Cervical spine MR imaging using multislice gradient echo imaging: Comparison with cardiac gated spin echo

Magnetk Resonance Imagmg, Vol. 6. pp. 517-525. Printed !n the USA. All rights reserved.

1988

0730-725X/88 $3.00 + .OO Copyright 0 1988 Pergamon Press plc

l Original Contribution

CERVICAL SPINE MR IMAGING USING MULTISLICE GRADIENT ECHO IMAGING: COMPARISON WITH CARDIAC GATED SPIN ECHO MADAN V. KULKARNI,* PONNADA A. NARAYANA, * CRAIG B. MCARDLE, * JOEL W. *University

YEAKLEY,*

NICOLAS F. CAMPAGNA,~ AND FELIX W.

of Texas Medical School, at Houston tGenera1 Electric Corporation,

and Hermann Hospital, Milwaukee, Wisconsin,

Houston, USA

WEHRLI~ Texas 77025,

Forty-one patients with suspected cervical spine disorders were studied using multislice gradient echo imaging (GE) technique, with a 1.5-T system. The images were compared to cardiac-gated spin echo (CGSE) images in the diagnosis of suspected cord and spinal disorders. Images were graded for ability to detect cord lesion, cord-CSF contrast, CSF-bone contrast and contrast between CSF and extradural abnormality. The signal-to-noise ratio and contrast-to-noise ratio were used to compare images. There was 44% decrease in contrast between cord lesion and normal cord on GE when compared to CGSE, except for spinal cord hemorrhage. There was a 40% improvement between bone and CSF contrast on GE compared to CGSE. GE images were significantly better qualitatively as well as quantitatively in the detection of extradural lesions. This effect was more marked in axial plane where CGSE images are extremely suboptimal. CGSE images are better than GE for spinal cord lesions, while GE are superior in the diagnosis of degenerative disease in the cervical spine.

Keywords: MR imaging, spine; MR imaging, technique; Spine disorders.

INTRODUCTION

pulse sequence.’ By appropriately selecting flip angle and timing parameters, the signal from CSF in the cervical canal can be enhanced. This sequence is termed Interleaved GE. This sequence can be pulse synchronized (PS) to the cardiac cycle to obtain images in the same phase of cardiac cycle. Results of 41 patients with various clinically suspected cervical spine disorders are presented. A comparison between GE technique and cardiac gated spin echo (CGSE) technique is made quantitatively and qualitatively for CSF-cord contrast, and CSF-extradural tissue contrast. These pulse sequences are also compared for their diagnostic ability in detecting cord lesions and degenerative disease.

Magneticresonance(MR) imaging

has been extremely useful in the diagnosis of spinal cord disease, and is now increasingly used as the scanning modality for degenerative disease of cervical spine. Although MR has excellent inherent soft tissue contrast between cerebrospinal fluid (CSF) and cord, the CSF pulsatile motion causes signal loss from CSF and create phase shift artifacts.‘,5’14 This results in loss of contrast between CSF, and cord and between CSF and extradural tissues, resulting in loss of diagnostic information.“,14 In addition phase ghosting artifacts overlap and distort the image and cause CSF signal intensity loss. CSF gating techniques have been utilized to eliminate these problems.3*5s7 Similarly gradient recalled acquisition in the steady state has also been utilized in the diagnosis of cervical spinal disorders.14 We have utilized gradient recalled echo imaging techniques incorporated into a multislice, multiecho

RECEIVED l/5/88;

METHODS AND MATERIAL Forty-one patients underwentMR imaging of the cervical spine using GE techniques. Imaging was performed using a General Electric Medical System

Miehl for manuscript preparation, Ms. Elizabeth Berry and Ms. Cynthia Brewer for technical assistance. Address correspondence to Madan V. Kulkarni, M.D.

ACCEPTED 2/l/88. authors wish to thank Ms. Alison

Acknowledgment-The

517

518

Magnetic Resonance Imaging 0 Volume 6, Number 5, 1988

SIGNA imager (Milwaukee, WI) operating at 1.5 T. Cervical spine imaging was performed using a saddle shaped surface coil built at our institution. Patient consent was obtained before MR imaging. In 28 patients MR imaging was performed by means of a multislice gradient echo technique described previously.’ At the shortest echo delay of 8.5 ms the pulse sequence has a sequence repetition time (time interval between successive echoes) of 14.4 ms, thus allowing excitation of 20 slices in a 300 ms TR period. The pulse sequence supports fields of view up to 48 cm and allows gating and acquisition of multiple gradient echoes. Among the patients studied nine were normal, 21 had disc disease or degenerative hypertrophic changes, 6 had cervical spine trauma and 3 had spinal tumor and 2 with congenital anomalies. When available, the diagnosis was confirmed either by surgery (3 tumors, 7 discs) or by other modalities, such as CT with or without metrizamide (12 degenerative and disc disease, 4 normals, 1 congenital anomaly and 5 trauma) and 9 by clinical and neurological examination and follow up. In 28 patients GE imaging was performed using the following protocol. Two echoes with TE of 9 and 25 ms, TR of 750 ms, and flip angle of 30”. In an additional 21 patients GE images were pulse synchronized to ECG “R” Wave. GE images were acquired in the sagittal and axial planes. CGSE images were acquired in all patients in the sagittal plane. Seven patients had CGSE imaging in axial plane. In addition, 11 patients who underwent axial CGSE before GE software was available, were also included for qualitative as well as quantitative evaluation. In the axial and sagittal planes, multiplanar images were obtained with both CGSE and GE technique using a slice thickness of 5 mm separated by a 1 mm gap. The acquisition matrix was 256 x 128 and field of view was 12 cm for axial and 16 cm for sagittal images corresponding to a pixel size of 0.094 x 0.047 mm and 1.25 x 0.63 mm respectively. Two excitations were used for sagittal and four for axial images. CGSE imaging was performed using TE of 30 and 70 ms. Three to four heart beats/TR were used depending on the patient’s heart rate, resulting in an effective TR between 2000 to 3000 ms. The midline sagittal images from CGSE, GE, and pulse synchronized (PS) GE as well as axial CGSE, GE and PS GE were qualitatively evaluated by one of the authors (MVK) using the rating scale4 as follows: 1 = CSF isointense with and indistinguishable from the spinal cord, 2 = CSF minimally higher than that of the cord, 3 = high signal CSF but indistinct interfaces with the cord and or dura and 4 = high signal

CSF with sharp cord and dura interfaces. The mean of the quality rating for CGSE, and GE studies was calculated. Quantitative analysis using signal-to-noise ratio (SNR), and contrast to noise ratio (CNR) were calculated. CNR were determined between cord/CSF, CSF/extradural tissue, CSF/disc herniation and cord/ intramedullary lesions. The signal and the background levels along with the standard deviations were measured in cord and CSF for all midline sagittal images. The SNR was computed using:9

SNR =

Signal - Background Standard deviation of the background

The CNR was computed using following formula”

‘%I- Sb CNR

=

(a,2 +

&l/2

where S, and Sb are the measured signal levels for tissues a and b with a, and (Jb representing the corresponding standard deviations. The denominator in the above equation represents the root mean square noise. The SNR and CNR were computed for all the patients. The means were determined and relative ratios in percentages were calculated and plotted. RESULTS The mean quality rating using CSF cord distinction and relative signal on a scale of 1 to 4 was best for GE. The rating for axial GE was 3.8 and for sagittal GE was 3.7. The rating for CGSE in axial and sagittal was 2.3 and 3.3 respectively. The quantitative SNR comparison between CGSE and GE for CSF and cord is shown in Fig. 1 and comparison between these two sequences in CNR for the cord lesion and for extradural tissues is shown in Figs. 2 and 3. On the sagittal images, the CSF and spinal cord distinction was consistently better on GE images compared to CGSE (Fig. 4). Axial GE images were also excellent for the demonstration of cord/CSF distinction (Fig. 5). The diagnosis of spinal cord lesions in general was superior with CGSE than with GE (Fig. 6). Lesions with increased signal intensity on T2-weighted images ( T2-WI), such as MS plaques, cord edema, compressive myelopathy and spinal cord neoplasm were bet-

Cervical Spine MR Imaging 0

MADAN V. KULKARNIET AL.

519

CGSE

fzl CGSE q GE

r

CSFA

SIGNALTO NOISERATIO Fig. 1. Relative SNR with GE compared to CGSE in spinal cord and CSF using quantitative evaluation.

d

IWTE CSF/DISC

Fig. 3. Quantitative comparison between CGSE and GE for cord/CSF and CSF/extraduraI tissues. GE has consistently better CNR compared to CGSE. Marked improvement in contrast between CSF and bone is not only due to enhancement of CSF signal on T,-WI but also due to further decrease in signal in the bone due to susceptibility which is marked on GE.

CGSE

LESION/CORD

HEtlORRHAGEKOfUI

Fig. 2. CGSE has better CNR for most of the cord lesions compared to GE. These lesions have increased signal intensity on r,-WI. Similar quantitative evaluation of spinal cord hemorrhage demonstrates better CNR with GE compared to CGSE.

ter seen on CGSE (Fig. 7). When lesions were seen on both CGSE and GE, the diagnosis was made with higher confidence level with CGSE than with GE. In one patient with acute spinal trauma, the cord hemor-

rhage was apparent on both sequences, although the size of the hemorrhage was larger and better seen on GE sequence (Fig. 8). The relative ratio of the contrast between hemorrhage and cord CGSE was 0.56 on GCSE compared to GE. Sagittal images in general were far superior in showing the spinal cord lesions and in demonstrating the extent of abnormality. Bone and CSF contrast was seen better on GE. In the diagnosis of hypertrophic degenerative changes or disc herniation, GE images were consistently superior to the CGSE images. Osteophyte/CSF and disc/CSF contrast ratios on &WI between GE and CGSE were 2.78 and 1.47 respectively. Osteophyte spurs had decreased signal compared to the bright signal of CSF on 7”-WI (Fig. 9). Similarly, foraminal stenosis due to hypertrophic changes and sclerotic changes in the uncinate process were seen on axial GE images (Fig. 10). Herniated discs also had decreased signal on GE and CGSE compared to CSF on T,-WI (Fig. 11) although signal from the herniated discs was variable depending on disc dessication and degenerative changes. Table 1 demonstrates the relative rating for MR study quality for sagittal and axial planes for both pulse sequences. CGSE in axial plane were significantly poor quality, when compared to GE axial images.

Magnetic Resonance Imaging 0 Volume 6, Number 5, 1988

B Fig. 4. Sagittal CGSE image with TE of 80 msec and TR of 2100 msec fails to demonstrate cord CSF distinction. No definite extradural defects can be identified (Fig. 4A). GE with TE of 25 and TR of 750 msec and flip angle of 30” shows excellent contrast between bulging discs (arrows) and CSF (Fig. 4B). Contrast between cord and CSF is also seen better with GE.

Table 1. Qualitative evaluation of GE and CGSE images ((r/o) Good

Average

Poor

Sagittal Axial

70 21

18 37

12 42

GE Sagittal Axial

90 81

3.5 15

CGSE

6.5 4

DISCUSSION MR imaging of the cervical spine is being increasingly used as a screening imaging modality. Because of its excellent inherent contrast, spinal cord and cord lesions are better seen with MR. If,-WI can frequently diagnose disease processes, in spite of only

minimal or no enlargement of the spinal cord. In addition, MR is also widely used for the diagnosis of degenerative processes involving the cervical spine. Although T,-WI can adequately demonstrate cord morphology, they are suboptimal in demonstrating spinal cord pathology. Because of inherent decreased signal or CSF on these images, the demonstration of

extradural defects such as hypertrophic osteophytic changes and disc herniation is frequently not possible. T&WI, on the other hand, are useful in demonstrating these extradural defects since relatively hypointense osteophytes and herniated discs should be well outlined by the hyperintense signal of CSF on T2-WI. In actual practice this is frequently not possible due to CSF oscillatory motion. 2-‘o Because MR images are reconstructed using 2 dimensional Fourier transformation, the oscillatory CSF motion results in image degradation in two different ways. There is signal loss due to CSF motion similar to that seen in blood vessels, and second is due to the “ghosting” artifact. The “ghosting” or “phase-shift image” is due to spatial mismapping of moving CSF along the phase encoding direction. These effects are seen at all field strengths but are more pronounced at higher field strengths. Frequency and severity of this phenomenon is worse in the cervical spine.14 Therefore Tl- as well as T2-weighted spin echo images are frequently suboptimal in diagnosing degenerative changes in cervical spine. 4*‘4 In addition, subtle changes in spinal cord abnormality are also difficult to identify due to overlapping of “phase-shift images” over abnormal spinal cord signal. This results in false negative or false positive study for cord diseases.2s’4 CSF gating techniques have been performed using

Cervical Spine MR Imaging 0

Fig. 5. Axial GE image of cervical spine demonstrates normal cord. Grey white matter differentiation and cord/CSF distinction in the cervical cord are seen on this 7’,-WI. CSF is bright (arrow) and bone has marked decreased signal (curved arrow) due to increased susceptibility effects on T,-weighted GE.

MADAN V. KULKARNI ET AL.

521

EKG gated “R” wave detection or by triggering off the capillary blush of a finger or toe by means of photoplethysmograph. 5,7 Previous reports have demonstrated improved sensitivity and specificity of diagnosing spinal cord, brain and pituitary disorders using CGSE images.3r5*7Although CGSE images were good in sagittal plane, the CGSE in axial plane had poor reproducibility and poor CSF/cord contrast in our experience. Poor results of CSF gating in axial plane have also been reported by others.4 GE imaging has been utilized in the past using the single slice method.4 This technique offers good contrast between CSF and cord, but was reported to have a poor signal to noise ratio.4 We have utilized a new software to obtain multislice images using GE sequence. This software utilizes relatively longer TR interval to obtain multiple GE images at different levels during single acquisition. Even with a long TR (i.e. 750 ms), the images showed surprisingly increased signal from CSF relative to cord. The relatively increased signal seen in CSF compared to cord using GE technique in single slice axial plane utilizes time of flight principles. lo CSF enhancement with short TR gradient echo images has also been describedI and is due to residual magnetization car-

Fig. 6. Sagittal 7’,-weighted CGSE image (Fig. 6A) demonstrates central cord edema (arrow) in the distal cervical cord. Normal cord (curved arrow) has decreased signal intensity compared to cord edema. Also note multiple “ghosting” artifacts along the long axis of the cervical spine. Evaluation of small cord lesion is sometimes difficult due to these artifacts projecting over the cord. GE T2-WI (Fig. 6B) performed at the same location demonstrates good contrast between CSF and cord but cord edema is not well identified in this image.

Magnetic Resonance Imaging l Volume 6, Number 5, 1988

522

A

B

Fig. 7. Spinal cord astrocytoma is well documented on Tz-weighted sagittal GSE image (Fig. 7A). Cord is enlarged and tumor/edema extension to the lower cervical cord (arrow) can be seen and distinguished from normal cord (curved arrow) on CGSE image. GE r,-WI at the same level shows cord enlargement and abnormal cord signal (Fig. 7B) but fails to demonstrate constrast between tumor and normal cord contrast as seen on CGSE image.

A

B

Fig. 8. CGSE images in acute spinal trauma patient demonstrates spinal cord hemorrhage (arrow) and spinal cord edema (open arrows) in two serial sagittal T2-WI (Fig. 8A and 8B). Acute hemorrhage has decreased signal, due to presence of deoxyhemoglobin, on T,-WI while edema has increased signal. Also note minimally increased signal intensity in the fractured C5 due to presence of hemorrhage. On GE sagittal images obtained in the same planes (Fig. 8C and 8D), spinal cord hemorrhage is seen better and appears larger. This is due to enhanced susceptibility effects seen on GE. Cord edema (curved arrow) is not as well documented on GE as in CGSE. Susceptibility effects also predominate in the vertebral body, and hemorrhage in fractured CS which is seen on CGSE is not well appreciated on GE (Fig. 8D).

Cervical

Spine MR Imaging

0 MADAN V. KULKARNIET AL.

523

Fig. 9. T2-weighted sagittal CGSE image (Fig. 9A) shows osteophytes and disc bulging at multiple levels in cervical spine. Focal canal narrowing is seen at C4-5, (25-6 and C6-7 levels. GE image (Fig. 9B) at same level demonstrates similar findings and contrast between osteophytes, CSF and spinal cord is better seen on GE.

ried over in the next excitation due to TR’s in the order of 20-25 ms. In this case the steady state transverse magnetization becomes a function of T, and T,. The relative T, influence is reduced by using small flip angles. Since CSF has long T, it provides a relatively increased signal compared to grey and white matter. Using GE technique, where multislice imaging is performed, the time of flight phenomenon probably does not occur, in all slices. Also TR’s are significantly prolonged and a steady state is not achieved. Hence, the phenomenon by which CSF has relatively higher signal than cord on axial single slice GE does not hold true for axial multislice GE sequence. Although the exact cause of relative increased signal of CSF on multislice GE sequence is not known, it is felt to be due to predominant spin density effects. Using a longer TE this contrast can be further improved by increasing the relative T,-weighting. Our experience demonstrates excellent contrast between cord and CSF and between CSF and extradural structures like bone, discs, and osteophytes. This contrast difference was seen on both sagittal and axial planes. Although CGSE images also showed good contrast between extradural tissues and CSF, GE images were slightly superior. This may not only be due to increased signal intensity from CSF but also to further decrease in signal from bone, osteophytes and the

fibrous annulus owing to susceptibility effects.‘2*‘5 These susceptibility effects are more marked on long TE and significantly more on GE than CGSE images. GE studies offered consistently good to excellent image quality when compared to CGSE in the sagittal plane (9OVo versus 70%). Although cord morphology was well seen on sagittal GE images, the spinal cord lesions were frequently not well identified and only faintly seen on these images. The sole exception was spinal cord hemorrhage. Spinal cord hemorrhage was seen superiorly on GE than on CGSE. This is due to inherent increased susceptibility seen on GE. Deoxyhemoglobin or intracellular methemoglobin in acute CNS hemorrhage results in decreased signal on T2-WI.‘* Part of the reason for this decreased signal is susceptibility which is better demonstrated on GE than on CGSE images. Diagnosis of acute spinal cord hemorrhage is important since it signifies poor prognosis for neurologic recovery.* Primary cord tumor, multiple sclerosis plaques involving spinal cord, compressive myelopathy and spinal cord edema showing increased signal on T,-WI were significantly better appreciated on CGSE sequences. The quantitative estimation of contrast between lesion and cord in these cases were approximately twice as good on CGSE than on GE images. When the cord lesion was identified on both se-

524

Magnetic Resonance Imaging 0 Volume 6, Number 5, 1988

Fig. 10. T2-weighted axial GE demonstrates disc herniation (arrow) and foraminal stenosis on the left (open arrow). Due to focal canal stenosis the subarachnoid space is markedly attenuated and only minute portion of CSF is seen laterally (curved arrow).

quences, the CGSE frequently diagnosed the lesion with higher confidence level. The susceptibility can cause image degradation. Occasionally a small compression fracture can be seen on the sagittal &-weighted SE images because of increased signal within the vertebral body in the relatively acute phase of trauma.a On two occasions this increased signal was seen on CGSE but not on GE images. This is probably because susceptibility effects dominate over the increased signal from the T2 contribution. Increasing TE to more than 25-30 ms in GE sequence can lead to more T2 weighting, but also results in image degradation. Decreasing the flip angle to less than 30” can also improve relative T2 contribution in the image but leads to marked decrease in SNR. One significant advantage of GE is seen in the axial plane. Our experience and that of other observers have shown the suboptimal quality of axial CGSE images.4 Systolic gated axial images demonstrate boundary layer phase dispersion in the anterior and anterolateral subarachnoid space.’ Since CSF flow in this situation is perpendicular to the imaging plane, rather than in plane as in the sagittal image, even-echo rephasing does not occur.’ Single-slice CGSE in the axial plane using diastolic gating can provide good

Fig. Il. Central disc herniation (arrow) has decreased signal compared to bright signal of CSF on GE in axial plane. r,-WI using GE is also useful in differentiating bone (open arrow) and disc (curved arrow).

contrast between CSF, cord and extradural lesions, but routine scanning using this sequence is impractical since evaluation in the axial plane requires multiple levels in the cervical spine. Using multislice CGSE sequences, these multiple slices in different phases of the cardiac cycle with only 1 or 2 images being in the end diastolic phase, provide adequate CSF/extradural tissue contrast. GE images in the axial plane were best for the CSF contrast. Visualization of disc herniation, osteophytes and foraminal stenosis was excellent with axial GE. These images were excellent or of good quality in 81% of cases. Lesser quality was evident with axial GE in cases of subarachnoid space narrowing either with cord enlargement or spinal stenosis. Similar to the sagittal plane, these axial GE images had poor contrast between cord lesion and cord except for cord hemorrhage. GE images also offer improved SNR and marked reduction in data acquisition times compared to CGSE. Imaging times are reduced by 45 to 68% using GE compared to CGSE images. Hence, GE imaging may be suitable when shorter imaging times are preferred, for example, in children, patients with claustrophobia and patients with trauma and life support equipment _ Although there was minimal objective improvement in CSF/cord contrast with PS GE compared to GE in both the sagittal and axial planes,

Cervical Spine MR Imaging 0

further experience is necessary to determine the improvement in contrast quantitatively. Our current protocol for cervical spine imaging includes (a) sagittal T,-weighted localize using SE 20/800 image, (b) sagittal CGSE technique with TE of 30 and TE of 70 and (c) axial GE images with TR of 750, TE of 9, TE of 25 and flip angle of 30”. Although CGSE images are excellent in detecting cord lesion, in about 30% of cases the image quality was not adequate for CSF/extradural tissue contrast. In those situations, additional imaging using sagittal GE techniques is beneficial. We conclude that using this protocol, CGSE and GE complement each other. This protocol combines the superiority of CGSE for detecting cord lesions and the enhancement of CSF with GE for evaluating extradural defects. GE techniques are excellent for imaging in the axial plane and also offer the advantage of shorter imaging times. REFERENCES Compagna, N.F.; Wehrli, F.W. Ultra-short-TE imaging by means of slice-interleaved gradient recalled echoes at high sampling frequency. Book of Abstracts, Society of Magnetic Resonance in Medicine Meeting, New York, August, 1987. Enzmann, D.R.; Rubin, J.B.; DeLapaz, R.; Wright, A. Cerebrospinal fluid pulsation: Benefits and pitfalls in MR imaging. Radiology 161:773-778; 1986. Enzmann, D.R.; Rubin, J.B.; O’Donohue, J.; Griffin, C.; Drace, J.; Wright, A. Use of cerebrospinal fluid gating to improve &-weighted images. Part II. Temporal lobes, basal ganglia, and brain stem. Radiology 162:768-773;

1987.

MADAN

V.

KULKARNI ET AL.

525

4. Enzmann, D.R.; Rubin, J.B.; Wright, A. Cervical spine MR imaging: Generating high signal CSF in sagittal and axial images. Radiology 163:233-238, 1987.7, AJR 1987. 5. Enzmann, D.R.; Rubin, J.B.; Wright A. Use of cerebrospinal fluid gating to improve Tz-weighted images I. Spinal cord. Radiology 162:763-767; 1986. 6. Gomori, J.M.; Grossman, R.I.; Goldberg, H.I.; Zimmerman, R.A.; Bilaniuk, L.T. Intracraniai hematomas: Imaging by high field MR. Radiology 157:87-93; 1985. 7. Kulkarni, M.V.; Lee, K.F.; McArdle, C.B.; Yeakley, J.W.; Haar, F.L. 1.5 tesla MR of pituitary microadenomas: Technical considerations and CT correlation. AJNR (in press). 8. Kulkarni, M.V.; McArdle, C.B.; Kopaniky, D.; et al. Acute spinal cord injury: MR imaging at 1.5 T. Radiology 164:837-843;

1987.

9. Kulkarni, M.V.; Patton, J.A.; Price, R.R. Technical considerations for the use of surface coils in MRI. AJR 147:373-378;

1986.

10. Mills, C.M.; Brant-Zawadzki, M.; Crooks, L.E.; et al. Nuclear magnetic resonance: Principles of blood flow imaging. AJNR 4:1161-l 166; 1983. 11. Motz, J.; Danos, M. Information content and patient exposure. Medical Physics 5:8-15; 1978. 12. Perkins, T.U.; Wehrli, F.W. CSF signal enhancement in short TR gradient echo images. Magn. Reson. fmaging 4:465-467; 1986. 13. Rubin, J.B.; Enzmann, D.R. Harmonic modulation of proton MR precessional phase by pulsatile motion: Origin of spinal CSF flow phenomena. AJNR 8:307-318; 1987;AJR

148:983-994;

1987.

14. Rubin, J.B.; Enzmann, D.R. Imaging of spinal CSF pulsation by 2 DFT MR: Significance during clinical imaging. AJNR 8:297-306; 1987; AJR 148:973-982; 1987.

15. Wehrli, F. Introduction to fast-scan magnetic resonance. Published by General Electric, Milwaukee, WI, 1986.