 93

16. Schematic diagram of the single-turn phosphorus transceive
10 cm diameter coil at 1.5 T tuned to 25.87 MHz ................ 93

17. Photograph of 1.5 T quadrature P-31 coil. ....................... 94

18. Schematic diagram of the quadrature phosphorus transceive
10 cm diameter coil at 1.5 T tuned to 25.87 MHz ................ 94

19. Axial images of gate-able phantom comparing images obtained
on 3.0 T using a 25 cm square proton surface coil with the
image pulse-sequences of (a) spin echo, (b) fast spin echo,
(c) gradient echo and (d) fast gradient echo imaging ........... 95

20. The STEAMCSI pulse sequence, GE's version of STEAM for
spectroscopy voxel localization. ................................ .96

21. The PRESSCSI pulse sequence, GE's version of PRESS for
spectroscopy voxel localization ................................... 96

22. Frequency domain of (a) PRESSCSI voxel localized phosphorus
spectroscopy with (b) diagram demonstrating localization ....... 97

23. Frequency domain of (a) STEAMCSI voxel localized phosphorus
spectroscopy with (b) diagram demonstrating localization....... 97

24. The ECHOCSI pulse sequence, one of GE's versions of
Spin Echo for spectroscopy acquisition ....................... 98

25. Frequency domain of (a) ECHOCSI voxel localized phosphorus
spectroscopy with (b) diagram demonstrating localization....... 98

26. Chemical shift imaging (CSI) voxel sizes versus time of
slice plus 2D CSI acquisition on GE Signa Advantage" ............ 99

27. The SPINECHO pulse sequence, one of GE's versions of
Spin Echo for spectroscopy acquisition. ...................... 100

28. Frequency domain of SPINECHO CSI multivoxel localized
phosphorus spectroscopy .................. ....................... 100

29. The ISISCSI pulse sequence (as shown for one gradient),
GE's versions of ISIS for volume, slice, column or voxel
localization ...................................................... 101

30. Frequency domain of (a) ISISCSI slice localized phosphorus
spectroscopy with (b) diagram demonstrating localization ...... 101

31. Frequency domain of (a) ISISCSI column localized phosphorus
spectroscopy with (b) diagram demonstrating localization...... 102

32. Frequency domain of (a) ISISCSI voxel localized phosphorus
spectroscopy with (b) diagram demonstrating localization...... 102


33. Visual display example of the acquisition where the total
number of acquisitions is 32 performed by (a) the current
GE ISISCSI technique, and (b) the modified ISIS technique..... 103

34. Frequency domain of (a) modified ISISCSI localized voxel
sequence for phosphorus (created from eight separate
acquisitions, added and subtracted appropriately during
post-processing) with (b) diagram demonstrating localization.. 104

35. The FIDCSI pulse sequence without phase encoding gradients
turned on, GE's versions of the simple single RF pulse
necessary to produce an FID for a spectroscopy acquisition ... 104

36. Charts comparing (a) relative signal obtained by varying
transmitter gain (TG) values at various depths for slice
localized FIDCSI at 1.5 Tesla, using the quadrature coil
and (b) relative optimized TG value versus depth for the
single turn coil ............................................. 105

37. Charts comparing relative signal obtained by varying
transmitter gain (TG) values at various depths for slice
localized FIDCSI at 3.0 Tesla, using (a) the quadrature
coil, or (b) the single turn coil .... ......................... 106

38. Frequency domain of FIDCSI slice localized phosphorus
spectroscopy of (a) gate-able phantom and (b) depth phantom... 107

39. Multivoxel phosphorus FIDCSI plus CSI of the (a) gate-able
phantom (b) depth phantom................. ................... 108

40. Comparison of 1.5 to 3.0 Tesla results of phosphorus FIDCSI
plus CSI localized voxel scaled by noise level ............... 109

41. Comparison of 1.5 to 3.0 Tesla results of phosphorus,
modified ISISCSI localized voxel (created from eight
separate acquisitions, added and subtracted appropriately
during post-processing) ..................... .................. 109

42. 1.5 T, P-31 single turn RF coil signal from a set of
25 mm thick, oblique DRESS slices (FIDCSI oblique slice)
moved across the internal phosphoric acid vial in the
Slice Profile Phantom.......................................... 110

43. 3.0 T, P-31 single turn RF coil signal from a set of
25 mm thick, oblique DRESS slices (FIDCSI oblique slice)
moved across the internal phosphoric acid vial in the
Slice Profile Phantom .......................................... 110

44. 3.0 T, P-31 quadrature RF coil signal from a set of
25 mm thick, oblique DRESS slices (FIDCSI oblique slice)
moved across the internal phosphoric acid vial in the
Slice Profile Phantom ........................................... 111

45. Human cardiac imaging with the spin-echo pulse sequence
at (a) 1.5 T with the body coil and at (b) 3.0 T with
a surface coil ................. ....... ...... .............. 141


46. Human cardiac imaging with the fast gradient echo pulse
sequence at (a) 1.5 T with the body coil and at (b) 3.0 T
with a surface coil .............................................. 142

47. Prone positioner for in-magnet cardiac spectroscopy ............ 143

48. A comparison of the heart's position in the (a) prone and
(b) supine positions, as shown from a 3.0 T axial slice ....... 144

49. Waveforms of (a) peripheral gating and (b) ECG gating on
the 3.0 T from a normal human volunteer (TEB) on
Sept 20, 1998, displayed at a rate of 21 mm/sec............... 144

50. A comparison of peripheral gating (pg) versus ECG gating,
and breathing during the image versus breath-hold images ...... 145

51. Human proton voxel localized spectroscopy of the heart
and chest wall obtained during one volunteer's shim using
the techniques of GE's (a) STEAMCSI and (b) PRESSCSI .......... 146

52. P-31 FIDCSI with CSI of a human subject at 3.0 T ............... 147

53. DRESS localization via oblique slice select combined
with sensitivity region of coil ............................... 148

54. Examples of 1.5 T cardiac phosphorus spectra localization
problems resulting in (a) liver contamination, or
(b) skeletal muscle contamination, in comparison with
(c) a non-contaminated cardiac spectrum ...................... 149

55. P-31 FIDCSI oblique slice localized human cardiac
spectroscopy (oblique DRESS) of the same subject,
at 1.5 and 3.0 T, on different days showing examples
of resting and exercise spectra, raw and fitted ............... 150

56. Series of cardiac region oblique DRESS spectra
representing decreased skeletal muscle contamination
with increase in depth of spectroscopy slice localization ..... 151

57. Relative size of oblique DRESS slice with (a) a 10-cm
diameter single-turn P-31 surface coil and (b) a 16 x 10
cm2 quadrature P-31 surface coil. ............ ................ 152

58. Water, hydraulic, hand-squeeze ergometer/static-exerciser,
modified from original design by North Coast Medical .......... 153

59. The linear response of the hydraulic handgrip to added
weight on the rubber bulb is illustrated ...................... 154

60. Dinamap" blood pressure and pulse monitoring equipment. ......... 154

61. Myocardial pH is proportional to the frequency
difference of the Pi and PCr peaks in the human, in-vivo
phosphorus NMR spectrum ..................... ........... ....... 155


62. Relaxation correction factors at 3.0 T for cardiac
[PCr]/[ATP] values based on repetition time (TR) values....... 173

63. ECG lead placement for 3.0 T gating. ........................... 230

64. Example 3.0 T data set analyzed by FITMASTER.. ................. 238

65. Example 3.0 T data set analyzed by Sage_IDL". ................... 241

66. Example 3.0 T data set analyzed by Sage_IDL' with baseline
correction points selected. .................. ............... 241

67. Example 3.0 T data set analyzed by Sage_IDL' with baseline
correction.......................................................... 242

68. Example 3.0 T data set processed by MRUIT. ..................... 244

69. Example 3.0 T data set analyzed by MRUI". ...................... 245

70. Example 3.0 T data set processed by FELIX.. ..................... 248

71. Example 3.0 T data set processed by FELIX' with peaks picked. .. 249

72. Example 3.0 T data set processed by FELIX' with baseline
correction.................... ........................... .....250

73. Example 3.0 T data set analyzed by FELIX". ..................... 250

74. Example 1.5 T data set with low SNR analyzed by SAGE_IDL". ..... 252

75. Example 1.5 T data set with low SNR analyzed by FITMASTERT. .... 254

76. 1.5 T Ti relaxation curve for MDPA (Ti = 5.53 sec) ............ 259

77. 3.0 T Ti relaxation curve for MDPA (TI = 6.04 sec) .............. 259

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



Angela Properzio Bruner

August 1999

Chairman: Katherine N. Scott
Major Department: Nuclear and Radiological Engineering

One of the newest methods with great potential for use in clinical

diagnosis of heart disease is human, cardiac, phosphorus NMR

spectroscopy (cardiac P-31 MRS). Cardiac P-31 MRS is able to provide

quantitative, non-invasive, functional information about the myocardial

energy metabolites such as phosphocreatine (PCr), adenosinetriphosphate

(ATP) and pH. In addition to the use of cardiac P-31 MRS for other types

of cardiac problems, studies have shown that the ratio of [PCr]/[ATP]

and pH are sensitive and specific markers of ischemia at the myocardial

level. In human studies, typically performed at 1.5 Tesla, [PCr]/[ATP]

has been relatively easy to measure but often requires long scan times

to provide adequate signal-to-noise (SNR). In addition, pH which relies

on identification of inorganic phosphate (Pi), has rarely been obtained.

Significant improvement in the quality of cardiac P-31 MRS was

achieved through the use of the General Electric SIGNA" 3.0 Tesla whole

body magnet, improved coil designs and optimized pulse sequences.

Phantom and human studies performed with many types of imaging and


spectroscopy sequences identified breathhold gradient-echo imaging and

oblique DRESS P-31 spectroscopy as the best compromises among SNR,

flexibility and quality of localization. Both single-turn and quadrature

10-cm diameter, P-31 radio frequency coils were tested. The quadrature

coil provided greater SNR, but had to be used at a greater depth to

avoid skeletal muscle contamination. Gated cardiac P-31 MRS obtained in

just 6 to 8 minutes, showed both improved SNR and discernment of Pi

allowing for pH measurement.

A handgrip, in-magnet exerciser was designed, created and tested

at 1.5 and 3.0 Tesla on volunteers and patients. In ischemic patients,

this exercise was adequate to cause a repeated drop in [PCr]/[ATP] and

pH with approximately eight minutes of isometric exercise at 30% maximum

effort. As expected from the literature, this exercise did not cause a

drop in [PCr]/[ATP] for reference volunteers.


The statistics placing heart disease as a leading killer are

remarkable. Considering all age groups and genders, heart disease is

the number one killer, above other common killers such as cancer,

accidents, and diabetes.' Cardiovascular disease kills about 2,500

people each day or more than one million each year in the United

States.2 Some form of heart disease affects one in four persons, with

the combination of costs from treatment and loss of productivity

approaching 50 billion dollars annually.2

One of the most common types of heart disease is myocardial

ischemia. In myocardial ischemia, individual cells in affected areas

of the myocardium can no longer function due to significant decreases

in blood flow to the region, which results in chest pain (angina

pectoris). The blood flow reduction is typically due to a gradual

blockage in the large and/or small vessels, which substantially

increases the risk for acute and total blockage infarctionn). Ischemia

limited to the small vessels is termed microvascular ischemia or

microvascular dysfunction (MVSF). MVSF seldom results in death but is

very disabling due to fatigue and anxiety resulting from the chest

pain.' Although patients with chest pain are usually checked for heart

disease, asymptomatic patients are also tested based on risk factors

such as age, weight, life-style (high fat diet, smoking) and family

history. Ischemia is often clinically silent or associated with

atypical symptoms. Unfortunately its presence is a significant risk

factor for a fatal heart attack. The American Heart Association

estimates that as many at 3 to 4 million Americans have silent or

asymptomatic ischemic episodes that are eventually diagnosed by testing

for reasons unrelated to the symptoms.2

Statement of Problem

None of the current clinical methods for diagnosing cardiac

ischemia is 100% accurate. This is especially true of the most common

cardiac test, measurement of the heart's electrical function via

electrocardiogram (ECG). Most especially for women, but also for men,

the ECG often does not assist in the diagnosis of cardiac dysfunction.4

In addition, although large vessel ischemia is commonly quantified by

the degree of stenosis in the coronary arteries via a coronary

angiography (CA) catheterization study, microvascular ischemia cannot

be diagnosed with CA because the vessels are too small to resolve.

Of all the tests clinically available to diagnose myocardial

ischemia, most do not provide a direct quantitative measure of ischemia

in the affected myocardial tissue. Most of the tests that look

directly at the myocardial tissue are qualitative imaging studies, such

as clinical nuclear medicine, ultrasound, computed tomography, and

magnetic resonance imaging methods, where information such as wall

motion and perfusion defects can be qualitatively determined. However,

these evaluations do not provide quantitative information about levels

of ischemia, although the technique of MR tagging (now at a research

stage only) may be used in the near future to quantify wall motion.

Those tests that do provide quantitative information do so indirectly,

rather than as a direct measure of the myocardial tissue, such as the

electrical signal measured via ECG or the percent stenosis of the

coronary arteries as measured in a CA study. Individually, none of

these tests can provide a direct diagnosis of ischemia. In this

context, it is common practice to perform multiple tests for added

accuracy in diagnosing myocardial ischemia.

Spectroscopy's Contribution to Diagnosing Myocardial Ischemia

Human, cardiac phosphorus NMR spectroscopy (cardiac P-31 MRS) is

a non-invasive technique that directly measures pH and the levels of

intracellular myocardial phosphocreatine (PCr), adenosine triphosphate

(ATP), phosphodiester (PDE), and inorganic phosphate (Pi) in a non-

destructive manner. Both the reduction in the ratio of [PCr]/[ATP] and

pH have been shown to be sensitive and specific markers of ischemia at

the tissue level in animal models'6 and in humans.'8 Cardiac P-31 MRS

is an additional tool that can add to the degree of sensitivity and

specificity in diagnosis of ischemic heart disease by providing

quantitative information as a measure of the myocardial tissue

directly, regardless of whether it is caused by macro- or micro-

vascular ischemia.

Cardiac P-31 MRS is currently in limited use clinically. The

reasons for this pertain partially to the degree of difficulty in

obtaining cardiac P-31 MRS in a reasonable amount of time and

uncontaminated by non-cardiac signal. Useful cardiac P-31 MRS requires

obtaining a phosphorus spectrum with good signal-to-noise ratio (SNR),

with the Pi peak discernible in order to measure pH, and without

contamination from skeletal muscle or liver. The spectrum must be

obtained from the heart, which is moving in a rhythmic motion with the

cardiac cycle. In addition, spectral acquisitions are sensitive to BO

homogeneity and generally require more swimming than do images to

enhance the SNR. Shimming of a moving, heterogeneous object is more

difficult than a stationary one. To compound the difficulties, the

anterior region of the heart is only 10 to 18 mm thick and at a depth

of 4 to 8 cm from the P-31 surface coil. The distance between the

heart and the surface P-31 radio frequency (RF) coil limits the

acquisition volume to the anterior wall of the heart because of

attenuation of B, and low SNR. Between the heart and the P-31 surface

coil is the chest wall skeletal muscle, below the heart is the liver,

and in the cardiac chambers is blood. Care must be taken so that

neither the skeletal muscle nor liver will contaminate the spectra

obtained from the cardiac muscle. Blood contamination can be corrected

for after the acquisition. The result of a cardiac P-31 acquisition is

typically a low SNR spectrum with an acquisition time of around 10 to

40 minutes. Because in-magnet exercise and/or drug-induced stress are

required for P-31 MRS studies of ischemic heart disease, a scan time of

more than 10 minutes is undesirable.

Research Hypotheses

The following hypotheses will be tested in this dissertation.

1. Human, phosphorus NMR spectroscopy of the heart can be implemented

at 3.0 Tesla (T) by overcoming technical problems to produce spectra

of higher quality than obtained at 1.5 T.

2. Mild exercise from an isometric, hydraulic handgrip, designed and

produced for this study, provides adequate stress on the heart to

cause a significant drop in [PCr]/[ATP] in the ischemic myocardium

and thus differentiates ischemic from non-ischemic myocardium.

Specific Obiectives

Localized phosphorus spectra will be obtained from both phantoms

and the anterior myocardium of human participants. All human

participants will be screened for MR incompatibility and will sign an

approved Institutional Review Board Informed Consent Form (Appendix A).

The optimal pulse sequence for obtaining human, cardiac

phosphorus spectroscopy will be determined from phantom and human

studies. All spectroscopy pulse sequences available on the General

Electric (GE) 3.0 T whole-body magnet will be compared using phantom

studies. Measurement of SNR and degree of localization will be made

for each pulse sequence. Select pulse sequences will then be tested

and compared using human subjects. Human study comparisons will be

based on SNR and the degree of contamination of cardiac muscle signal

from other sources (skeletal muscle and liver).

A 10-cm diameter, single-turn P-31 RF coil will be compared with

a 16 x 10-cm, quadrature P-31 RF coil (two, 10-cm diameter RF coils

overlapping). Comparisons will be based on quality factors, isolation,

SNR at different depths, and degree of signal contamination.

Select tests will be performed on both 1.5 and 3.0 T GE Signa

AdvantageM systems to compare the resulting SNR and spectral dispersion

obtained using the same pulse sequence, coil, and sample studied

(phantom or human). The human cardiac phosphorus spectra obtained at

3.0 T should show greater SNR and spectral dispersion allowing for the

discernment of the Pi peak and thus the measurement of pH.

A hydraulic handgrip exerciser, designed and produced for this

study, will be tested for adequacy as a cardiac stressor that should

cause a drop in [PCr]/[ATP] in ischemic myocardium but not in non-

ischemic myocardium. The adequacy will be demonstrated by obtaining

data on heart rate and blood pressure changes with isometric handgrip

exercise of 30% maximum effort. These data will be compared with

literature values where similar devices were used during cardiac

phosphorus spectroscopy acquisitions as well as compared with the

responses to clinical cardiac tests (treadmill and dobutamine) obtained



This work will build on past knowledge and technology for

assessing cardiac metabolites and thus heart function. It will be

assumed, in most cases, that data from previous publications are

correct, especially at 1.5 T, and this data will be used to help

validate the work developed in this dissertation by comparing it with

past publications and work on similar patients and procedures performed


In order to assess the procedure used to stress the participants

in the magnet, both suspected ischemic and non-ischemic subjects will

be tested. It will be assumed that these participants will be properly

categorized by risk factor assessment, current symptoms, and when

available ECG treadmill and cardiac catheterization results.

Scope of the Proiect

This work will be limited to creating a procedure for obtaining

human, in-vivo cardiac P-31 NMR spectroscopy optimized for the GE Signa

Advantage'" 3.0 T whole body magnet located in the tunnel between Shands

at UF and the Veterans Affairs hospitals in Gainesville, FL. The

ability to improve the results of acquisitions on the 3.0 T scanner is

limited by the capabilities of the system, such as gradient strength (1

Gauss/cm) and specific absorption rate (SAR) limitations.

This work also will not use extensive pulse programming to

improve the system performance. Instead, the available sequences on

the 3.0 T system will be optimized, most often via the choice of

parameters and protocols. When necessary, small pulse programming

changes may be performed by Dr. Hee-Won Kim.

The numbers of subjects in the studies will be limited but should

still be adequate to demonstrate the feasibility and reproducibility of

the technique. A true set of controls free of ischemic heart disease

would first have to be evaluated by a cardiologist, at a cost that is

not available for this project. Taking any willing volunteer would

potentially introduce bias into the results as some might have silent

ischemia. In addition, a cardiologist will refer all participants

identified with ischemic heart disease. This work is intended as a

feasibility study where the results could then be used to justify

further research.


Human cardiac spectroscopy has been commonly performed at 1.5 T91'

and occasionally at 4.0 T,16." but never before at 3.0 T. With all

other factors the same, an increase in magnet field strength will

theoretically result in a linear increase in the spectral dispersion

and at least a linear increase in the SNR. In addition, the use of a

quadrature surface coil is expected to further increase the SNR

compared to the simpler single loop surface coil currently used in

previously reported human cardiac spectroscopy studies..-24

Finally, the use of in-magnet exercise while non-invasively

measuring human cardiac phosphorus metabolites has been accomplished by

just a few research groups. 13'2'25-28 In-magnet exercise with our

technique and device is easy to incorporate and looks very promising

for helping to distinguish normal from diseased cardiac muscle,

especially in cases of myocardial ischemia.

In terms of potential research and clinical use of human cardiac

phosphorus spectroscopy, there are many potential future benefits.

First, this test may provide increased accuracy of diagnosis of

ischemia and microvascular ischemia, for which new treatment modalities

may exist. Because cardiac P-31 MRS is a quantitative measure of

ischemia at the tissue level, it could be used to monitor or follow-up

treatment regimes or to help develop new treatments for ischemic heart

disease. In addition, a noninvasive MR diagnostic test of ischemia may

eliminate unnecessary invasive cardiac catheterization procedures,

reducing both risk of medical complications and cost. Based on data

from the Health Care Finance Administration, National Physician Fee


Schedule Relative Value File," a cardiac catheterization study costs

four times as much as a cardiac MR exam.


Magnetic resonance (MR) is a technique for using the interactions

of atoms and molecules with external magnetic fields to extract image

and chemical data from a sample. Using the classical description, the

proton possesses a spin angular momentum, S and a gyromagnetic ratio,

y, where the product is the magnetic dipole moment, .

S= yS Equation 1

Hydrogen and nuclei with either an odd number of protons (such as

phosphorus-31 with 15 protons and 16 neutrons) or an odd number of

neutrons possess magnetic moments whereas even-even nuclei have zero

magnetic moment. NMR takes advantage of the spin magnetic moment to

obtain a signal from these nuclei, especially if they are already

somewhat plentiful in the human body like proton (H-l) and phosphorus

(P-31) .

Placing a bulk of material, having nuclei with spin magnetic

moments, into a strong and uniform external magnetic field (Bo) causes

the nuclei's magnetic moment to attempt to align with the applied

magnetic field. This results in the spin processing around the

magnetic field analogous to a spinning top.30 The precession of the

nuclei in response to an applied magnetic field proceeds at a known

frequency described by the Larmor equation:

o = yB0 Equation 2

where c0 is the rate of precession in radians per second, and B0 is

the main magnetic field. The application of radio frequency energy at

the Larmor frequency for a nucleus constitutes a condition of

resonance, hence explaining the terms of Nuclear Magnetic Resonance or


In practice, the distribution of electrons about any given

nucleus provides some shielding of the nucleus from the Bo field.

Since the distribution of electrons is a function of the molecular

structure in which the atoms (and their nuclei) are located, the actual

field experienced by the nucleus differs from the Bo field by some

small amount.30 Equation 2 can be modified to include the shielding

constant for chemical shift, G for a nucleus in a specific molecular


io = yB0(1-) Equation 3

A basic NMR experiment consists of placing a sample in a constant

external magnetic field (Bo) where the nuclei's magnetic moment

attempts to align with Bo, also called the equilibrium condition.

After applying an oscillating radio-frequency (BI) identical to the

precession frequency of the nucleus of interest (the Larmor frequency),

the nuclei absorb energy and tip away from their alignment with the

external magnetic field. This displacement with the Bi field causes the

individual spins to coalesce, creating a combined magnetic moment in

the transverse plane. Once the BI field is removed, the magnetic moment

rotates at the Larmor frequency, slowly losing phase coherence due to

magnetic field variations affecting the individual spins." A signal is

induced in a receiver RF coil (either the same or different from the RF

transmitting coil) by the rotating magnetic moment in the transverse

plane. This produces a damped sinusoidally varying signal of positive

and negative polarity at the Larmor frequency, known as the free

induction decay (FID). Only when the net magnetic moment in the

transverse plane is nonzero, and the spins are in phase coherence will

a signal be generated."

The MR technique is both non-invasive and sensitive to the

molecular environments of the atoms. These factors have led to the use

of MR in several areas, primarily for imaging and spectroscopy. In MR

imaging (MRI) the signals from fat and water hydrogen atoms are mapped

according to their location and their characteristics. The second area

of use, MR spectroscopy (MRS), is a technique that was first used in

chemistry and physics laboratories in the analysis and identification

of chemical compounds.32 Biologically, MRS has also been used in the

identification and analysis of protein and macromolecular structures

and conformation using high resolution NMR." Another biological use of

MRS has been the examination of cell samples or organs, ex vivo." The

most clinically relevant use of MRS, however, is in vivo where the

patient's tissue metabolites can be examined directly and non-

invasively using techniques to localize to the region of interest

within the body.

The two main isotopes that are most studied in human, in-vivo

spectroscopy are proton (H-l) and phosphorus (P-31) because of their

natural abundance, relative sensitivity and chemical significance

within the body. H-1 MRS is the most commonly performed clinical MRS

patient examination because it can be done using standard MRI equipment

and software available on most clinical systems. It has been shown to

be useful in measuring markers of a variety of brain abnormalities such

as stroke,3' tumors6 and epilepsy,3 through the relative increase or

decrease of the metabolites (such as NAA, choline, and creatine). P-31

has a larger chemical shift range (-30 ppm) than H-1 (-10 ppm) and a

much lower sensitivity, as shown in Table 1. Water suppression is not

necessary in P-31 MRS, but lower spatial resolution and/or increased

scan times are required to obtain the same signal to noise (SNR) as

with H-1 MRS. P-31 MRS can detect a number of metabolites involved in

cellular energy metabolism, such as phosphocreatine (PCr),

adenosinetriphosphate (ATP) and inorganic phosphate (Pi). The chemical

shift of the Pi peak is pH dependent. In addition, relative peak areas

of PCr, Pi and ATP peaks have been used to evaluate fatigue and/or

ischemia in muscle, as will later be discussed in this chapter.

Table 1. Nuclear Spin Parameters".

Gyromagnetic Resonance Resonance
% Natural Ratio Frequency Frequency Relative
Isotope Abundance (MHz/T) at 1.5 T at 3.0 T Sensitivity

H-1 99.985 42.58 63.86 127.74 100

P-31 100 17.25 25.88 51.75 6.6518

P-31 Spectroscopy Pulse Sequence Options for the Heart

A variety of pulse sequences have been employed to localize to

the myocardium when performing in-vivo cardiac spectroscopy. The most

basic pulse sequence, a square excitation pulse, uses no gradients and

thus performs no localization, but instead takes in all signal within

the sensitive volume of the coil. This sensitive volume is generally

defined, for a surface coil, as the volume of a sphere of one coil

diameter." Such unlocalized acquisition methods have been used with

animal studies where open chest experiments allow for small surface

coils to be placed directly on the heart.40'.1 This uses the surface

coil as a localizer. For human studies, where we prefer to work non-

invasively, a combination of coil and pulse sequence localization

techniques have been used. This presents a technical challenge to

optimize a localization technique for obtaining spectra from a moving

heart, at a depth into the body, without adding signal from the chest

skeletal muscle that lies between the coil and the heart.

Slice Localization Techniques

Slice selection is accomplished in MR by simultaneously turning

on a slice-selective RF pulse and a gradient along the direction of the

slice. The gradient is on for just a few milliseconds. The RF pulse

designed for slice selection has a time-varying shape in the form of a

sin t/t or since function,424 which is used for both spectroscopy and

imaging. The RF pulse is generated at the resonance frequency of the

nuclei of interest. With the gradient on, this allows the since RF

pulse to excite just those frequencies in a narrow bandwidth to either

side of the center resonance frequency. A since pulse can produce a

sharp cutoff of frequencies thus exciting just within a slice region.

The thickness of the excitation slice is related to the bandwidth of

the RF pulse and the gradient strength. For example, the steeper the

gradient or the narrower the RF pulse bandwidth (the longer the RF

pulse) the thinner the selected slice. In terms of spectroscopy

acquisitions, the shorter the RF transmit pulse the larger the range of

frequencies acquired.

Slice localized pulse sequences, such as DRESS (Depth Resolved

Surface Coil Spectroscopy) ,4 SLIT-DRESS (SLice INterleaved DRESS) ,56"4

Rotating Frame MR,48 FROGS (Fast Rotating Gradient Spectroscopy)," and

1D-CSI (chemical shift imaging)," allow for acquisition of signal from

a slice at a depth parallel to the coil. DRESS for cardiac studies

involves acquiring from a coronal or oblique slice through the cardiac

muscle at a depth from the chest, thus avoiding contamination from the

skeletal muscle directly under the surface coil. Sometimes it requires

outer volume suppression to avoid contamination of the signal from the

adjacent skeletal muscle at the sides of the body. Most studies have

not used outer volume suppression, but relied on reduced regions of

coil sensitivity to prevent contamination from the sides of the body.

Because it is simple and not extremely motion sensitive, DRESS has

been widely used for obtaining spectroscopy slice profiles of human

cardiac muscle -12 1 '.24s51 56

Multi-Voxel Localization Techniques

The spectrum of the cardiac muscle is easily contaminated by the

surrounding skeletal muscle and blood, therefore localization and

suppression of unwanted signal outside the volume of interest is

important and accomplished with single- or multi-voxel localization

techniques. Multiple voxel acquisition via chemical shift imaging

(CSI) involves the use of the gradients to split up the selected slice

or slices into a number of smaller voxel areas. Spatial localization

is done by phase encoding gradients in one (1D-CSI), two (2D-CSI) or

three dimensions (3D-CSI)." There are no gradients on, however, during

signal acquisition. The advantage of CSI is that multiple voxels are

sampled as part of the same acquisition protocol. However, this is

accomplished at the cost of increased scan time for the phase encoding

for each axis. CSI is generally utilized to further partition a volume

already selected by another pulse sequence.

There are a number of technical problems associated with

acquiring spectra with CSI. One problem is achieving a good shim over

a large region of interest and possible changes in magnetic

susceptibility encompassing the multivoxel acquisition volume. In

addition, the point-spread function from the spatial Fourier transform

reconstruction implies that any given voxel in the multivoxel

acquisition contains contributions from neighboring voxels.3 This

leads to lesser ability to prevent contamination outside the voxel of

interest, such as the skeletal muscle contaminating the cardiac muscle.

In addition, depending on the size of the voxel of interest and

size of the field of view, the time of the acquisition can be

dramatically increased beyond the reasonable limit for performing in-

magnet exercise during the study. This is primarily due to the

increased time necessary for the CSI phase encoding steps. For

example, the total measurement time for a 3D CSI acquisition is given


Time (seconds) = TR Nacq Nx Ny N

Equation 4

where TR is the repetition time, Nacq is the number of acquisitions,

and Nx, Ny and N1 are the number of phase encoding steps in the x, y

and z directions."

Despite these faults with CSI, a number of studies of cardiac

phosphorus spectroscopy have utilized 1D, 11',227,'-64 2D,10'65s-7 and 3D

CSI,64'70-7 and techniques similar to CSI such as 3D fourier series

window.7 Among these studies, some were performed on animals

alone,62 69~'7374 but the majority were performed on humans.

Single Voxel Localization Techniques

Single voxel localization techniques rely on the pulse sequence

to obtain a localized voxel of data, selected from within a larger

volume. This protocol of obtaining a single voxel of data per

acquisition typically takes less time than the CSI multivoxel approach

because of the added phase encoding steps needed for CSI. Single voxel

acquisition pulse sequences that have been reported include PRESS

(Point Resolved Spectroscopy Sequence) ,7 PROGRESS (a version of

PRESS),7 STEAM (Stimulated Echo Acquisition Mode)," ISIS (Image

Selective In-Vivo Spectroscopy),s, 1'0-83 and modified ISIS pulse

sequences such as CRISIS (combination of ISIS and DRESS with x- and z-

selective 1800 pulses followed by a 900 y-slice selection pulse; which

eliminates y-direction motion artifacts and benefits from reduced cycle

time),84 FLAX-ISIS,41 and 2D ISIS plus Outer Volume Suppression (OVS).8"

With PRESS and STEAM, three slice selective RF pulses are used to

select three intersecting orthogonal planes. Only spins in the voxel

defined by the intersection of the three planes experience the three RF

pulses and contribute to the final signal. Both STEAM and PRESS

acquire a voxel with each acquisition (also called single-shot

localization),"8 thus being less motion sensitive than techniques

requiring multiple acquisitions to localize. STEAM uses a 900 RF pulse

for all three excitations thus creating a stimulated echo, as shown in

Figure 1. PRESS creates a double echo by using a 900 and two 1800 RF

pulses, as shown in Figure 2.38.86 Echo time (TE) for STEAM is

designated as twice the time between the first and second RF pulse.

PRESS has two echo times, TEl and TE2, where TE1 is equal to twice the

time between the first (900) and second (1800) pulses and the total time

to produce the first signal echo.3 TE2 is twice the time from the

first echo to the second 1800 RF pulse and the full time between the

first and second signal echo. In general, each TE for both STEAM and

PRESS must be relatively short to successfully acquire phosphorus

spectra, so that the signal losses due to transverse relaxation (T2) are

small. PRESS offers the advantage of having twice the SNR of STEAM,

but cannot be run with short TE's,"6 while STEAM can handle some shorter

TE's down to the system limit.

PRESS and STEAM differ primarily in the nature of the echo signal

created. PRESS forms the echo from 180 RF refocusing of the net

magnetization, whereas in STEAM, only part of the available signal is

used to form the stimulated echo via the use of 90 RF pulses. This

theoretically results in a factor of 2 increase in SNR for PRESS over

STEAM. Experimentally, the factor of 2 increase in SNR has also been


Both PRESS and STEAM have occasionally been accomplished

successfully at the phosphorus frequency, but never on the heart.

STEAM has been shown to work on a phantom (400 mmolar solution of

NasP3O0 in H20) at 1.5 T using a TE of 3.1 msec"8 and on human brain at

2.0 T using a TE of 3.0 msec." PRESS has also been used to obtain

phosphorus spectroscopy of a newborn human brain at 2.4 T using a TE of

10 msec.90 Note that in each of these cases, the TE was set at a

minimum value. Neither PRESS nor STEAM are well suited for phosphorus

MR spectroscopy due to reasons related to the characteristically short

T, times of phosphorus metabolites, but STEAM does work marginally

better than PRESS." The echo times achievable with PRESS are too

long for the acquisition of signals from nuclei with short T, relaxation

times." In contrast, STEAM's shorter TE values reduce signal loss from

T, relaxation and allow some observation of short T2 metabolites, such

as P-31 metabolites."

ISIS allows for volume, slice, column or voxel selection based on

one, two, four or eight acquisitions.92 Zero to three selective-

inversion pulses in the presence of gradients precedes each acquisition

exciting different areas within the volume. A combination of areas,

added and subtracted, results in the localization achieved with ISIS as

shown in Figure 3.7"9 ISIS has proved to be one of the best voxel

localization techniques for P-31 spectroscopy. Unlike PRESS and STEAM,

ISIS is not sensitive to the T2 values of the acquired resonances but is

sensitive to the Ti values, as part of the time delay between the

selective excitation and signal acquisition." Most P-31 biological T,

values are on the order of a few seconds, therefore obtaining in vivo

ISIS spectra from living tissue should not be a problem in terms of

relaxation times.9 Also, for any localized spectroscopy pulse

sequence, when long repetition times are used to allow for Ti

relaxation, there is no T, error in terms of signal cancellation. ISIS

may require eight separate acquisitions in order to localize to a

voxel, but the final SNR is still equivalent to an acquisition taken

with eight acquisitions (i.e. no signal is lost due to localization).

This is also true for CSI acquisitions. ISIS is also flexible in that

fewer than eight acquisitions can be used to obtain a volume (1

acquisition), a slice (2 acquisitions), a column (4 acquisitions), or a

voxel (8 acquisitions).

ISIS uses an adiabatic RF pulse that allows for uniform

excitation of signals over a larger volume of the sample.9u"96 The

adiabatic pulse, a modification of the since pulse, was specially

designed for use with surface coils where uniform excitation is

difficult to obtain.94

The adiabatic pulse is frequency selective and B, insensitive, and

can achieve uniform excitation in an area of B, field inhomogeneities.

The adiabatic pulse is designed using frequency modulation to satisfy

two constraints. First, the RF pulse is designed to cause negligible

decay of the transverse magnetization during the pulse. Second, the

rate of change of the net magnetization orientation is considerably

slower than its rate of precession."

In addition, adiabatic pulses are different from since or square

RF pulses in that they are designed to excite maximally at a high gain

and remain maximal even when the gain is further increased."'"96 Sine or

square RF waves excite an area based on the gain of the signal with an

optimal signal peaking at some point between low and high levels of

gain. Unfortunately, the ISIS adiabatic pulse require a greater power

than since or square pulses, therefore it is important to have a quality

amplifier designed for the higher requirements. Since power

requirements increase with field strength, special consideration needs

to be made to provide enough power for ISIS while still remaining

within the safety guidelines. Specifically, the Food and Drug

Administration (FDA) has limited the specific absorption rate (SAR) of

the RF power to a maximum of 8 Watts/kg in the head or torso for any

period of 5 minutes." This applies to all pulse sequences and all

field strengths used on human subjects.

ISIS is prone to many possible causes of signal contamination."

As ISIS selects outside the volume of interest and attempts to subtract

this signal, only partial cancellation of the unwanted signal is

achieved." This effect is greater when the volume of interest (VOI) is

much smaller than the complete volume detected by the coil. Since

signal is detected from a large volume after each acquisition, the

receiver gain cannot be optimized for the smaller VOI.98 In addition,

the adiabatic pulse does not acquire equally all the chemically shifted

resonance's at the acquired frequency." Finally, T, smearing" caused

by residual magnetization left from the observation pulse from the

previous acquisition, also results in less than total subtraction of

unwanted signal outside the ISIS selected volume of interest." ISIS is

also motion sensitive due to its volume subtraction method of

localization involving eight separate acquisitions to obtain the

spectrum from one voxel." In addition, ISIS is affected by spatial

displacement of chemically shifted species in the slice selection at

high field at the P-31 frequency."

The issues of contamination with ISIS have been tested in a

computer simulation and published." Despite its faults, only ISIS and

ISIS derived pulse sequences are useful as single-voxel localization

pulse sequence techniques for P-31 MRS. A number of cardiac

spectroscopy studies have been done using ISIS or ISIS derived pulse

sequences. "4",117',23.2s,80-82,1-117 All except one11 of this list of studies

was performed on humans.

Cardiac P-31 Spectroscopy Results in the Literature

Animal studies

Animal studies have been essential for predicting potential areas

for human use. A majority of cardiac spectroscopy animal studies have

been done with the animal's chest open, thus allowing the surface coil

to be placed directly on the heart,40,41 via a catheter coil inside the

heart,11 or by excising the heart and maintaining it artificially in an

isolated heart perfusion studys~51 122. There have also been closed-

chested animal experiments 67123'124 that more closely match what would

happen if the same test were performed on humans. The promise of the

open-chested experiments is that a great deal of information from the

heart via spectroscopy can be obtained when SNR is maximized.

Unfortunately, outstanding SNR is generally only achieved when the RF

receive coil is placed directly on the heart, allowing for unlocalized

acquisitions. The open-chested in situ mode can also take advantage of

the increased SNR and use localization acquisition methods, such as 2D

ISIS, 1D spectroscopic imaging (SI), or FLAX-ISIS, to obtain

spectroscopic information about discrete layers of the myocardium, sub-

endocardium and sub-epicardium individually. 4112s 26

Animal hearts can be stressed to a much greater degree while

under anesthesia 27 than we would ethically stress a human volunteer.

Although most human studies of normal myocardium have shown no change

in the phosphorus metabolites with the stress levels that do produce

changes in ischemic hearts, normal animal hearts have often been

stressed beyond what can be done in humans to the point where a change

in the phosphorus metabolites can be induced. Care must be taken,

therefore, when making extrapolation to humans based on these animal

studies, because often the heart is damaged by the level of stress or

by the deliberate infarction..24'128

Animal studies allow for tighter control of the disease state and

measurement than can be ethically accomplished in humans and therefore

reveal the potential for P-31 MRS to provide feedback of specific

cardiac disease states. Studies of permanent coronary occlusion in

dogs with open chest, measured up to 6 hours after occlusion.2 and up to

5 days after occlusion.24 show that residual Pi remains in the region of

the infarction for a period of days although tissue pH returns to

normal within a day.' Human cerebral infarction measured by phosphorus

MRS has shown similar behavior.7'2 A study on open-chested dogs showed

that phosphorus spectroscopy could differentiate between viable and

non-viable myocardium from 6 to 54 hours after an ischemic insult based

on measurements of PCr and [Pi]/ [PCr]." Another animal study of

assayed myocardium after coronary occlusion or low-flow ischemia showed

a depletion of PCr and ATP with irreversible injury.31 Animal studies

have shown reduced phosphorus metabolites in cases of hereditary

cardiomyopathy,132'.1 cardiomyopathy from chronic chemical exposure,134-'15

dietary deficiencies, 13 and diabetes3'. In another animal study, PCr

was shown to be preserved until the blood flow was reduced to about 50%

of that in an originally healthy heart."18 This is another example of

the type of understanding that can be achieved with animal studies but

would be unethical for human studies.

Specifically, both a decrease in the [PCr]/[ATP] ratio and a

downfield shift of the inorganic phosphate resonance due to acidosis

have been measured occurring with the onset of ischemia.56 In cases

with mild reductions in blood flow (approximately 17%), only the Pi and

pH changed significantly when compared to a control group. When the

blood flow reduction was more substantial (on the order of 50% or

more), the [PCr]/[ATP] ratio was also shown to decrease while ECG

monitoring was also abnormal (reduction in segment shortening).139'14

Human Studies

Nuclear medicine thallium scans clinically show a reversible

thallium defect where the area fills with thallium during rest but does

not during stress indicating an area of ischemia. If the defect is not

reversible, meaning the area doesn't fill with thallium during rest or

stress, then the area is infarcted. P-31 exercise tests on subjects

with severe CAD and/or reversible thallium defects showed a significant

decrease in the [PCr]/[ATP] ratio.12,49 Although a change in the

[PCr]/[ATP] ratio is seen in the ischemic heart with stress, tests on

normal controls and patients with non-coronary cardiomyopathy show no

significant change with exercise. In patients with stenosis of the

macrovessels, a drop in the [PCr]/[ATP] ratio is no longer present

after revascularization." These findings prove that in the absence of

blood flow reduction or scar to an area of the myocardium, the

metabolite energy values will remain constant. These studies showed a

direct correlation between ischemia and a decrease in [PCr]/[ATP] and

pH with stress as measured via phosphorus NMR spectroscopy.

A number of research groups from a variety of locations around

the world have been utilizing human, in-vivo cardiac NMR P-31

spectroscopy. Assorted magnet systems, coils and techniques have been

employed. In addition, a wide number of patient types have been

studied. Even through there is a wide variation in heart problems

studied, some trends in the type of techniques and RF coils used have

started to appear that may prove diagnostically and clinically

feasible. In the least, P-31 MRS should provide information that

scientists and physicians can utilize to better understand the workings

of the human heart under all types of disease conditions.

A condensed summary of all the research locations that have

conducted in-vivo cardiac phosphorus NMR spectroscopy acquisitions of

the human heart are listed in Table 2. This table is given in an

effort to simultaneously demonstrate how much and how little work has

been done since the late 1980s. Since each type of cardiac problem

will have its own trends and results in terms of cardiac P-31 MRS,

research on each of the main types of heart disease has also been

detailed in the remainder of this chapter.

Table 2. Research Published on Human, In-Vivo

Spectroscopy of Studies on

Normal-Controls or

Cardiac P-31 NMR
Patients and Related

Major Magnet P-31 P-31
Location Players Tesla Method Coils References



Otsu, Mie,

New York




1.5 T ISIS
Philips CSI

diam R/T

1.5 T DRESS 5-15cm
GE CSI diam R/T,
20cm T +
12cm R

1.5 T DRESS 40cm T +
GE CSI 7cm R,
7cm diam

2.0 T PMRFI 5-6.5 cm
Bruker CSI diam R/T,
DANTE 15cm T +
7cm R

1.5 T ISIS 9-14cm
2.0 T CSI diam R/T

1.5 T CSI
Philips ISIS

Birmingham, Evanochko 1.5 T ISIS
Alabama Hetherington Philips
den Hollander 4.1 T

117 141 142
82 14 71 105
68 100 143
111 109 65
104 10 144
113 145 18

56 55 147 148
52 54 13 51
12 27 60 53

9 149 70 63
67 49 11 150
151 152 153
154 20

48 155 25 156
157 57 58 158
159 59 21 160

83 15 161 64
162 163





de Roos
den Hollander

Minneapolis Menon Uc

Philadelphia Whitman

Durham, Herfkens

Paulo, Kalik-Fi

Gainesville, Bruner
Florida Scott, Kim

purbil 4.0 T CSI
Siemens ISIS

1.9 T Coil
Oxford localized

S 1.5 T CSI

lbo 1.5 T ISIS

1.5T GE Oblique

10-14cm 116 164 108
diam R/T 165 28 22 166
167 112 114

10-14cm 103 106 169
diam R/T 81 102 170
171 80 28 16

llcm diam

5cm diam

6cm diam


10cm diam,

17 76




174 175

P-31 Spectroscopy Acquisition Techniques. In terms of techniques

for acquiring human, in vivo, cardiac P-31 MRS, there is a trend that

emerges. For those locations with clinical Philips systems, the most

common technique used is ISIS followed by CSI. For clinical GE sites,

the most common technique is DRESS followed by CSI. The two Siemens

sites conducting human, in-vivo P-31 MRS primarily utilize CSI.'144'17

This chapter summarizes the results from studies of various types

of cardiac diseases. When available, the acquisition times are listed

in Table 3 through Table 7. Notice that DRESSn'149 and ID-CSI"49"' are

the shortest duration sequences, generally being under 10 minutes,

although running them for longer further increases SNR. ISIS typically

is run for slightly longer periods of time at around 20

minutes,B2.'10.16s,17 although it can be run the same amount of time as

DRESS at around 10 minutes1'. Two dimensional and 3D-CSI are the

longest duration,1" with some scans lasting an hour for just a single P-

31 acquisition'".

A few truly unique techniques for obtaining human cardiac P-31

MRS, SLOOP and PMRFI, are utilized only at their originating locations.

Spectral localization with optimal pointspread function or SLOOP was

created and used by a group in Germany, utilizing a Siemens 1.5 T

system. a'6B1'411314 The SLOOP technique is a combination of a 3D-CSI

acquisition with sophisticated post-processing that uses anatomical

information from the proton image to group the spectra by tissue type.

Anatomical compartments, such as for cardiac muscle, skeletal muscle,

and liver are defined with the proton image. The overlaying P-31

spectral data is then separated into each of these compartments. The

papers that have been published from this group using SLOOP seem

outstanding as they achieve cardiac spectra with greater SNR, due to

the larger volume they can utilize to obtain the signal. In addition,

they seem to have solved the problems with contamination.

Unfortunately, until the technique is tried and evaluated at other

locations, judgment on the true accuracy of the method cannot be fully


Phase modulated rotating frame-depth imaging selection technique

or PMRFI is a technique used for cardiac P-31 MRS solely at Oxford,

England.4's .1s5' PMRFI allows for a 2D-spectroscopy signal to be

collected from the sample within the sensitivity of the coil via a set

of free induction decays. The data is collected from a volume formed

by a stack of disc shaped slices at various depths into the chest with

the entire data set taking approximately 35 minutes to acquire."4 The

PMRFI technique claims to produce high-resolution, high SNR spectral

images with limited spatial distortion."

In addition, at Oxford, they have also utilized DANTE for cardiac

P-31 MRS of human volunteers."' DANTE allows for the selective

excitation/suppression of individual peaks in the frequency domain in

this case used to distinguish the intracellular and extracellular Pi.

This group also utilized an initial saturation pulse to eliminate any

contamination from skeletal muscle."' Many of these types of unique

pulse sequences were originally developed on research magnets for

animal studies and then recently extended to human use on a Bruker

research magnet system. This may be the reason why these techniques

have not been published for use on clinical 1.5 T systems, such as from

GE, Philips or Siemens. Users often have the ability to pulse program

clinical systems, but with less flexibility and ease in programming


P-31 Radio Frequency Coils. Surface coils have been used for all

human cardiac spectroscopy studies to date. The reasons are simple.

Compared to a volume coil that would need to encompass the chest,

surface coils provide a much greater signal-to-noise ratio (SNR) than

the volume coil. The simplest surface coil, a single loop of wire,

provides a means of obtaining spectra from a volume of tissue adjacent

to the coil.4 The sensitivity of the surface coil is greatest for the

tissue closest to the coil, with sensitivity decreasing rapidly with

depth.4 This sensitivity range extends approximately one radius away

from the center of the surface coil," with the majority of the signal

obtained from a disc-shaped region at the plane of the coil and

decreasing in size with depth.4 Typically, the result is relatively

good SNR from the anterior wall of the heart, but not enough depth

penetration to cover the entire heart. Surface coils can be adapted to

a variety of specific tasks by changing the size and shape of the coil.

Surface coils are commonly used in both MRI and MRS as receivers

because of their high sensitivity, providing good SNR. In

spectroscopy, especially with nuclei other than H-l, surface coils are

frequently used both as RF transmitters and receivers." Their

applications evolved from small animal P-31 spectroscopy studies17 to

spectroscopy investigations of muscle and superficial organs in

humans.'" Unfortunately, surface coils suffer from non-uniform RF

excitation and inadequate spatial localization when used without a

localized pulse sequence. The sensitivity of the surface coil to a

point in the sample is proportional to the B, field achieved by the coil

at that point, with the most signal received from sample points closest

to the coil.41"'17 Fortunately, pulse sequences designed for

localization, utilizing selective RF pulses or gradients such as

ISIS179 '.8 and CSI,B1-1'B or multiple RF pulses"'8 can localize spectra to

regions where the B, field is relatively constant such as in a localized

voxel volume."'7 For example, the ISIS pulse sequence uses an adiabatic

excitation pulse that with enough power is specifically designed to

excite spins in a manner that is independent of the B, field for more

uniform excitation.41 Such a pulse has been shown to improve the

localization abilities when used with a surface coil."4 In addition,

there is the option of using a larger transmit coil to provide better

homogeneity of the B, field, with a smaller receive coil for optimal


The B, inhomogeneity inherent with using a surface coil can

degrade the performance of the RF pulses resulting in an incorrect

tipping angle, decreasing signal from the volume of interest (VOI), and

potentially increasing signal contamination from regions outside the

VOI.15, 'I9,8 Contamination is most probable from the tissue closest to

the coil since such tissue will have the highest degree of sensitivity

by the coil." In the case of cardiac spectroscopy, the skeletal muscle

has a large potential to contaminate the cardiac muscle signal in this

manner. Larger coils provide higher homogeneity at the expense of

reduced sensitivity to small, shallow VOI's.93'185 This compromise

between increased-size resulting in increased homogeneity versus

decreased-size resulting in increasing sensitivity has led to many

different sizes of transceive and a few combinations of transmit and

receive coils being built.

A trend that appears when looking at Table 2 involves the design

of the P-31 coils. The greater majority (over 95%) of human, in-vivo,

cardiac P-31 MRS patient studies are conducted using a simple single

turn P-31 coil of 5 to 15 cm diameter to act as a transceive radio

frequency (RF) coil. Beyond this, the next largest minority uses a

larger transmit (15 to 40 cm) to excite and a smaller 6.5 or 7 cm

diameter coil to receive.

The purpose of a larger RF transmit coil used with a smaller

receive coil is to improve the homogeneity and still have a small

spatial sensitivity.7 The two coils can be arranged coplanar'86 or

slightly displaced.4 The sensitive volume is determined by the overlap

region of the two coils.' Unfortunately, there are more drawbacks to

using separate transmit and receive coil for localized phosphorus

spectroscopy. The complexity of the setup creates difficulties when

trying to control accurately the position and thickness of a plane or

volume of a voxel. Multiple acquisitions are required in order to

localize the signal.' This is due to the fact that the placement of the

RF coil is critical within a few centimeters over a prime spot anterior

to the heart. Often, the placement is not correct the first time

and/or the pulse sequence localization position and/or transmitter gain

must be modified for better localization to the cardiac muscle.

There are also designs in the literature for quadrature coils and

array designs for human P-31 cardiac spectroscopy with a theoretical

improvement in SNR, but there is no known literature on such coils

being used with patient studies to date. Phased array and quadrature

coils offer the advantage of a larger sensitive coil region without

increased acquisition time or a decrease in SNR.152'18' Quadrature and

phased array coils have been used more extensively in imaging than

spectroscopy modes of human in-vivo data acquisitions. Several

experimental designs have been considered for cardiac imaging and

spectroscopy including (1) a 4-coil, diamond shaped design by Hardy et

al.,"2 (2) two pairs of surface coils placed on the chest and back,

surrounding the heart by Constantinides et al.,"' (3) a single pair of

coils designed for placement on the chest, ~."' and (4) a double-tuned

quadrature surface coil by Menon et al.76 There has also been a

theoretical study performed to determine the optimum configuration of 2

to 10 circular coils combined in a phased array system designed for

cardiac imaging."' This paper argued that theoretically the best

results would be obtained from a 4 coil array, producing 560% signal

improvement relative to a whole body coil and up to 360% improvement

over some commercial cardiac imaging coils.19o

Cardiac RF surface coils are placed on the front of the chest,

over the heart. Coil placement is critical for optimum signal from the

intended volume of interest. Incorrect coil placement can result in

decreased signal from the desired VOI.' This is especially true of

small (5 to 10 cm diameter), transceive, single-turn surface coils.

Incorrect coil placement can also increase contamination from areas

outside the VOI.' The problem with positioning can be theoretically

minimized by using a phased array or quadrature coil that will cover a

larger area with the improved SNR up to square root of two.7,'12,191

Patient Studies. In a normal, disease-free heart, a phosphorus

spectrum will show a consistent level of each of the phosphorus

metabolites"'" ~'1"'63 of phosphomonoester (PME) inorganic phosphate

(Pi), phosphodiester (PDE), phosphocreatine (PCr), and the three

adenosine triphosphate peaks (y a -ATP). Conversely, in patients

and animals with myocardial infarction or myocardial ischemia, in vivo

phosphorus cardiac spectroscopy will demonstrate either a reduction in

the overall amounts of each metabolite192"'1 or a change from the normal

ratios of one metabolite to another12'4'60"124'194 Such a measurement has

the potential to differentiate a healthy heart from one with myocardial

infarction or ischemia. In addition, some cardiac diseases, such as

cardiomyopathy or hypertrophy, both involving thickness changes of the

heart wall, are easier to diagnose with cardiac images. Even these

types of heart problems can potentially benefit diagnostically from the

added feedback of the chemical information provided by P-31 MRS.

Patient studies have proven over and over again that P-31

spectroscopy is a sensitive marker for clinical cardiac disease

states.82'1 A quick overview will be presented here to prove how solid

the P-31 spectroscopy measurement has been shown to be. As part of a

study on patients with severe stenosis of the left anterior descending

coronary artery (greater than 70% blockage), the [PCr]/[ATP] ratio

decreased significantly with exercise." A group of normal controls and

patients with non-ischemic heart disease showed no significant change

in the [PCr]/[ATP] ratio when they underwent the same stress as part of

the same study. A repeat P-31 NMR spectroscopy measurement was made of

five of the patients with severe stenosis of the left anterior

descending coronary artery after revascularization. In all five cases,

a significant improvement in the [PCr]/[ATP] measurement was seen in

terms of a smaller decrease in [PCr]/[ATP] during in-magnet stress. In

a similar study by Yabe et al.,12 P-31 NMR spectroscopy with in-magnet

exercise was performed on patients with ischemia, patients with

blockages infarctionn), and normal controls. The patients were first

categorized based on results from radionuclide testing with thallium-

201. Those patients with reversible thallium defects (ischemia),

showed a reduction in [PCr]/[ATP] during in-magnet handgrip exercise,

that normalized during rest. Those patients with irreversible thallium

defects infarctionn or dead myocardial tissue) and normal controls

showed no change in the [PCr]/[ATP] ratios with exercise. It is

believed that the [PCr]/[ATP] ratio does not change in cases of pure

infarction because the affected tissue is completely dead while the

remaining tissue is essentially sound. Therefore, the resulting signal

is less overall due to a lesser amount of viable tissue, but the ratio

remains the same. It is theorized that more complex techniques that

would measure the absolute quantity of the metabolite levels would be

able to differentiate even myocardial infarction as the overall amounts

would be lower, assuming measurement of a known volume of cardiac


Healthy Volunteers. Multiple studies have confirmed that for

healthy volunteers free of substantial coronary artery disease, in-

magnet exercise stress tests with non-drug stimuli have not produced a

significant change in the anterior myocardial [PCr]/[ATP]

ratio. 12'25s'49.s~, 16 Such results are shown in Table 3 where both leg

and handgrip exercises show no drop in [PCr]/[ATP] with normal control

groups. These results have been repeated in animal studies, without

handgrip exercise. No significant change in the [PCr]/[ATP] ratio was

found in the normal intact dog heart over a five-fold range of rate-

pressure products.127 In this case the stress was in the form of a

pacemaker which increased the rate of the heartbeats by threefold and

increased the rate-pressure product by 1.5.

There is however, the potential to use drug stimulus to stress

the healthy human heart to the point that a drop in [PCr]/[ATP] can be

seen. The first attempt at doing this was by Schaefer et al.16 where 2

to 16 pg/kg/min of dobutamine was used. Dobutamine has the effect of

increasing the heart rate, contractility and blood pressure. In this

case, this amount of dobutamine alone was enough to increase the rate-

pressure product (heart rate times systolic blood pressure) in normal

controls from 7,000 to 15,000. Despite this increase of stress on the

heart with dobutamine, the [PCr]/[ATP] at stress was not significantly

different from [PCr]/[ATP] at rest for the normal control group, as

shown in Table 3. In the same study a group of dilated cardiomyopathy

patients under the same stress also did not produce a significant

change in [PCr]/[ATP] with dobutamine stress ([PCr]/[ATP] at rest =

1.630.24, with drug = 1.570.24, p=0.38). Two other published

studies repeated these experiments with a slightly different protocol

and succeeded in dropping the [PCr]/[ATP] ratio even in normal

controls. Lamb et al.112'16 and Pluim et al.,16e both of Oxford, England,

first utilized 10 pg/kg and 0.03 mg/kg, respectively, of atropine

sulfate to block the cholinergic nervous system which will allow for

increased heart rates. They then both used 10 to 40 pg/kg/min of

dobutamine to achieve a steady heart rate, based on subject age. This

protocol produced a significant drop in the [PCr]/[ATP] ratio of the

normal controls being studied.

Table 3. Cardiac [PCr]/[ATP] Ratio of Healthy Volunteers (Normal
Controls) at Rest and During Stress.

[PCr]/[ATP] [PCr]/[ATP] Magnet P-31
at Rest at Stress Stress Form Tesla Method Reference
1.5+0.2 1.58+0.14 leg exercise 1.9 T Not 25, 157
(n=6) listed

1.51+0.03 1.51+0.03 Bicycle 1.5 T DRESS 13
(M=6) ergometer
1.72 0.15 1.74+0.17 handgrip 1.5 T 1D-CSI 49
(n=ll) exercise 5-14 min

1.77+0.16* 1.74+0.19* handgrip 1.5 T 1D-CSI 63
(n=8) exercise 8-16 min

1.80+0.28* 1.84+0.26* handgrip 1.5 T DRESS 54
(n=ll) exercise

1.85+0.28* 1.900.23* handgrip 1.5 T DRESS 12
(n=11) exercise 7-8 min
1.86+0.17 1.900.22 dobutamine 2.0 T 1D-CSI 163
(M=14) (drug) 7 min

1.420.18* 1.220.20* Atropine/ 1.5 T 3D-ISIS 112,167
(F=2, M=18) dobutamine 10 min

1.41+0.18* 1.16+0.13* Atropine/ 1.5 T 3D-ISIS 168
(M=12) dobutamine
* blood corrected; M = Male; F = Female

There has been one interesting study by Lamb et al.28.'66 where

reproducibility of each method and comparisons of the results from

different methods were determined for the Philips 1.5 T system. Using

16 normal controls, the same area of the anterior left ventricle

produced a [PCr]/[ATP] ratio of 1.3110.19 for ISIS alone, 0.98+0.20

for CSI alone and 1.41+0.20 for ISIS plus CSI. Lamb et al. concluded

that the CSI contained liver contamination resulting in a low

[PCr]/[ATP] ratio. In addition, the intra-examination difference of

repeated studies of each volunteer produced a smaller difference than

the inter-examination, volunteer-to-volunteer [PCr]/[ATP] difference.

Myocardial Infarction. In some of the studies in patients with

myocardial infarction, the phosphorus spectrum shows a normal ratio of

[PCr]/ [ATP] 82.9.'93 but the overall concentrations of the phosphorus

metabolites are lower than normal .2'3 as described by Bottomley et al.7

and Luney et al.8' These characteristics are theorized to occur since

the dead myocyctes can not contribute a metabolic signal to the

observed spectrum.7 The observed PCr and ATP signals are derived from

surviving myocytes surrounding or interspersed with the infarcted

myocytes.2~049 More recent publications are providing evidence that the

[PCr]/[ATP] ratio can be lower for infarction than in normal

controls,12jss o0. and reduce [PCr]/[ATP] slightly during stress'~"4. It

has been suggested that the reduction in the [PCr]/[ATP] ratio for the

patients with myocardial infarction represents an ongoing metabolic

stress in the myocytes remaining within the scarred region of the

myocardium." It is still not clear how the expected ratio of the

myocardium with infarction should present, but it is known that the

amounts of the metabolites will differ significantly from normal.

Therefore, measurements of the absolute concentrations of the

phosphorus metabolites, rather than the ratios, should provide

appropriate characterization of myocardial infarction.

The published results for [PCr]/[ATP] of myocardial infarction

are summarized in Table 4. Notice that there is not a significant

change in [PCr]/[ATP] with handgrip exercise. Table 5 compares the

amounts of [PCr] and [ATP] in normal controls to patients with

myocardial infarction. Observe that in more than one case the overall

amount of each metabolite (pmol/g) is significantly reduced for cases

of myocardial infarction60 'i'1" compared to normal controls."D0"' There

is even a slight reduction in the amount of [PCr] and [ATP) for the

case of myocardial ischemia."o This data shows that P-31 cardiac

spectroscopy has the potential to evaluate myocardial viability.

Table 4. Cardiac [PCr]/[ATP] Ratio of Patients with Myocardial
Infarction at Rest and During Stress.

Degree of [PCr/ [ATP] [PCr]/[ATP] Magnet P-31
Disease at Rest at Stress Tesla Method Reference
Infarction 1.6 0.2 1.5 T DRESS 149
After (ENDO) 2.5-9 min
angioplasty (M=2, F=2)
and drug

Infarction 1.8+0.2 1.5 T DRESS 149
After (EPI) 2.5-9 min
angioplasty (M=2, F=2)
and drug

Myocardial 0.48+0.21 1.5 T 2D-ISIS + 81 **
"Scar" (M=5) 1D-CSI

Chronic Normal 1.5 T ISIS 82
anterior wall (M=6) 32 min

Chronic Normal 1.5 T ISIS 82
posterior (M=4) 32 min

Fixed T1-201 0.940.41* 1.5 T DRESS 60
defects (M=8, F=4) 12-15min

Fixed T1-201 1.18 0.28 1.12 0.24 1.5 T DRESS 54
defects (n=12) handgripp)

Fixed T1-201 1.240.28 1.19+0.28 1.5 T DRESS 12
defects (M=9, F=3) handgripp) 7-8 min
* = blood corrected; M = Male; F = Female
ENDO = localized to endocardium; EPI = localized to epicardium;
** = 3 out of 5 patients had a prior history of heart failure

Table 5. Cardiac [PCr]
Myocardial Infarction,

and [ATP] Amounts at Rest in Patients with
Ischemia and in Normal Controls.

Patient [PCr] [ATP] Magnet Quantify P-31
Type pmol/g pmol/g Tesla to: Method Reference
Infarction < normal < normal 1.5 T External DRESS 193

Infarction < normal < normal 1.5 T External DRESS 192

Infarction 3.94+2.21 4.351.52 1.5 T External DRESS 60
(M=8, F=4) Standard 12-15 min

Ischemia 7.64 3.00 6.35 3.17 1.5 T External DRESS 60
(M=22, F=7) Standard 12-15 min

Normal 11.7 2.5 7.2 1.2 1.5 T External 1D-CSI + 67
Standard 2D-Phase

Normal 12.14 4.25 7.72 2.97 1.5 T External DRESS 60
(n=11) Standard 12-15 min

Normal 10+2 5.8+1.6 1.5 T Internal 1D-CSI 20
(n=21) Water
M= Male; F = Female

There is also potential value in observing the Pi and pH of

patients with myocardial infarction. One study at 1.5 T of six

patients with anterior myocardial infarction, with blood contamination

corrections performed on the spectra, showed a slight but non-

significant elevation of Pi in the patient set.49 This study was early

on in human, cardiac P-31 MRS, used only a simple 1D-CSI acquisition,

and may have mistaken blood contamination for Pi. There are not many

studies that follow this one in identifying changes with Pi at 1.5 T.

Despite this fact, there is interest in identifying the Pi peak as a

means of identifying a decrease in the pH of the myocardium that would

also be a sign of ischemia.

Difficulty with blood contamination has made the measurement of

Pi more complicated in human cardiac spectroscopy studies. This is due

to the fact that at 1.5 T or below, the Pi is most often overlapped by

the 2,3-DPG (2,3-diphosphoglycerate) peaks from the blood

contamination. The Pi has been repeatedly seen at 1.5 T with the use

of specialized pulse sequences such as DANTE selective excitation,"9

nuclear Overhauser effect (NOE) for signal enhancement, magnetization

transfer," and proton decoupling for enhanced spectral resolution."

These options are not currently available on the GE 3.0 T system.

Also, long acquisition times can increase the probability of

visualizing the Pi peak by increasing the SNR. These problems are

overcome at 3.0 T. Work at 3.0 T allows for the Pi peak to be

distinguished even over relatively short scan times (i.e. 6 to 8

minutes) due to the enhanced SNR and wider spectral dispersion achieved

at the higher field strength.

Myocardial Ischemia. One of the most common types of heart

disease is cardiac ischemia. In the diseased heart affected by

myocardial ischemia, individual cells in affected areas of the

myocardium can no longer function due to significant decreases in blood

flow to the region. The blood flow reduction is typically due to a

gradual blockage forming in the coronary arteries. These patients have

symptoms of angina which are usually temporary and brought on by stress

when the required blood flow to the heart is inadequate. Although

typically these patients will suffer from angina and fatigue without

triggering a deadly heart attack, a diagnosis of ischemia does

substantially increase the risk of a heart attack. Since the blood

flow is already reduced, there is greater potential for acute and total

blockage infarctionn). The result is a disorder that is very disabling

partially due to the fatigue but also due to the anxiety related to the

chest pain.

The typical diagnosis of cardiac ischemia is based on signs,

symptoms and laboratory tests that can be performed by a family-

practice or emergency room physician. Cardiologists and radiologists

at the local hospital can also perform further diagnostic exams. In

addition to patients with angina being tested for ischemia,

asymptomatic patients at risk for heart problems based on risk factors

such as age, weight, life-style (high fat diet, smoking) and family

history are also tested. This is due to the fact ischemia is often

clinically silent or associated with atypical symptoms. The AHA

estimates that as many at 3 to 4 million Americans have silent or

asymptomatic ischemic episodes that are eventually diagnosed by testing

for reasons unrelated to the symptoms, such as a routine physical


Ischemia is primarily quantified by the identification of

stenosis in the larger vessels of the coronary arteries, but ischemia

can also be confined to the microvessels. Microvascular ischemia can

involve angina-like chest pain that is coincidental with exertion and

thus may resemble typical angina pectoris but without stenosis of the

main coronary arteries. More often, however, the chest pain associated

with microvascular ischemia has characteristics that differentiate it

from typical angina. For example, the pain is generally prolonged,

repetitive, occurring at night, and poorly responsive to rest and

medications. Dr. Carl Pepine, a cardiologist at the University of

Florida, was one of the first to observe that patients' with symptoms

of microvascular disease are most often not predictive of a life

threatening disorder. Whereas in age-matched patients with coronary

artery disease such symptoms are indications of a life-threatening

event.' Despite microvascular disease not being life threatening, it

still has a great impact on the quality of life. In addition,

undiagnosed conditions, lethal or not, traditionally place added costs

on the medical system as the patient will continue to seek diagnosis,

go to new physicians and have more and more tests performed. Often the

quality of life is diminished to an extent that many remain unemployed

or retire from work and limit their activities, with obvious

socioeconomic implications. From a clinical and financial point of

view, these patients need to be diagnosed and treated.

In cases of myocardial ischemia, the localized phosphorus

spectrum will show a small level of depletion in the [PCr]/[ATP] ratio

at rest, compared with reference normal controls (i.e.: Ischemic

Anterior Myocardium [PCr]/[ATP]: 1.450.31, n=16; Disease-Free Controls

[PCr]/[ATP]: 1.720.15, n=11)." A more significant difference occurs

when comparing the [PCr]/[ATP] ratio during minor stress test (leg or

handgrip exercise) to that with the resting value, as shown in Table 6.

Similar results have been shown in both animal124t'1 and human


Table 6. Cardiac [PCr]/[ATP] at Rest and Stress in Patients with
Myocardial Ischemia.

Degree of [PCr]/[ATP] [PCr]/[ATP] Magnet P-31
Ischemia at Rest at Stress Tesla Method Reference
>= 70% 1.45 0.31 0.910.24 1.5 T 1D-CSI 49
stenosis (n=16) handgripp 5-14 min

>=70% 1.46+0.39* 0.94 0.28* 1.5 T 1D-CSI 150
stenosis (M=14) handgripp 8-16 min

>=75% 1.560.19 0.94+0.27 1.5 T DRESS 54
stenosis (n=15) handgripp

>75% 1.600.19* 0.960.28* 1.5 T DRESS 12
stenosis (M=1, F=4) handgripp 7-8 min

*= blood corrected; M = Male; F = Female

Table 7 demonstrates the descriptive value of cardiac P-31 MRS

for ischemia that has been treated. The first two rows show a study

where five ischemic patients were tested with and without in magnet

exercise handgripp) after revascularization surgery. In the case of

the patients with revascularization, the [PCr]/[ATP] ratio does not

drop when exercised, although the same patients did drop their

[PCr]/[ATP] ratio prior to surgery.49 This provides a quantitative

means for evaluating the success of revascularization surgery. Another

method for treating a heart attack that is thought to be primarily

ischemic is to provide drug therapy intravenously, such as thrombolytic

agents. The third row shows such a study where the region of the

heart, even after being treated, remained in a stunned state, where the

wall motion in the region is still impaired even after intervention.17

In this case, investigators were able to show that despite the

myocardium remaining stunned, the tissue [PCr]/[ATP] values were not

significantly depleted as they would normally be under an ischemic

situation under stress, showing the potential of cardiac P-31 MRS as a

feedback measure after treatment."' Unfortunately, no pre-treatment

data was obtained so it is difficult to conclude that the ratio of

[PCr]/[ATP] alone was a marker for ischemia. The last two rows show a

study done only at rest of ischemic patients before and after

angioplasty, which showed no change in [PCr]/[ATP]. This study would

have been more effective if in magnet exercise or drug stress was used

during the [PCr]/[ATP] acquisition.

Table 7. Cardiac [PCr]/[ATP] at Rest and Stress in Patients with
Myocardial Ischemia with Some Type of Intervention.

Disease Description [PCr] / ATP] [PCr] / ATP] Magnet P-31
at Rest at Stress Tesla Method Source
Ischemia before 1.510.19 1.020.26 1.5 T 1D-CSI 49
revascularization normal < normal 5-14 min

Postischemic after 1.600.20 1.620.18 1.5 T 1D-CSI 49
revascularization normal normal 5-14 min
Drug infused 1.51 0.17* 1.5 T ISIS 173
Postischemic, normal 22 min
Stunned Myocardium
(M=15, F=6)
Ischemia >= 75% 1.50.7* 1.5 T DRESS 51
stenosis before normal 15 min

Postischemic after 1.41.0* 1.5 T DRESS 51
angioplasty normal 15 min
* blood corrected; M = Male; F = Female

Microvascular Ischemia Study: WISE. An NIH sponsored study

concentrated on women's ischemic syndrome evaluation (WISE) is studying

microvascular ischemia in women with phosphorus NMR spectroscopy as

part of a multi-center study including the University of Florida. The

specific feasibility of phosphorus NMR spectroscopy to look at

microvascular ischemia was initially demonstrated during a pilot study

performed on a 1.5 T Philips system at the University of Alabama at

Birmingham (UAB).1"' In this study, women were identified who had

angina-like chest pain characteristic of ischemia and CA tested to

measure the degree of stenosis in their macrovessels. The patients

with insignificant stenosis (less than 30% stenosis) were also

evaluated for all other possible causes for their chest pain, without

diagnosis. UAB's study identified micro-vascular dysfunction (MVDS) in

30% of the women with previously unidentified chest pain. This data

was also compared with 17 normal volunteers (ages 21 to 53 years;

average age 32 10; 10 males, 7 females), who underwent 8 minutes of

isometric handgrip exercise at 30% maximum force. There was no

significant change (-0.1%10.3%) in the [PCr]/[ATP] ratio as compared

to rest. These values are consistent with similar handgrip exercise

literature where Weiss et al.4 and Yabe et al.12 have similar estimated

percent change statistics in their normal population samples of

1.9%7.0% and 0.5%9.0%. The determination of significance was set at

two standard deviations (20.4%) of a percent drop in the [PCr]/[ATP]

ratio of the normal volunteers with handgrip exercise." In eight

patients (7 male, 1 female) with CA proven stenosis greater than 70%

blockage, the same amount of exercise resulted in a -24%2% drop in

[PCr]/[ATP] This comparison of normal volunteers versus patients with

proven stenosis proves that the phosphorus NMR spectroscopy test is

viable for differentiating non-ischemic versus highly ischemic hearts.

The Birmingham study also evaluated 17 women with less than 30%

stenosis (very little blockage) and undiagnosed chest pain (suspected

microvascular ischemia). The same amount of exercise produced little

[PCRI/[ATP] change (-2%7%) in 12 of these women, but 5 women had a PCr

change of -27%. This change is attributed to MVDF, because of what is

expected by the ischemic angina-like symptoms and the lack of

macrovascular disease.'1 Accordingly, cardiac P-31 MRS shows

significant potential as a quantitative test for myocardial ischemia

that does not depend on the presence of macrovascular disease. No

other current diagnostic modality is able to quantitatively assess the

degree of ischemia of the heart in this way.

In-Magnet Exercise. For patients with myocardial ischemia, there

is a significant difference between phosphorus cardiac spectra obtained

at rest and during stress (in-magnet exercise).4" It is clearly

valuable to use some type of in-magnet exercise during one period of

the cardiac spectroscopy acquisition protocol. Studies using in-magnet

exercise have been done by exercising the legs, 1',5's7.197 arms with a

hand-grip,49'5,'i098 ~ 9s and drug-induced (dobutamine infusion) cardiac

stress13. The leg exercises can allow for prone position exercises

(prone position is best for cardiac imaging and spectroscopy to reduce

respiratory motion artifacts). One design by Conway et al.1s7 has the

subject lie in a prone position, lifting 5-kg weights with the legs by

bending at the knee. This stress test tends to produce an increase in

the heart rate pressure product of around 70%.25'157 Isometric, hand-grip

exercise, where the subject squeezes continuously at a constant 30% of

the subject's maximal force,"4 can also be performed simply in any

position. In-magnet exercise tests have been also done with devices as

simple as a bottle of water. Widmaier et al."' successfully implemented

a dynamic hand-grip finger flexion exercise consisting of squeezing a

50-mm diameter, water-filled, plastic bottle at a Maximum Voluntary

Contraction (MVC) approximately once every second for a 130 seconds

acquisition. Such an exercise increases the heart rate pressure

product by approximately 30 to 35%, which is still enough to increase

coronary vasoconstriction in the presence of critical levels of

coronary stenosis.19' The two types of handgrip exercise used are either

dynamic, where the grip is released and regrabbed up to the 30% maximum

level repeatedly during the test, or isometric, where the handgrip is

held constant at the 30% of maximum effort level. As can be seen from

Table 8, both dynamic and isometric handgrip raise the heart rate and

blood pressure, but the isometric method is a harder level of work and

therefore responds with a greater heart rate and blood pressure

response. Finally, dobutamine infusion (drug infusion) has the ability

to increase the rate-pressure product by 60 to 130%.7'16

Table 8. Literature Review of In-Magnet Handgrip Exercise Response.

Avg Rest Avg Rest Rate
30% Max & & Pressure Avg Rest &
Handgrip Exercise Exercise Product Exercise Patient
Exercise HR SBP (HRxSBP) [PCr] / [ATP] Type Reference
dynamic 67+8 117 12 7839 1.850.28 control 12
77+11 13113 10087 1.900.23

dynamic 6812 11813 8024 1.600.19 ischemia 12
7513 134+16 10050 0.960.28 >= 75%

dynamic 63+11 11514 7245 1.24+0.30 infarction 12
74+13 128+13 9472 1.190.28 >= 75%

isometric 6712 143 9600 1.720.15 control 49
81+10 156 12600 1.740.17

isometric 77+ 13 132 10200 1.45+0.31 CAD and 49
8916 151 13400 0.910.24 ischemia
>= 70%

isometric 75+13 132 9900 1.59+0.31 Non- 49
85+14 159 13500 1.55+0.24 ischemic

F = Female; M = Male

Post-Processing Calculations and Corrections

Simply obtaining a phosphorus spectrum from a voxel in the

myocardium is not enough to ensure useful spectral data. The cardiac

spectrum is generally contaminated from unwanted signal from

surrounding blood and skeletal muscle. Also, T, relaxation corrections

must be made when TR < 5 x T1 and T2 relaxation corrections when

acquiring an echo instead of an FID. Finally, calculation of absolute

concentrations of metabolites may show changes undetected by metabolite


Starting with the first human study of cardiac phosphorus

spectroscopy, the techniques for gathering phosphorus cardiac spectra

have gradually changed and improved. The first cardiac spectrum was

obtained by Bottomley et al. in 1984,45 corrections for relaxation by

1987 by Bottomley et al.,14 and blood contamination corrections as early

as 1991 by Sakuma et al.55 The pulse sequences and post-processing

methods have been gradually refined and duplicated by different

research groups and yielded very comparable results. In each case the

ratio of myocardial [PCr]/[ATP] was comparable for reference controls

during rest: 1.800.21 (n=12)," 1.930.21 (n=17),150 1.95+0.45

(n=19),82 1.650.26 (n=9).'65 Uncorrected cardiac spectra produce

unreliable results,7 therefore studies of human cardiac spectroscopy are

no longer publishable without corrections for blood contamination and

relaxation effects. In addition, since there is no method for

correcting for skeletal muscle contamination, such contamination

invalidates the study results.

Skeletal Muscle Contamination

Skeletal muscle contains the same phosphorus peaks as cardiac

muscle, but in different quantities. The primary method for ensuring

that there is no skeletal muscle contamination in the cardiac spectrum

is to use good methods of localization with an appropriate pulse

sequence.5" This dissertation project will evaluate ISIS and CSI

derived pulse sequences for elimination of skeletal muscle


Blood Contamination

Blood contains ATP and PDE as does myocardium, but no PCr."

Blood also contains 2,3-DPG (2,3-diphosphoglycerate), which produces a

doublet in the phosphorus spectrum at chemical shift positions of 5.4

and 6.3 ppm, near Pi and phosphomonoester resonances. Blood

contamination causes the [PCr]/[ATP] ratio to appear reduced." Also,

the myocardial Pi peak in normal heart is small and difficult to

resolve from the blood DPG signal.7.148

The correction method for blood contamination is to determine the

relative signal contributions of 2,3-DPG and ATP from blood and use

this knowledge to correct for blood's contribution to ATP. This is

accomplished with the use of a correction factor, which is the ratio of

blood [ATP]/[2,3-DPG]. Such a correction factor for the cardiac muscle

[PCr]/[ATP] ratio typically increases its value by 13% 6% at 1.5

T.,..2.14,is5oI The blood correction is considered small enough that a

substantial error in [ATP]/[DPG] bid ratio will not severely compromise

the final [PCr] / [ATP] ratio, although the final value of

myocardial [PCr] /[ ATP] caac spectrum will still be better than the

uncorrected value.' The ratio of blood ATP to DPG obtained from basic

spectrometer methods is 0.30.200-202

Spectra from pure blood are usually obtained in vitro using

heparinized (anti-coagulant) blood samples. Unfortunately, one study

using proton decoupling of blood samples resulted in spectra with the

contribution from 2,3-DPG overestimated and ATP underestimated.202 In

addition, most of these types of blood studies were conducted days and

weeks after the sample was obtained, since most studies were performed

to test the survival of stored blood. These studies showed increased

Pi which indicate the breakdown of 2,3-DPG. However, ATP was shown to

remain constant when the cells were maintained under appropriate

conditions of 37 C temperature and the appropriate gas mixture..o..20

The excess Pi signal appears as soon as two hours after obtaining the

blood sample.2"' As the 2,3-DPG breaks down over time the ATP/2,3-DPG

ratio increases.202 Consequently, the published ratio of 0.30 may be

elevated due to the delay before acquisition. There are thus several

reasons to distrust the results of these experiments when determining

the proper ratios to correct for blood contamination in the in vivo

cardiac muscle spectrum.

At least two human cardiac NMR research groups have dealt with

the problems of obtaining adequate blood spectra by determining their

own correction factors for blood ratios. They extracted venous blood

(-50 milliliters) from their volunteers and obtained phosphorus spectra

of the blood in the same magnet used for the cardiac study,2.'14s in one

case using the same NMR acquisition techniques as well.148 In both cases

studies were conducted at 1.5 T and the blood [ATP]/[2,3-DPG]

concentration was much lower than the reported values from the standard

spectrometer experiments: 0.110.0282 and 0.14 0.02.18 These research

groups were also able to correct for the blood 2,3-DPG contamination of

the muscle PDE peak with correction ratios of [PDE]/[DPG] of 0.19+/-

0.0382 and 0.21+/-0.02.'" Theoretically, the blood could be measured

in-vivo directly from inside the heart, but has not been documented in

the literature to date.

Ischemia causes a reduction in systolic wall thickness. It has

been debated that because of the reduced volume of myocardium, the

phosphorus spectrum will be more contaminated with blood for ischemic

patients. As blood contains ATP, blood contamination can alter the

[PCr]/[ATP] ratio. Fortunately, the amount of blood contamination can

be corrected for, and it has been shown that the [PCr]/[ATP] ratio is

still reduced with exercise for ischemic patients even after blood


Relaxation Corrections

Two components of the macroscopic magnetization in NMR are

subject to time dependent exponential relaxation effect. The

longitudinal magnetization or spin-lattice relaxation along the z-axis

is an increasing exponential function (Ti dependent) with a maximum of

magnetization of Mo. The transverse magnetization or spin-spin

relaxation in the x,y plane is a decreasing exponential function (T,

dependent) with a minimum magnetization of near zero.208

The relaxation time T0, is the time required for the net

magnetization (M) to return to 63% of its original value following an

excitation pulse.38 Ti relaxation rates depend on the presence of

molecular interactions in the vicinity of the excited spin that

modulates with an intrinsic frequency (wJ) ." When w, is near in

frequency to the resonance frequency (w) the interaction will more

readily absorb the resonant energy and this energy transfer will occur

more frequently. This allows the collection of spins to return to the

equilibrium configuration sooner, resulting in a shorter T, value.38 In-

vivo metabolites as studied with P-31 spectroscopy are usually small

molecules where the rate of molecular motion is rapid. This results in

a poor match between w, and w0 and thus relatively long Tl relaxation In addition, because of the relationship with resonant

frequency, Ti is also dependent on the main magnetic field strength.

In a typical P-31 MR spectroscopy experiment, the time between

successive radio frequency pulses (TR) is usually insufficient for

complete Ti relaxation. Successive acquisitions applied at a short TR

result in a steady state of M where the spins are partially saturated

and the resulting MR signals are reduced from their completely relaxed

values." This situation makes quantitation difficult since the correct

MR signal from each P-31 metabolite is directly proportional to the

number of spins only when the collection of spins is at equilibrium.

Different metabolites relax at different rates. For example, the

[PCr]/[ATP] ratio will typically be too small due to the faster

relaxation of ATP. Fortunately, there is a way to correct this


T, relaxation corrections are necessary for cardiac spectroscopy

studies since the pulse repetition time, TR, is generally much shorter

(minimum 1 sec when gated) than the total time for complete relaxation,

five times T,"1 (Ti = -4 sec for PCr and ~2 sec for P-ATP at 1.5 T).s5

The method for determining the relaxation factor for cardiac

spectroscopy is to obtain two sets of phosphorus spectra, one at short

TR (TR < T,) and one at a fully relaxed TR (TR >> Ti) Then a simple

division of the ratio of [PCr]/[ATP] measured at the fully relaxed TR,

over the [PCr]/[ATP] ratio measured at the shorter TR, provides the

relaxation factor, assuming the same TR for each relaxation corrected

experiment.209 This factor can in turn be multiplied by the [PCr]/[ATP]

ratio of the localized spectra obtained with the shorter TR value to

obtain the relaxation corrected results.

The relaxation correction factor (RCF) ideally should be obtained

directly from the myocardium, with localized techniques. This requires

a set of volunteers be used to gather data to estimate the relaxation

correction factor for all studies at that field strength and frequency.

This is because it would take an extra 30 minutes to an hour to gather

data for RCF for each subject, an unreasonable request to be tacked on

after a current two-hour study. However, Bottomley et al.6,209.210

obtains the relaxation data from each subject by using unlocalized

acquisitions at long and short TR times, adding only six minutes of

scan time2' and allowing the calculation of RCF for each participant.

This method also assures that the flip angle, pulse power, RF coils,

and patient is the same for the localized spectrum and for the

measurement of the correction factor.20 In doing so, a large assumption

is made that the P-31 metabolites of skeletal and cardiac muscle have

the same relaxation rates and thus the same RCFs. This assumption is

based on animal studies in rat skeletal muscle211 and canine cardiac

muscle. -

Those research groups that have measured animal and human

localized cardiac P-31 metabolite Ti values directly have shown that it

is possible to obtain relatively consistent values although there are

still discrepancies between research sites. This is demonstrated in

Table 9 where the standard deviations of some published values of Ti are

relatively low with greater discrepancies between reports, such as

between Neubauer et al." and most of the other human studies164'193212 at

1.5 T. There is no need to rely on assumptions about cardiac and

skeletal muscle metabolites, when the Ti relaxation values of the P-31

metabolites can be measured from the cardiac muscle directly. A

correction factor of 1.28, has been measured at UAB'95 and is used in the

WISE study to correct for relaxation effects at 1.5 T. To use this

correction factor, simply multiply it by the blood corrected

[PCr]/[ATP] ratio.

Table 9. Published Spin-Lattice Relaxation Times of Myocardial PCr and

Field TI (PCr) Ti (y-ATP) TI (P-ATP)
Subject (Tesla) (sec) (sec) (sec) Reference
Dog 1.9 4.40.1 1.8+0.2 1.60.1 118

Pig 4.7 4.80.9 3.0+1.7 2.61.7 213

Pig 2.0 6.30.4 2.2+0.8 2.2+0.7 213

Human 4.0 5.31.6 2.7.6 -17

Human 1.5 4.2 1.7 193

Human 1.5 4.0 1.80.2 212

Human 1.5 4.1 2.70.8 164

Human 1.5 6.1 5.4+0.5 5.81.0 14

The relaxation time T., is the time required for the transverse

component of M to decay 37% of its initial value. At equilibrium, Mo

is oriented only along the z (Bo = main magnetic field) axis and no

portion of Mo is in the x,y plane. The coherence or uniformity of the

spins is entirely longitudinal with no transverse component.38 A 90

radio frequency pulse causes Mo to rotate entirely into the xy plane,

so that the coherence is in the transverse plane at the end of the

pulse. After the pulse, the coherence gradually disappears, the spins

lose phase coherence, and reorient themselves along Bo. The

disappearing coherence produces the free induction decay (FID) with a

dephasing time of T2 or T2*, where T2 is always less than T, and T2* is

less than T2.3'

After the application of the 90 radio frequency pulse, when M is

oriented in the transverse plane, each spin processes at the same

frequency wo, and the spins are in phase. Each nearby spin of the same

type and the same molecular environment will have the same w,. The wo

will not remain the same, however, as intra- and inter-molecular

interactions will cause the local magnetic field to modulate around

each spin causing wo to vary. The variations will produce a gradual,

irreversible loss of phase coherence and a reduction in the transverse

magnetization." In addition, non-uniformity in the Bo field and

magnetic susceptibility differences can cause additional loss in

transverse phase coherence and T,* relaxation." Fortunately, neither T2

nor T2* are factors that need to be corrected for quantitation of a P-31

spectrum using ISIS or DRESS, but they can affect the quality of the

spectrum. The Ti component of relaxation determines the amplitude of

the metabolite signal while the T, or T2' has an effect on the decay of

signal with time and the linewidth. Shimming of the region of interest

can attempt to correct for some losses in phase coherence, but P-31

metabolites have inherently fast and unalterable T2 times.

Calculation of pH

Some values of cardiac pH as found in the literature are shown in

Table 10. The pH is proportional to the frequency difference of the Pi

and PCr P-31 metabolite peaks. In the literature it is well stated

that it is not always possible to see the Pi peak in every case due to

SNR differences between cases. None of the publications listed were

able to obtain pH values on every single subject. There are also no

known publications of human cardiac pH during in-magnet stress.


Table 10. Myocardial pH in the Literature as Measured by Human, In-vivo
Phosphorus NMR Spectroscopy.

Type of patient n pH at rest Reference
Normal Control 4 7.150.03 165

Normal Control ? 7.15 0.02 149

Normal Control 1 7.17 159

Infarction 4 7.15 0.06 149

ox-- A A------

,.Q___ A r



Figure 1. STEAM Pulse Sequence


RF 1800
RF -t A -
.~ rv




TE1/2 TE1/2 TE2/2 >< TE2/2
Figure 2. PRESS Pulse Sequence


1t 2nd 3rd 4th

5th 6tt 7"h 8th

Volume: Slice: Column: Voxel:
(1") l) (1 2d ) ( 2nd+3' _-4th) (L'-2"d+3 Cd -4 t h
+5th-6th+7th-8 h)

Figure 3. ISIS Volumes for 8 Acquisition Voxel Localization.


This chapter covers the design and implementation of the phantom

techniques for evaluating the 3.0 Tesla GE Signa Advantage' magnet in

Gainesville, Florida for purposes of performing cardiac spectroscopy on

human subjects. The 3.0 Tesla magnet located in the tunnel between

Shands at the University of Florida (UF) and the Veterans Affairs

Medical Center (VAMC) is owned by the VAMC, Shands at UP, and the Brain

Institute at UF. Phantom and coil combinations were tested with a

variety of imaging and spectroscopy pulse sequences. The purpose of

utilizing a phantom is to take the opportunity to test all the possible

pulse sequences and options with a standard sample (the phantom) to

compare each method. In addition having a phantom prevents the need to

have an endless supply of human volunteers to test out each pulse

sequence and option. In addition, it is often difficult for humans to

lie perfectly still in the magnet for a long period of time, as they

are generally required to do for all MRI and MRS studies (with the

exception of exercise studies, where special procedures are

implemented). When new protocols are initially being tested, they tend

to take longer to run than after they are better defined, again

pointing to the advantage of using a constant, stationary phantom.

Although there is not enough room in each figure caption to capture the

full protocol under which each image and spectrum were obtained, this

information can be found in the appendix, identified with its reference


Phantom Design

Four phosphorus phantoms have been identified or built for the

purpose of initially testing the parameters for each pulse sequence and

protocol, before use on human subjects.

Gate-able and Depth Phantoms

Two phantoms, the gate-able and depth phantoms consist of two

compartments, each containing a different phosphorus compound, which

present as separate chemical peaks in the phosphorus spectrum.

Photographs and images of the gate-able and depth phantoms, in each of

the three planes can be found in Figure 4, Figure 5, Figure 6, and

Figure 7. The outer and inner compartments (OC and IC) are filled with

the phosphorus containing chemicals of sodium dihydrogen phosphate

(NaH2PO ) and methylenediphosphonic acid (MDPA), respectively, each

diluted in distilled water. The compartment sizes and concentrations

for each phantom are described in Table 11. Two sections are necessary

to test the ability of each pulse sequence to localize to the desired

voxel (represented by the IC) and exclude the outer voxel signal

(represented by the OC) .1 The OC represents the unwanted signal from

skeletal muscle and blood that surrounds the myocardium. The IC

represents the phosphorus signal from the myocardium. In addition, the

depth phantom's IC is surrounded by a layer of water (compartment 2 in

Table 11) which is just thick enough to eliminate most of the

contamination from just outside the selected volume. For this reason

the depth phantom is especially useful in demonstrating complete

elimination of the OC signal of NaH2PO, when the pulse sequence is able

to do so. Both phantoms were designed to load the H-1 and P-31 coils

similar to a human chest load. The estimation of loading equivalency

is demonstrated by the resulting quality factors (Q-factor) values of

the loaded H-1 and P-31 coils compared to a human chest load, as shown


Table 12. The Q-factor will be explained further in the section

of this chapter on coil designs. The center of the gate-able phantom's

IC is at a depth of 5 cm, to approximate the depth of the anterior wall

of the heart, while the depth phantom has variable depth IC.

Table 11. Gate-able and Depth Phantom Compartment Sizes and

Inner Inner Outer Outer
Compartment Compartment Compartment Compartment
Phantoms Size Concentration Size Concentration
Gate- Container: Container:
able 4.7-cm length, 80 mM MDPA 11-cm max height, 80 mM NaHPO,
2.7-cm ID*; 28-cm ID*;
35 ml partially filled
with 4 liters

Depth Container 1: Container:
3-cm length, -80 mM MDPA 16-cm max height, 30 mM NaHPO,
2.3 cm ID* 30-cm ID*; plus 70 mM
partially filled NaCl
Container 2: with 7 liters
3-cm length, water
3.6 cm ID* barrier
Abbreviations: ID* = inside diameter

The gate-able phantom was initially designed to test the degree

of localization of each pulse sequence and to do this while the IC was

moved in and out of the volume of acquisition in a gated fashion. A

volunteer standing outside of the magnet would gate the magnet

acquisition with a peripheral-gating device reading the pulse rate in

the volunteer's finger. While viewing the pulse waveform, the

volunteer would use an extension rod attached to the IC of the gate-

able phantom to coincide the movement of the IC with the peripheral

pulse gating, pushing and pulling the IC in or out of the acquisition

volume (AV) with each beat. The system gated the sequence to the peak

of the peripheral pulse waveform. It was found that placement of the

IC in the AV at the peak of the pulse waveform resulted in a signal

that was equivalent to that achieved by placing the IC in the AV

without movement. In addition, when the cycle was reversed such that

the IC was outside of the AV at the peak of the peripheral pulse, and

moved inside between gated pulses, the result was equivalent to having

the IC outside of the AV for the entire sequence without movement.

Therefore, the gating was found to be predictable and accurate.

The depth-changing phantom (or simply "depth phantom") was

borrowed from Hee-won Kim's dissertation work. This phantom is

designed with an inner compartment that can be moved to different

depths away from the coil surface, thus simulating different cardiac

depths. Unlike the gate-able phantom, the IC position is fixed before

the start of acquisition and cannot be moved during acquisition,

although it can be moved to different depths between acquisitions. The

IC of the depth phantom consists of two separate containers, an inner

container of MDPA surrounded by a second container providing a layer of

water between the IC and OC. The water layer is thin (-0.7 mm) but

large enough to help eliminate a majority of signal just outside the

selected volume of interest. This phantom is therefore ideal to use

when demonstrating the ideal localization ability of a pulse sequence

and to show virtually no external contamination from the external

NaHPO4. The primary use of the depth phantom, however, is to estimate

the optimal transmitter gain (TG) which determines the optimal flip

angle at each depth. For the human studies, these phantom measurements

will provide initial guesses of TG values for human cardiac P-31 MRS


Slice Profile Phantom

A third phantom, the slice profile phantom, as shown in Figure 8,

consists of a 2 ml vial of 14.7 M phosphoric acid (HPO4), placed at an

angle of 40 at depth of 5.5 cm with a plastic arch that is glued in

place. The vial is surrounded by an outer compartment (22 x 22 cm2 base

x 10 cm tall) of 4 to 5 liters of 70 mM sodium chloride to load the

coil. This phantom was used to more precisely detect the degree of

contamination in a slice from a source (the 2 ml vial of HPPO4) just

outside the slice.

GE Phosphoric Acid Phantom

A GE plastic bottle phantom consisting of a larger volume (-450

ml) of phosphoric acid (H3PO4) provided a strong P-31 signal for use in

comparing the different types of P-31 coils. This phantom provided

enough P-31 signal for adequate imaging at the P-31 frequency. It was

used for quick working checks of coils and equipment, as well as to

verify dimensionally the depth ranges of each phosphorus coil used in

this work. The bottle is 18 cm tall (including cap) and has a 7.3 cm

diameter at its widest. GE's 14.7 M Phosphorus Phantom is labeled as

"GE Medical Systems, 46-317299G2, Spectroscopy Service Phantom, 14.7 M

H3PO4 Phosphoric Acid" as shown in Figure 9.

Radio Frequency Coil Design

Sets of proton and phosphorus radio frequency coils were selected

in order to obtain the best performance from a study. The proton coil

is used for imaging, swimming and ensuring correct positioning of the

phosphorus coil. To prevent changes in shim values and localization

due to patient movement, it is essential to have the proton coil and

phosphorus coils in place during the entire study, independent of

whether the magnet is set at the proton or phosphorus frequency. In

terms of RF coils, this means having proton and phosphorus coil(s) that

are compatible and maximize the SNR for the phosphorus spectrum. The

coils were designed dimensionally by the author, with the author

sometimes providing the platform and placing the copper tape on the

coil former with proper dimensions. All of the capacitor, resistor and

cable placements for proper coil tuning and usage, in addition to some

initial platform creations, were performed by Dave Peterson and Bryan

Wolverton in Dr. Fitzsimmons' coil lab, at the Veterans Affairs Medical

Center, Gainesville, FL. In addition, Dave Peterson provided

assistance in measuring coil parameters as are shown in Table 12 and

Table 13.

3.0 Tesla Square Proton Coil Paired with Quadrature Phosphorus Coil

The GE 3.0 Tesla magnet does not have a body coil, therefore it

is necessary to build a proton surface coil in addition to the

phosphorus surface coil. A 10 x 16 cm2 quadrature transceive phosphorus

coil (two 10-cm diameter coils overlapping), coplanar with a 25-cm

square transceive proton coil was created and was shown to provide

adequate imaging and spectroscopy performance. The quadrature coil

provides an advantage over the basic single-turn coil. Theoretically,

a quadrature coil will provide up to 2 times the signal to noise of a

single turn coil."' In practice, the signal to noise improvement can be

better or worse depending on the quality of the coils that are

compared. A quadrature coil, acting as receive only, provided 1.4

times the signal-to-noise ratio of a single turn coil having the same

shape and total dimension, in a cardiac imaging acquisition.'18

Originally, the quadrature phosphorus coil and the square proton coil

were permanently positioned together on one holder as shown in Figure

10a. A schematic diagram of the proton square coil is shown in Figure

lla, whereas a schematic diagram of the phosphorus quadrature coil is

shown in Figure llb. It was found that this pairing, when positions

were permanently placed, offered limited flexibility in centering the

quadrature coil over the heart, where the coil holder was often

protruding into the chin of the volunteer. A modification was made to

mount the two RF coils on separate holders so each could be optimally

and comfortably positioned. This modification is shown in Figure 10b.

3.0 Tesla Phosphorus Single-Turn Coil

Most cardiac spectroscopy work is currently done with a simple,

single turn surface coil. It is for this reason that tests were

performed with a single-turn phosphorus coil for comparison with the

quadrature coil results. The single turn coil is a 10 cm diameter

transceive design tuned to 51.71 MHz for phosphorus at 3.0 Tesla. A

photograph of this coil is shown in Figure 12, and a schematic diagram

of this coil is found in Figure 13.

3.0 Tesla Coil Comparisons

Phantom tests with the GE 14.7 M phantom show that the quadrature

coil overall performs significantly better than the single turn

phosphorus coil, especially at greater depth.

Table 12 lists values of the basic coil parameters of quality (Q-

factor = quality factor) and isolation for each of the 3 Tesla tuned

coils. The quality factor is defined as the center frequency over

bandwidth, where the bandwidth is the frequency +3dB and -3dB from the

center frequency."' The parameters were measured with help from David

Peterson in Dr. Fitzsimmons' coil lab, using the HP 8752A Network


Table 12. A Comparison of 3.0 Tesla Coil Parameters.

Square Quadrature Single-Turn
H-1 P-31 P-31
Q-factor Unloaded: 16 44 30
Q-factor Human* Chest Load: 2.2 24 19
Q-factor Depth Phantom Load: 2.8 12 13
Q-factor Gate-able Phantom Load: 2.5 16 16
Isolation of Quadrature: 15.4 dB
Isolation of P-31 coplanar with H-: 36 dB 42 dB
human chest load of 29 year old male with body mass index of 25.5.

In addition to measurements of the quality factor and isolation,

each phosphorus coil's performance was mapped. Comparative mapping of

the quadrature versus single-turn coil performance was achieved by a

series of P-31 spectroscopy slice acquisitions where the TG was

optimized at each slice. Using a larger proton slab phantom beneath

the coil for loading, and placing the 14.7 M phantom on top of the

coil, which provided excellent SNR, a DRESS slice (25 mm thick) was

obtained at 5 mm intervals from each coil, as shown in Figure 14. At

each slice, the TG was optimized. The maximum distance from each coil

measured was at 96 mm, just under 10 cm, where the 14.7 M P-31 phantom

bottle's neck formed. The bottle also had curved edges at the bottom

and top as well as a slight concave area at the bottom, which can

explain part of the reduction in signal in the first few slices from

the RF coils. Based on the results, the single-turn RF coil has

optimal SNR at a depth of 25 mm while the quadrature RF coil has an

optimal SNR at a depth of 42 mm. This data reinforces the fact that

the quadrature coil penetrates deeper with greater SNR and is therefore

more ideal for reaching the depths of hearts of thicker chested

individuals. This result is duplicated when the ratio of quadrature to

single-turn signal is plotted, showing the quadrature coil peaks with

the greater signal at both the coil surface and at a depth of 42 mm.

In addition, until the depth of 60 mm, the quadrature coil outperforms

the single-turn coil in terms of relative signal. Increase in signal

from the depth of zero to 3 cm in Figure 14 can also be explained.

Starting at the coil position (zero distance from the coil), only half

of the first slice contains the phantom material. As the depth

increases, the entire phantom is in the slice by a depth of 1.25 cm.

Acquisition of a spectroscopy slice is not 100% accurate at the edges,

where a Gaussian like function describes the contamination of signal

from outside the slice. At the first slice depth where the entire

phantom is in the selected slice, only one edge is contributing signal

contamination. As the depth increases, eventually both edges will

contribute a small bit of contamination. In addition, although the

14.7 M P-31 phantom is only 73 mm in diameter, this test is not 100%

accurate because the profile of each coil (single turn and quadrature)

may not have equally obtained signal from each slice.

1.5 Tesla Coils

One main argument for using the 3.0 Tesla magnet is its superior

performance when compared to 1.5 Tesla. All comparisons between the

1.5 Tesla and 3.0 T were done on GE scanners with identical software

(version 5.4) and with components as similar as possible. The 1.5

Tesla GE Signa Advantage" has a body coil, therefore no proton coils

were created. A single turn phosphorus coil of 10 cm diameter was

created. In addition, a quadrature coil of the same dimensions as the

3.0 Tesla P-31 quadrature coil was constructed. The same parameter

measurements taken from the 3.0 Tesla coils, were also performed on the

1.5 Tesla coils as shown in Table 13, with the exception of the 1.5

Tesla body coil, which cannot be moved for such testing in the coil

lab. In addition, a photograph of each of the coils can be found in

Figure 15 and Figure 17, and a schematic for each can be found in

Figure 16 and Figure 18.

Table 13. Comparison of 1.5 Tesla Coil Parameters.

Quadrature Single-Turn
P-31 P-31
Q-factor Unloaded: 77 48
Q-factor Human* Chest Load: 45 33
Isolation: -18.6 dB not applicable
Match: 68 pF per side not applicable
reflections: -18 dB / -29 dB not applicable
* human chest load of 29 year old male with body mass index of 25.5.

Coil Ideas for Future Cardiac Spectroscopy Studies

Future coil sets may consist of a smaller proton coil or a set of

phosphorus coils with a separate, larger transmit coil, thus reducing

spatially dependent spectral distortions from the excitation field.2n1

There is also the possibility of other coil designs, if one is found to

significantly improve the performance of the cardiac imaging and

spectroscopy acquisition.


Both Spin Echo and Gradient Echo imaging pulse sequences are the

most common MRI techniques for proton imaging on the GE system. The

goal of the phantom imaging was to predict which pulse sequence would

provide the best image uniformity and depth of penetration, when used

with a simple surface coil. In this chapter, imaging is briefly

examined at 3.0 T and only in phantoms, but in the next chapter images

taken at 1.5 Tesla will be compared with 3.0 T and explained in greater

depth. Imaging with the spin echo pulse sequence on the 3.0 Tesla with

a surface coil has its limitations. Even with a maximum transmitter

gain (TG) the signal depth is not as good as the gradient echo images.

This is shown in Figure 19 where the top corners of the phantom are cut

off when spin echo imaging is used, but are still visible when gradient

echo imaging is used. In contrast, the spin echo pulse sequence works

better at producing images than the gradient echo sequence on the 1.5

T, where a body coil provides more uniform excitation of the axial

slice and requires less overall power. Human cardiac images shown in

the next chapter emphasize this point. Also note the slight asymmetry

in signal brightness in the gradient echo images, which is due to the

slightly off center placement of the 25 cm proton coil with the center

of the gate-able phantom. Gradient echo imaging, especially with a

flip angle of less than 90 (600 was used) was found to work best for

imaging an axial slice with a simple surface coil.

Phosphorus imaging can only been done with a phosphorus sample of

extremely high concentration, such as with the GE 14.7 M phosphorus

phantom. The phosphorus concentrations in the human body, or in the

depth or gate-able phantoms, are not high enough to produce an image

within the standard phase and frequency steps designed for proton

imaging. The result of imaging a phosphorus sample with normal tissue

amounts of phosphorus in the millimolar range was simply an image of

noise. The only phosphorus imaging that is presented in this write-up

is of the GE 14.7 M phosphorus phantom. The GE 14.7 M phantom images

were used to compare the depth penetration abilities of each of the

phosphorus coils.


GE offers a number of spectroscopy pulse sequences of different

characteristics and qualities. Due to characteristic differences of

metabolites at different frequency ranges, there are pulse sequences

that are more appropriate for either H-l or P-31. In addition, some

pulse sequences are designed to acquire spectra unlocalized, while

others can slice, column or voxel localize. There is no direct purpose

for using unlocalized H-1 or P-31 spectroscopy acquisitions, assuming T,

corrections will be based on spectra localized to the cardiac muscle.

The discussion will be limited to localized spectroscopy sequences that

would be of use in a cardiac MRS protocol. Using phantom studies the

GE pulse sequences will be described and evaluated for the later

purposes of H-1 localized acquisition for swimming and P-31 localized

acquisition for evaluation of cardiac disease. Quality of the

phosphorus spectra from the phantom studies will be determined based on

acquisition time, SNR, and degree of localization (i.e. increased

signal from the VOI and reduction of signal outside the VOI).

Note that most figures of complex P-31 spectral data in the

figures of this dissertation are graphed as real, phased data, as noted

by "REAL" as labeled by GE's SAGE_IDL"' spectral post-processing


Localized Proton Spectroscopy

Although the magnet's homogeneity is optimized upon installation,

any individual person or object that goes in the magnet will distort

the main magnetic field to some degree. SNR and the resolution of

peaks with frequencies that are close together depend partially on good

field homogeneity. Homogeneity can be increased in a region of

interest with a technique called swimming. Shimming involves obtaining

an unsuppressed H-1 MRS signal from the region of interest and

adjusting the gradients to optimize the H-1 MRS signal. The H-1 MRS

signal is used because the water peak provides high SNR and optimizing

the magnetic field for the water protons will also improve the signal

for the P-31 metabolites. Since the linewidth of a resonance peak is

inversely proportional to T,*, the homogeneity can be optimized by

either minimizing the full width at half maximum (FWHM) of the water

peak or maximizing the extension of the T:* dependent free induction

decay (FID) of the H-1 MRS signal.

The two spectroscopy pulse sequences on the GE system most

appropriate for voxel localized H-1 spectroscopy are STEAMCSI and

PRESSCSI, based on the standard STEAM and PRESS techniques in the

literature. GE's STEAMCSI pulse sequence incorporates three slice-

selective 90 RF pulses and a set of crusher gradients, as shown in

Figure 20. GE's PRESSCSI voxel localized pulse sequence is obtained

through three slice-selective RF pulses and utilizes a spin-echo with

900, 1800 and 1800 pulses and two sets of crusher gradients, as shown in

Figure 21. The PRESSCSI pulse sequence was preferred because of its

general ability to provide twice the SNR of the STEAM sequence, as

explained previously in the literature review. The FWHM values of the

water peak from the phantom studies with PRESSCSI and STEAMCSI were

5.78 Hz (0.045 ppm) and 14.09 Hz (0.110 ppm), respectively.

Localized Phosphorus Spectroscopy

Using the best possible scenario for obtaining a phosphorus

spectrum, namely using a phantom, it is easy to definitively compare

the quality of the results of each of the available GE pulse sequences,

and some modified GE pulse sequences. The use of a phantom is ideal

because it allows for a standardized cross-comparison of pulse sequence

acquisition results with a non-moving, non-changing phantom. Each of

the available GE pulse sequences, PRESSCSI, STEAMCSI, ECHOCSI,

SPINECHO, ISISCSI, and FIDCSI are compared in increasing order of

quality of localized results within a reasonable amount of time

(usually less than ten minutes per acquisition). Such an increase in

quality of MDPA localization can be seen through the progression of

quality of all GE spectroscopy pulse sequences via progressively

smaller localization volumes, as seen in Figure 22 through Figure 32,

and Figure 34 through Figure 38. Also presented are some modifications

of the most viable phosphorus localized pulse sequences, ISISCSI and

FIDCSI, which produce even better results.


GE's STEAMCSI and PRESSCSI pulse sequences, as explained

previously, allow for the acquisition of a single voxel with each TR

acquisition. In addition, both of GE sequences can be used along with

CSI phase encoding gradients to further divide the field of view into

multivoxels. As expected from the literature on PRESS and STEAM,

PRESSCSI doesn't work as well as STEAMCSI, although neither was ideal

for obtaining localized phosphorus spectra, as shown in Figure 22 and

Figure 23, respectively. Note that the STEAM sequence was acquired in

a fourth of the time of the PRESS sequence, but is better able to

display the P-31 peaks.

Phosphorus ECHOCSI

ECHOCSI is GE's modification of the standard spin echo 90-1800

pulse sequence combined with 2-D CSI, as shown in Figure 24.9" GE's

version of ECHOCSI for localized spectroscopy employs a slice selection

and CSI option. The use of a surface coil with this pulse sequence

combines an inhomogeneous B, field with a pulse sequence requiring

somewhat accurate 90 and 180 pulses. In addition, the reliance on CSI

for voxel localization leads to intervoxel signal bleed. The result of

a P-31 ECHOCSI CSI voxel localized acquisition is shown in Figure 25

and demonstrates low SNR and inadequate localization to the MDPA inner

compartment of the phantom.

In addition, reliance on CSI is done at an additional cost of

increased scan time. A specific example of acquisition times, for

various field of views and voxel sizes is shown in Figure 26 for a

sample data set of 128 acquisitions and TR of 2 seconds with the

ECHOCSI pulse sequence. These parameters were chosen based on the same

parameters being used to obtain a localized P-31 spectrum successfully

with GE's FIDCSI pulse sequence, as described later in this chapter.

Notice when the voxel sizes are decreased to reasonable values for

localizing to the cardiac muscle (2 x 2 x 2 cm3 or less) the scan time

increases substantially from 18 to over 60 minutes, depending on the

field of view (FOV). The FOV is centered with the magnet bore so the

region of interest must be within the FOV to obtained spectral data.

Considering this protocol will be used for cardiac spectroscopy

acquisitions, the heart is not centered but to the left of the center 5

to 10 cm (depending on the person). In addition, this figure does not

take into account the increased number of acquisitions that would be

needed to keep the SNR constant as the voxel size decreases, thus

further increasing the scan time.

Phosphorus SPINECHO

GE has available a second spectroscopy pulse sequence based on

the spin echo idea. Like, ECHOCSI, GE's SPINECHO is based on the 90-

1800 spin-echo pulse sequence, except that the pulse and the parameters

associated with the sequence have been optimized with the RF pulse

reformed, as shown in Figure 27. The oddly shaped amplitude (Rhol) and

phase (Theta) modulated RF pulses are the result of a back calculation

of the RF pulse based on an input of the echo time and slice profile"'.

The pulse is a composite of the initial 900 pulse and the refocusing

1800 pulse. These pulses are combined together because if these two

pulses are optimized separately, the resulting echo time is limited by

the length of each individual pulse. The SPINECHO pulse available on

the GE system has been calculated for a specific set of parameters and

is not available for the user to change. The excitation pulse has an

effective flip angle of 60 and an echo time of 2.5 msec. The radio-

frequency pulse is optimized for the acquisition of phosphorus spectra

from 3.0 x 3.0 x 3.0 cm3 CSI volumes. The reported benefit of the

optimization was the elimination of the need for baseline correction

during post-processing. A test of this sequence on 3.0 T is shown for

a CSI experiment in Figure 28. The limitations of a 3 x 3 x 3 cm' voxel

areas positioned with CSI would make it difficult to localize to the

anterior myocardium, where a rectangle would be more appropriate. It

is also not clear that this pulse sequence was optimized for use at 3.0


Phosphorus ISISCSI

The ISISCSI pulse sequence, GE's version of ISIS, is the most

appropriate voxel localization pulse sequence for phosphorus as

provided by GE for the 3.0 Tesla scanner, as shown in Figure 29. On

the GE system, ISISCSI uses an adiabatic RF pulse that allows for

uniform excitation of signals over a larger volume of the sample."

The ISISCSI sequence allows for various options in acquisition

areas such as volume, slice, column, and voxel acquisition. As the

acquisition volume size decreases, the ISISCSI sequence is successful

in eliminating more and more of the outer phantom volume (NaHPO,).

This causes the NaHPO, signal to decrease with increasingly smaller

localized volumes, such as the slice and column and shown in Figure 30

and Figure 31. One step smaller than the column is the voxel. The

voxel acquisition is the most desirable localization technique for

cardiac spectroscopy, because of the need to avoid contaminating the

cardiac muscle signal with signals from skeletal muscle and blood.

ISISCSI, however, is plagued by poor localization due to short TR

times between acquisitions of the next of eight separate volumes, not

allowing for complete relaxation to occur between volume acquisitions.

If the localization was ideal, the spectrum shown in Figure 32 would

show just the peak on the left, MDPA, with no added signal from the

outer volume signal of NaHPO4.

A modified ISISCI sequence can be developed to overcome the

relaxation error problems of the original GE pulse sequence. The

original GE pulse sequence gathers each of the eight volumes (for a

voxel acquisition) that will be added and subtracted from each other in

sequence, with the same TR between each. The acquisition proceeds in

the manner that all eight parts are acquired, and then the process is

repeated until the number of acquisitions (a number divisible by eight)

has been acquired. Without rewriting the ISISCSI pulse sequence, a

modification that improves the ISISCSI localization results has been

made. By acquiring an average of each of the eight volumes separately,

with a delay time of at least 15 seconds between acquisition of each

volume, a much improved voxel acquisition can be acquired with very

little compromise in overall acquisition time. This idea is explained

visually in Figure 33 for an acquisition example of 32 NEX. When all

eight volumes were acquired separately with a short TR time during each

acquisition, but a longer TR between acquisitions, the localization

results were significantly improved, as shown in Figure 34.

Phosphorus FIDCSI

GE's FIDCSI spectroscopy pulse sequence is the most basic of the

spectroscopy sequences involving a single RF pulse, as shown in Figure

35." In addition, CSI is an option to gather multivoxel acquisitions.

The FIDCSI pulse sequence provides a good signal and slice

localization. Slice localization without phase encoding can be done at

a minimum thickness of 25 mm (also called depth resolved surface coil

spectroscopy or DRESS). FIDCSI can do multi-voxel acquisition via

phase encoded CSI. Slice localized spectroscopy that is acquired in

the manner of FIDCSI is often referred to by the general term of DRESS

(depth resolved surface coil spectroscopy), regardless of the machine

or pulse sequence designer. This technique is often used for cardiac

spectroscopy because it is a simple sequence with one spectrum output

per acquisition. As long as the slice is sufficiently deep, the

surface coil sufficiently narrow, and the slice properly positioned,

there is only a small risk of skeletal muscle contamination. In

addition, the transmitter gain (TG) must be optimized to ensure maximum

signal from the slice. The proper TG values have been pre-measured

using the depth phantom, as shown in Figure 36 at 1.5 Tesla and Figure

37 at 3.0 T. These measurement can then be used as a first estimate in

optimizing the signal from human FIDCSI slice localized studies at the

same distance from the coil.

The DRESS slice localized spectrum of a phantom as obtained using

the standard GE FIDCSI pulse sequence is shown in Figure 38.

Unfortunately, the pulse sequence suffers from a significant delay time

before the free induction decay is recorded. Some delay, 1 to 4 msec,

is essential to prevent eddy current signals from contaminating the

signal of interest. Long delays, as 20 msec for human FIDCSI oblique

slice acquisitions at 3.0 T, are less desirable because they cut off

too much signal from the ATP part of the phosphorus spectrum, which has

the shortest T2 relaxation time. Fortunately, there are post-processing

programs, such as FITMASTER" (Philips) which can estimate the missing

part of the FID and provide excellent results.

All parameter options were evaluated for oblique and coronal

FIDCSI slice selections. The oblique slice uses more than one gradient

set to specify the slice and therefore can have more conservative

parameter limits. Available parameter options with the FIDCSI slice

selection protocol include changing a parameter called SQUEEZE, which

reduces or increases the overall time taken for the RF pulse. If the

time is reduced, as it is for SQUEEZE = 2, the delay time until the

start of the FID is also reduced. A delay in the FID creates a

increase in the frequency dependent phase shift (first order phase

correction). A 1800 phase shift will be created for each dwell

period."2 The delay time issue is even greater with the oblique slice

as the system must be "tricked" by changing the variable pwgph to 4

msec to even do an oblique slice localization. This is due to oblique

slice using multiple gradients and having the GE software at the most

conservative level for error messages. The pw_gph variable actually

increases the delay time to 20 msec when used at 3.0 Tesla with a

spectral width of 4000, and with the SQUEEZE parameter set to 1.

After careful evaluation of the FIDCSI pulse sequence and after

speaking with both Napapon Sailasuta and Ralph Hurd, both of GE, a

rewriting of the FIDCSI pulse sequence was necessary to fix this delay

problem. The FIDCSI pulse sequence program file was modified (by Dr.

Hee-Won Kim using GE's EPIC) to decrease the delay time for an oblique

slice and renamed FIDOBL on the GE console. This modification

basically allows for a larger gradient strength, still within the

system and safety limits, so that the RF and gradient pulses are as

tall as possible, with the area and thus power remaining the same. A

taller pulse takes less time, therefore, the delay time before

acquisition was reduced. In addition, the need to change the pwgph

control variable is eliminated.

The FIDCSI pulse sequence can be utilized in one other way. The

sequence can be set up to obtain a localized slice, which is segmented

into multiple voxels. The limitations of this procedure come from

inflexible placement of the multivoxels, timing necessary to acquire

CSI phase encoding steps, and unwanted spectral bleed due to point-

spread function inherent with CSI and the Fourier transform. As shown

in Figure 39, the multivoxel FIDCSI with CSI option allows for a set of

voxels to be placed within a set field of view. The field of view is

centered on the image and the voxel placement within the field of view

is dependent on how many times in the x and y direction the field of

view is broken up by phase encoding gradient steps. It should also be

noted that the standard GE FIDCSI pulse sequence on the 3.0 Tesla does

not allow for less than a 36 cm FOV if you let the console continue to

think it is operating at the proton frequency while the acquisition is

at a phosphorus frequency. This was the initial setup recommended by

GE. In this case it is also necessary to change control variables

"asfov" to 48 and "GAM" to 1723.5 to ensure the correct localization

dimensions. It was found that with a few tricks with the control

variables, a minimum FOV of 14 cm can be achieved. This is

accomplished by setting the console to run at the P-31 frequency and

then changing the following control variables to prevent system errors:

"pibbandfilt" equals 0 and "pixmtband" equals 1. These control

variables correct for the absence of a separate RF amplifier for P-31.

The 3.0 Tesla system uses one amplifier for all frequencies, unlike the

default of separate amplifiers expected by the software.

Finally, with some pulse programming corrections, as has been

done with the modified FIDCSI protocol locally called FIDFOVH, FOVs

below 14 cm can be achieved. Dr. Hee-Won Kim performed the

modification of the FIDCSI pulse sequence program with EPIC pulse

programming. The modification basically removed the protection limits

for the gradient amplifiers. The FIDFOVH, modified FIDCSI pulse

sequence, should only be used without the autoprescan, as parts of the

autoprescan (where transmitter gain and receiver amplifiers are

maximized) will exceed the limits of the gradients and do so without

warning. Autoprescan is generally useless for phosphorus spectroscopy,

therefore it would be a mistake to use autoprescan with phosphorus.

This allows the FIDFOVH modified sequence to be used safely.

1.5 Tesla to 3.0 Tesla Phosphorus Spectroscopy Comparisons

One of the main arguments for using 3.0 T over using a 1.5 T

magnet is the significant improvement in the amount of signal that is

obtained with an increase in field strength. To prove this point,

proton and phosphorus spectra of the depth phantom were taken at 1.5 T

and 3.0 T with similar parameters and compared. In addition, the same

parameter was again compared at 3.0 T to compare the single-turn and

quadrature RF coils.

The first comparison was of an FIDCSI acquisition with CSI voxel

localization of the depth phantom, as shown in Figure 40. In each case

the parameters were set as follows: 256 acquisitions, 25 mm thick

slice, 8 x 8 x 1 CSI, 16 cm field of view, 2 x 2 x 2.5 cm' voxels, scan

times = 4:26, and 10Hz line-broadening. In each case the phantom was

positioned so that one of the voxels would select just the inner

compartment of the depth phantom, at a depth of 5.5 cm.

Experimentally, the resulting SNR of the MDPA peak of the P-31 spectrum

for the single turn RF coil at 1.5 T was 7.4 while at 3.0 T it was

21.65, and the SNR of the quadrature RF coil at 3.0 T was 26.1. This

shows a significant improvement in signal from 1.5 T to 3.0 T, but only

a moderate improvement of the quadrature over the single turn coil.

The MDPA TI relaxation rate was also measured at 1.5 T and 3.0 T and

found to be 5.52 and 6.04 seconds, respectively (Appendix G). These

relaxation rates are negligibly different from each other and do not

significantly alter the MDPA SNR at each field strength.

The second comparison was of the modified ISISCSI acquisition of

a voxel positioned over the center compartment (5.5 cm depth) of the

depth phantom, as shown in Figure 41. The parameters used for the

ISISCSI acquisitions were as follows: 256 acquisitions, 16 cm field of

view, 2 x 2 x 2 cm3 voxel, scan times = 2:13 for each of 8 acquisition,

and 10Hz line-broadening. Experimentally, the resulting SNR of the

MDPA peak of the P-31 spectrum for the single turn RF coil at 1.5 T was

6.2 while at 3.0 it was 10.1, and the SNR of the quadrature RF coil at

3.0 T was 14.5. This shows a moderate improvement in signal from 1.5 T

to 3.0 T and another moderate improvement in the results of the

quadrature over the single turn coil.

To verify the extent to which the slice profile of the oblique,

FIDCSI modified slice selection is accurate, a series of tests on the

1.5 and 3.0 Tesla were conducted using the slice profile phantom.

Slices of 25 mm thick, tilted at an angle though the vial, were

obtained across the vial, as shown in Figure 8. The results of the

signal profile over the slice are shown in Figure 42 for the 1.5 Tesla

with the single turn P-31 coil and Figure 43 and Figure 44 for the 3.0

Tesla for the single turn and quadrature P-31 coils, respectively. The

slice profile tails (where none of the H2PO4 phantom was in the slice)

widened slightly more at 3.0 T, most likely due to the same gradient

pulse from the same pulse sequence on both systems despite different

magnet field strengths. This is shown via slightly larger area of

signal when no part of the phantom was in the slice (representing

potential contamination by skeletal muscle or blood) at 3.0 T, as shown

in Figure 43 and Figure 44, and compared to 1.5 Tesla as shown in

Figure 42.

The volume where the slice selection was below the phantom

represents the potential contamination from skeletal muscle. This is

due to the fact that the phantom, in that case being above the selected

Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EFW5Y9FNT_JU1PFD INGEST_TIME 2013-01-23T15:32:21Z PACKAGE AA00012991_00001