Australas Radio1 1992; 36: 192-197
A Comparison Between M.R.I. and C.T. in Acute Spinal Trauma MORRY SILBERSTEIN, M.B., B.S. Department of Radiology Austin Hospital, Heidelberg, Victoria,Australia. BRIAN M. TRESS, F.R.A.C.R. Department of Radiology Royal Melbourne Hospital, Parkville, Victoria, Australia OLIVER HENNESSY, F.R.C.R. Department of Radiology Austin Hospital, Heidelberg, Victoria,Australia.
ABSTRACT Magnetic Resonance Imaging (MRI) at 0.3T and Computed Tomography (CT) were compared in the retrospective evaluation of 34 patients with acute spinal cord injury. MRI was highly accurate in the imaging of vertebral body fracture, and spondylitic changes, and is the method of choice for imaging ligament injury, traumatic disc protrusion and spinal cord compression. It was also useful for the identification of subtle subluxations in the sagittal plane. CT remains the method of choice for imaging neural arch fractures. MRI at 0.3T is a valid technique for assessing patients with acute spinal trauma.
Address for correspondence: Dr M. Silberstein University of Melbourne Fellow in Magnetic Resonance and Radiology Department of Radiology Austin Hospital Heidelberg 3084 Victoria Australia
192
INTRODUCTION Magnetic Resonance (MR) imaging has demonstrated great promise in the evaluation of the spinal cord following acute injury, both in correlating the appearances of the spinal cord with the level of neurologic deficit, and in identification of parameters predictive of neurologic outcome (1,2,3). If, as Mirvis et a1 (2) believe, MR is the method of choice in imaging symptomatic patients with acute spinal trauma, will other studies such as plain film radiography, computed tomography (CT) and myelography become obsolete in these patients, just as MR has relegated CT to a minor role in a wide variety of intra cranial and spinal conditions (4,5)? Although sporadic data has appeared in the literature comparing MR and CT in acute spinal trauma at lower field strengths (6,7), the only detailed comparison of these two modalities in acute spinal trauma was made at 1.5 Tesla (T) (3). This study was undertaken to compare MR and CT in the identification of imaging findings associated with acute spinal injury, and to determine the accuracy of MR at a lower field strength (0.3T) in the demonstration of parameters usually assessed with CT. MATERIALS AND METHODS The 34 patients included in the study were derived from patients admitted to the Austin Hospital Spinal Injuries Unit between 1987 and 1990. There were 22 males and 12 females with ages ranging from 12 to 70 years, with a mean of 34 years. The mechanisms of injury were: motor vehicle accident (15), fall ( 8 ) , diving (3), recreational sport (3), assault (2), and 3 miscellaneous. Twenty-two patients had cervical and 12 had thoracic injuries. Patients were not included if they had neurological deficits below
T12 (conus medullaris) of if images were technically unsatisfactory. This resulted in the exclusion of one patient who had image degradation by bullet fragments following a gunshot wound. All patients included had single level injury to the spine. MR imaging was performed on a 0.3 Tesla MR unit (B 3000, Fonar; Melville NY) on 31 patients, and a 1.5 Tesla super conducting MR unit (Magnetom 63; Siemens; Germany) on 3 patients. Images were obtained with a spin-echo (SE) technique consisting of two sagittal sequences (repetition time msec/echo time msec = 600/15 and 2500/80) and two axial sequences using similar acquistion times. Flow compensation was used on the long TR sequences. Slice thickness was 4mm with lm m interslice gap. Average time from injury to MR was 11 days (range 1 day to 42 days), with over 70% of examinations performed within 7 days. All examinations were performed during initial hospital admission. CT was performed on a Siemens unit (Somatom DR, Germany) with 30 patients having non-contrast studies, and 4 having CT with intrathecal contrast. Examinations consisted of contiguous 2mm slices with sagittal reformatting following identification of the site of abnormality on plain radiographs, or myelography. When these were normal, contiguous 4mm slices were obtained through the clinically suspected level and further 2mm slices were obtained through any abnormality revealed. All patients were examined during initial hospital admission.
Submitted for publication on: 13th August, 1991 Resubmitted for publication on: 23rd December, 1991 Accepted for publication on: 21st January, 1992
Australasian Radiology, Vol. 36, No. 3. August, 1992
M.R.I. AND C.T. IN ACUTE SPINAL TRAUMA
FIGURE 1A - Burst fracture L1 vertebral body. Axial CT scan shows comminuted fracture with displacement of fragment into spinal canal.
The MR and CT studies were retrospectively examined independently. The MR studies were assessed for the presence of vertebral body fracture; posterior element fracture; spondylosis (defined as posterior osteophyte formation in relation to an intervertebral disc) or disc narrowing, subluxation, prevertebral swelling, ligament injury by recognised criteria (3), disc hemiation, extramedullary haematoma (a localised collection of epidural or subdural blood with signal intensity greater than CSF on the T1 weighted images) and spinal cord compression (non - visualization of an area of subarachnoid space in relation to the spinal cord). In patients who had undergone CT myelography, the MR examinations were also assessed for the presence of cord transection and cord swelling based on previously published criteria (3, 8). The CT studies were assessed for the same imaging characteristics with the presence of ligament injury indentified by subjective criteria - significant degree of subluxation on sagittally reformatted images, anterior disc space widening or significant interspinous widening. Cord transection and cord swelling were assessed in the 4 patients with CT myelography using recognised criteria (9). Cord transection was defined as failure to identify the spinal cord on an axial image, and extramedullary haematoma was defined as abnormal soft tissue in the spinal canal not related to an inter-vertebraldisc (9). Australasian Radiology, Vol. 36, No. 3, August, 1992
FIGURE :1B - Sagittal T1 weighted M R image demonstrates fracture line with T12-L1 subluxation compressing lower spinal cord.
RESULTS
neimaging modality which reput-
edly offered optimal identification of a particular abnormality was used as the standard of reference. For fracture, both vertebral body and posterior elments,
subluxation and spondylosis, CT was used as the standard (Table 1). For Ob~rvationsrelating to the Paraspinal Soft tissues, MR was used as the standard *) as as for ObseNatiom to the cord*
TABLE 1 Value of CT in Detection of Soft Tissue Iniuries ( n = 34 ) Predictive Value Pos. Neg. Accur. Sens. Specif. Prevertebral Swelling
(%)
(%)
(%)
(%)
(%)
88
94
94
89
91
21
1M)
100
74
76
77
77
53
53
60
60
n=17
Ligament Injury n = 11 Disc Hemiation n=7 Extramedullary Haematoma n = 14 Cord Compression
n = 12
No intrathecal contrasr ( n = 30 ) 0 100 0 No intrathecal contrasr ( n = 30 ) 0 100 0 No intrathecal contrasr ( n = 30 ) 0 100 0
TBLE2 Value of MR in Detection of Osseous In.iuries( n = 34 ) Predictive Value Pos. Neg. Sens. Specif. Vertebral body fracture n = 10 Posterior element fracture n = 13 Subluxation n=8 Spondylosis n = 10
Accur.
(%)
(%I
(%)
(%)
(%)
91
96
91
96
94
23
100
100
68
71
100
100
100
100
100
100
100
100
100
100
193
M. SILBERSTEIN et a1
FIGURE 2A - T11 vertebral body and neural arch fractures. Axial CT scan shows comminuted left sided neural arch fracture, and oblique vertebral body fracture. Note ossific fragment in spinal canal anteriorly.
FIG.RE 2B T1 weighted MR image shows fraciure and associated cord compression.
For vertebral body fracture, MR was very sensitive and had very high specificity (Figure 1). One patient 194
FIGURE 2C - Axial T1 weighted image at similar level to A. The intraspinal fragment (linear area of low signal, mowed) is visualised, but the neural arch fracture is not identified.
with a false positive MR was originally reported as having a bone bruise in the T3 vertebral body but subsequent comparison with the plain films and CT upon conclusion of the study failed to demonstrate any bony trabecular abnormality. Although this does not exclude a bone bruise, the sharp definition of the margins of the lesion, its signal characteristics, and the fact that the patient’s clinical neurologic deficit did not correspond to this level, favoured an incidential haemangioma (Figure 2). h4R had a significant false negative rate for posterior element fractures (Figure 3). Non-contrast CT was an accurate method for identification of prevertebra1 swelling but failed to identify 63% of patients with ligament injury. It failed to identify any of the patients with disc herniation or extramedullary haematoma. Although the two patients with cord compression by extramedullary haematoma were correctly identified using CT with intrathecal contrast, only one of the two patients with disc herniation was correctly identified with this technique. Only one patient with cord transection, and one with cord swelling, had CT with intrathecal contrast, although both cases were Correctly identified (Figures 4 and 5).
Excluding the 3 patients with 1.5T MR imaging had very little effect on the overall statistical analysis. None of these 3 patients had CT with intrathecal contrast. Two of the ,three had posterior element fractures, only one of which was identified on MR. DISCUSSION Magnetic resonance has major advantages in spinal imaging when compared with other modalities. Direct sagittal images can be obtained without the necessity of reformatting images from axial data with the resultant loss of resolution. The spinal cord and subarachnoid space are visualised without intrathecal contrast, and specific tissue characteristics such as haemorrhage o r oedema can be demonstrated (10). In the acute spinal trauma setting, the major disadvantages are the prolonged imaging time for a full MR examination and the difficulty in examining a patient connected to monitoring equipment and head tongs ( 1 1). In our patients a typical examination was completed within 30-45 minutes and no major difficulties were encountered. One patient was excluded from the study on the basis of artefact from gunshot pellets resulting in unsatisfactory imaging of the spinal cord on MR. This also hampered CT assessment. Australasian Radiology, Vol. 36, No. 3, August, 1992
M.R.I. AND C.T. IN ACUTE SPINAL TRAUMA
FIGURE 3A - Spinal cord compression due to osteophytes and subluxation. Sagittal midline reformatted CT image shows C b c 7 subluxation with osteophytes projecting into spirial canal at C-C7 and C7-TI.
FIGURE 3B - Sagittal T1 weighted MR image clearly demonstrates cord compression as well as anterior epidural haernatoma. Note disruption of anterior longitudinal ligament at
FIGURE 3C - T2 weighted image. Osteophytes are visualised as low signal due to cortical bone. Note associated cord abnormality.
C5-C6.
Our results indicate that MR is an adequate method for assessing vertebral body fractures in the acute trauma setting. This was a better result than that of Flanders et a1 (3) who had a 20% false negative rate, in comparison to our 10% at 0.3T. When a fracture was identified on MR, this finding had 94% specificity in Flanders series which is similar to our result. In the series of Tracy et a1 (7) at OST, all 10 fractures in 13 patients with acute spinal trauma were identified on MR Australasian Radiology. Vol. 36, No. 3 , August, 1992
where CT was regarded as the standard. In another report of 14 patients with vertebral body fracture examined at OST, MR identified all cases (6). In view of the ability of MR to demonstrate bone bruises as marrow signal intensity changes even when plain radiography and CT are normal (12), this modality should be considered complementary to CT in the investigation of vertebral body fractures, although our false positive example of a vertebral haemangioma should be
kept in mind when making this diagnosis. Vertebral haemangioma is usually well defined with high signal intensity on both T1 and T2 weighted images and is frequently an incidental finding (13). Our results support those of Flanders et a1 (3) who found a 75% false negative rate for identification of posterior element fractures on MR. We similarly had a high positive predictive value (100%) but both series suggest that posterior element fractures are best evaluated by CT. This is less optimistic than the results of Tarr et a1 ( 6 ) who identified 57% of posterior element fractures in 14 patients, but the relevance of identification of these fractures to procedures implies that any false negative rate is unacceptable in the acute trauma setting-( 14). Although the ability of MR to demonstrate cortical bone was in doubt in the early days of imaging (15), we had no difficulty in identifying posterior osteophytes in patients with spondylosis and our 100% sensitivity and specificity compare favourably with the 95% recorded by Flanders et a1 (3). In contrast to Flanders, we used CT as the standard of reference for subluxation with 100% accuracy. In the series of Flanders, CT failed to identify 28% of patients with subluxations, and had a false positive rate of 13% using MR as standard. Tracy et a1 (7) had 6 patients with subluxations, with CT and MR providing equivalent estimates of displacement. Theorectically, CT might be deficient in imaging sagittal displacements because of the axial plane of section and loss of resolution in sagittal reconstructions (16), a problem not found in MR which has direct sagittal imaging. It may be that our 0.3T system failed to identify subtle subluxations in the sagittal plane because of its reduced inherent resolution, and that MR at 1.5T would have identified other patients with subluxations missed at 0.3T. However, none of the 3 patients with scans at 1.5T had subluxations, and review of the plain films performed at the conclusion of our study failed to reveal any subluxations not previously visualized. CT was accurate in identifying soft tissue swelling in the prevertebral region in our study. We did not find the high (65%) false negative rate of Flanders er a1 (3). These authors do not state the time interval between CT and MR, which may have influenced detection rate, as their specificity (97%) is similar to ours. The ability of MR to image liagment injury non195
M. SILBERSTEIN et a1
FIGURE 4A - Spinal cord compression by haematoma. Axial CT with intrathecal contrast demonstrates anterior epidural lesion.
invasively has made it the first risk free technique able to do this (8). Our technique for identification of this parameter on CT was relatively crude, involving inference of subluxation on the basis of subjective assessment of degree of displacement, and although we failed to identify 73% of ligament injuries, we had 100% specificity with our technique. Clearly, however, MR is the method of choice and may make late functional views redundant. We failed to identify any of the patients with disc herniation, extramedullary haematoma, or cord compression with unenhanced CT. This is similar to the results of Tracy et a1 (7) but differs from the experience of Flanders et a1 (93) who identified 44% of disc protrusions and 77% of patients with cord compression using CT. We did not routinely photograph CT images of the spine on narrow soft tissue windows, and the analyses in this study involved retrospective review of previously photographed images, which is the likely reason for our poor rate of detection. Nevertheless, MR is likely to be the method of choice for identification of these findings (3). In the patients who had CT with intrathecal contrast, there was no difficulty in identifying extramedullary haematoma, cord compression, swelling or transection, although 1 of the 2 patients with disc herniation who 196
FIGURE 4B - Axial T1 weighted MR image yields similar findings. Acute epidural haematoma was suspected clinically.
had this study was not identified. The non-invasive nature of M R , however, will almost certainly make installation of intrathecal contrast redundant in the management of acute spinal trauma. In addition to the parameters assessed in this study, comparing CT and MR, specific identification of spinal cord lesions such as haemorrhage and oedema is possible with MR, a feature not offered by any other non-invasive imaging modality (1,2). Comparison of our results with previous studies at 1.5T, comparing CT and MR in acute spinal trauma suggests that M R at 0.3T yields a similar accuracy to previously reported series. Therefore, for the parameters assessed by this study, MR at 0.3T is a valid method of imaging in acute spinal trauma. One area in which low field MRI may be less than optimal is in the demonstration of the time-related changes in haemorrhage. Spin-echo sequences, as used in this study, frequently fail to demonstrate classical T2 shortening induced by intracellular deoxyhaemoglobin, which may only be visible using gradient recalled sequences (17). Evaluation of acute spinal trauma at lower field strengths should, therefore, include images obtained with gradient-echo techniques, as well as standard spin-echo sequences.
Rapid advances in high field strength MR are occurring, and with the increasing use of volume acquisitions allowing contiguous slice reconstructions, fat saturation techniques, and gradient echo techniques, resulting in high contrast between CSF and surrounding tissues, the ability of h4R to image subtle fractures and neural arch fractures is likely to improve to a stage where CT will no longer be required in the acute trauma setting (18, 19). CONCLUSION MR at 0.3T is highly accurate in the imaging of vertebral body fractures and spondylosis in the acute trauma setting. For ligament injury, prevertebral swelling, disc herniation and cord compression, it is the method of choice, out-performing CT. However, CT will still be required for demonstration of posterior element fractures. Although CT and MR were equally accurate in the demonstration of subluxations in our study, the results of previous studies at 1.5T suggest that an element of caution be maintained when excluding subtle subluxations using CT alone, and the theoretically poorer resolution of 0.3T MR should be kept in mind when attempting to make this diagnosis at lower field strengths.
Australasian Radiology, Val. 36, NO. 3 , August, 1992
M.R.I. AND C.T. IN ACUTE SPINAL TRAUMA
REFERENCES
FIGURE 5A - Spinal cord transection at T3. Axial CT images with intrathecal contrast demonstrate contrast within the cord superiorly, with cord discontinuity below.
1. Kulkarni MV, Bondurant FJ, Rose SL, Narayana PA. 1.5 Tesla magnetic resonance imaging of acute spinal trauma. Radiographics 1988; 8: 1059-1082. 2. Mirvis SE, Geisler FH, Jelinek JJ, Joslyn JN, Gellad F. Acute cervical spine trauma: evaluation with 1.5T MR imaging. Radiology 1988; 166: 807-816. 3. Flanders AE, Schaefer DM, Huynh TD, Mishkin MM, Gonzalez CF, Northrup BE. Acute cervical spine trauma: correlation of MR imaging findings with degree of neurologic deficit. Radiology 1990; 177: 25-33. 4. Hyman RA, Gorey MT. Imaging strategies for MR of the brain. Rad Clin North Am 1988; 26: 471-503. 5 . Hyman RA, Gorey MT. Imaging strategies for MR of the spine. Rad Clin North Am 1988; 26: 505-533. 6. Tarr RW, Drolshagen LF, Kerner TC, Allen JH, Partain CL, James AE Jr. MR imaging of recent spinal trauma. J Comput Assist Tomog 1987; 11: 4 12-417. 7. Tracy PT, Wright RM, Hanigan WC. Magnetic resonance imaging of spinal injury. Spine 1989; 14: 292-301. 8. Goldberg AL, Rothfus WE, Deeb ZL et al. The impact of magnetic resonance on the diagnostic evaluation of acute cervicothoracic spinal trauma. Skelet Radio1 1988; 17: 89-95. 9. Cooper PR, Cohen W. Evaluation of cervical spinal cord injuries with metrizamide mydography - CT scanning. J Neurosurg 1984; 61: 281-288. 10. Han JS, Benson JE, Yoon YS. Magnetic resonance imaging in the spinal column and craniovertebral junction. Rad Clin North Am 1984; 22: 805-827. 11. McArdle CB, Wright JW, Provest WS, Dornfest DJ, Amparo EG. MR imaging of the acutely injured patient with cervical traction. Radiology 1986; 159: 273-274. 12. Deutsch AL, Mink JH. Magnetic resonance imaging of musculoskeletal injuries. Rad Clin North Am 1989; 27: 983-1001. 13. New PFJ. Shoukimas GM. Thoracic s h e and spinal cord. In: Stark DD, Bradley WG, (eds). Magnetic resonance imaging. St Louis; CV Mosby: 1988: 662. 14. Jacobs RR, Casey MP. Surgical management of thoraco-lumbar spinal injuries: general principles and controversial considerations. Clin Orthop Re1 Res 1984; 189: 22-35. 15. Harms SE, Greenway G. The musculoskeletal system. In: Stark DD, Bradley WG (eds.). Magnetic resonance imaging. St Louis, CV Mosby; 1988: 1323. 16. Murphy MD, Batnitzky S, Bramble JM. Diagnostic imaging of spinal trauma. Rad Clin North Am 1989; 27: 855-871. 17. Zimmerman RD, Heier LA, Snow RB, Liu DPC, Kelly AB, Deck MDF. Acute intracranial haemorrhage: intensity changes on sequential MR scans at 0.5T. AJNR 1988; 9: 47-47. 18. Burke DL Jr, Dalinka MK, Schiebler ML, Cohen EK, Kressel HY. Strategies for musculoskeletal magnetic resonance imaging. Rad Clin North Am 1988; 26: 653-672. 19. Kricun R, Kricun ME. Advances in spinal imaging. Rad Clin North Am 1990; 28: 321-339.
FIGURE 5B -Axial TI weighted MR image confirms CT finding.
Australasian Radiology, Vol. 36, No. 3 , August, 1992
197
A Comparison Between M.R.I. and C.T. in Acute Spinal Trauma MORRY SILBERSTEIN, M.B., B.S. Department of Radiology Austin Hospital, Heidelberg, Victoria,Australia. BRIAN M. TRESS, F.R.A.C.R. Department of Radiology Royal Melbourne Hospital, Parkville, Victoria, Australia OLIVER HENNESSY, F.R.C.R. Department of Radiology Austin Hospital, Heidelberg, Victoria,Australia.
ABSTRACT Magnetic Resonance Imaging (MRI) at 0.3T and Computed Tomography (CT) were compared in the retrospective evaluation of 34 patients with acute spinal cord injury. MRI was highly accurate in the imaging of vertebral body fracture, and spondylitic changes, and is the method of choice for imaging ligament injury, traumatic disc protrusion and spinal cord compression. It was also useful for the identification of subtle subluxations in the sagittal plane. CT remains the method of choice for imaging neural arch fractures. MRI at 0.3T is a valid technique for assessing patients with acute spinal trauma.
Address for correspondence: Dr M. Silberstein University of Melbourne Fellow in Magnetic Resonance and Radiology Department of Radiology Austin Hospital Heidelberg 3084 Victoria Australia
192
INTRODUCTION Magnetic Resonance (MR) imaging has demonstrated great promise in the evaluation of the spinal cord following acute injury, both in correlating the appearances of the spinal cord with the level of neurologic deficit, and in identification of parameters predictive of neurologic outcome (1,2,3). If, as Mirvis et a1 (2) believe, MR is the method of choice in imaging symptomatic patients with acute spinal trauma, will other studies such as plain film radiography, computed tomography (CT) and myelography become obsolete in these patients, just as MR has relegated CT to a minor role in a wide variety of intra cranial and spinal conditions (4,5)? Although sporadic data has appeared in the literature comparing MR and CT in acute spinal trauma at lower field strengths (6,7), the only detailed comparison of these two modalities in acute spinal trauma was made at 1.5 Tesla (T) (3). This study was undertaken to compare MR and CT in the identification of imaging findings associated with acute spinal injury, and to determine the accuracy of MR at a lower field strength (0.3T) in the demonstration of parameters usually assessed with CT. MATERIALS AND METHODS The 34 patients included in the study were derived from patients admitted to the Austin Hospital Spinal Injuries Unit between 1987 and 1990. There were 22 males and 12 females with ages ranging from 12 to 70 years, with a mean of 34 years. The mechanisms of injury were: motor vehicle accident (15), fall ( 8 ) , diving (3), recreational sport (3), assault (2), and 3 miscellaneous. Twenty-two patients had cervical and 12 had thoracic injuries. Patients were not included if they had neurological deficits below
T12 (conus medullaris) of if images were technically unsatisfactory. This resulted in the exclusion of one patient who had image degradation by bullet fragments following a gunshot wound. All patients included had single level injury to the spine. MR imaging was performed on a 0.3 Tesla MR unit (B 3000, Fonar; Melville NY) on 31 patients, and a 1.5 Tesla super conducting MR unit (Magnetom 63; Siemens; Germany) on 3 patients. Images were obtained with a spin-echo (SE) technique consisting of two sagittal sequences (repetition time msec/echo time msec = 600/15 and 2500/80) and two axial sequences using similar acquistion times. Flow compensation was used on the long TR sequences. Slice thickness was 4mm with lm m interslice gap. Average time from injury to MR was 11 days (range 1 day to 42 days), with over 70% of examinations performed within 7 days. All examinations were performed during initial hospital admission. CT was performed on a Siemens unit (Somatom DR, Germany) with 30 patients having non-contrast studies, and 4 having CT with intrathecal contrast. Examinations consisted of contiguous 2mm slices with sagittal reformatting following identification of the site of abnormality on plain radiographs, or myelography. When these were normal, contiguous 4mm slices were obtained through the clinically suspected level and further 2mm slices were obtained through any abnormality revealed. All patients were examined during initial hospital admission.
Submitted for publication on: 13th August, 1991 Resubmitted for publication on: 23rd December, 1991 Accepted for publication on: 21st January, 1992
Australasian Radiology, Vol. 36, No. 3. August, 1992
M.R.I. AND C.T. IN ACUTE SPINAL TRAUMA
FIGURE 1A - Burst fracture L1 vertebral body. Axial CT scan shows comminuted fracture with displacement of fragment into spinal canal.
The MR and CT studies were retrospectively examined independently. The MR studies were assessed for the presence of vertebral body fracture; posterior element fracture; spondylosis (defined as posterior osteophyte formation in relation to an intervertebral disc) or disc narrowing, subluxation, prevertebral swelling, ligament injury by recognised criteria (3), disc hemiation, extramedullary haematoma (a localised collection of epidural or subdural blood with signal intensity greater than CSF on the T1 weighted images) and spinal cord compression (non - visualization of an area of subarachnoid space in relation to the spinal cord). In patients who had undergone CT myelography, the MR examinations were also assessed for the presence of cord transection and cord swelling based on previously published criteria (3, 8). The CT studies were assessed for the same imaging characteristics with the presence of ligament injury indentified by subjective criteria - significant degree of subluxation on sagittally reformatted images, anterior disc space widening or significant interspinous widening. Cord transection and cord swelling were assessed in the 4 patients with CT myelography using recognised criteria (9). Cord transection was defined as failure to identify the spinal cord on an axial image, and extramedullary haematoma was defined as abnormal soft tissue in the spinal canal not related to an inter-vertebraldisc (9). Australasian Radiology, Vol. 36, No. 3, August, 1992
FIGURE :1B - Sagittal T1 weighted M R image demonstrates fracture line with T12-L1 subluxation compressing lower spinal cord.
RESULTS
neimaging modality which reput-
edly offered optimal identification of a particular abnormality was used as the standard of reference. For fracture, both vertebral body and posterior elments,
subluxation and spondylosis, CT was used as the standard (Table 1). For Ob~rvationsrelating to the Paraspinal Soft tissues, MR was used as the standard *) as as for ObseNatiom to the cord*
TABLE 1 Value of CT in Detection of Soft Tissue Iniuries ( n = 34 ) Predictive Value Pos. Neg. Accur. Sens. Specif. Prevertebral Swelling
(%)
(%)
(%)
(%)
(%)
88
94
94
89
91
21
1M)
100
74
76
77
77
53
53
60
60
n=17
Ligament Injury n = 11 Disc Hemiation n=7 Extramedullary Haematoma n = 14 Cord Compression
n = 12
No intrathecal contrasr ( n = 30 ) 0 100 0 No intrathecal contrasr ( n = 30 ) 0 100 0 No intrathecal contrasr ( n = 30 ) 0 100 0
TBLE2 Value of MR in Detection of Osseous In.iuries( n = 34 ) Predictive Value Pos. Neg. Sens. Specif. Vertebral body fracture n = 10 Posterior element fracture n = 13 Subluxation n=8 Spondylosis n = 10
Accur.
(%)
(%I
(%)
(%)
(%)
91
96
91
96
94
23
100
100
68
71
100
100
100
100
100
100
100
100
100
100
193
M. SILBERSTEIN et a1
FIGURE 2A - T11 vertebral body and neural arch fractures. Axial CT scan shows comminuted left sided neural arch fracture, and oblique vertebral body fracture. Note ossific fragment in spinal canal anteriorly.
FIG.RE 2B T1 weighted MR image shows fraciure and associated cord compression.
For vertebral body fracture, MR was very sensitive and had very high specificity (Figure 1). One patient 194
FIGURE 2C - Axial T1 weighted image at similar level to A. The intraspinal fragment (linear area of low signal, mowed) is visualised, but the neural arch fracture is not identified.
with a false positive MR was originally reported as having a bone bruise in the T3 vertebral body but subsequent comparison with the plain films and CT upon conclusion of the study failed to demonstrate any bony trabecular abnormality. Although this does not exclude a bone bruise, the sharp definition of the margins of the lesion, its signal characteristics, and the fact that the patient’s clinical neurologic deficit did not correspond to this level, favoured an incidential haemangioma (Figure 2). h4R had a significant false negative rate for posterior element fractures (Figure 3). Non-contrast CT was an accurate method for identification of prevertebra1 swelling but failed to identify 63% of patients with ligament injury. It failed to identify any of the patients with disc herniation or extramedullary haematoma. Although the two patients with cord compression by extramedullary haematoma were correctly identified using CT with intrathecal contrast, only one of the two patients with disc herniation was correctly identified with this technique. Only one patient with cord transection, and one with cord swelling, had CT with intrathecal contrast, although both cases were Correctly identified (Figures 4 and 5).
Excluding the 3 patients with 1.5T MR imaging had very little effect on the overall statistical analysis. None of these 3 patients had CT with intrathecal contrast. Two of the ,three had posterior element fractures, only one of which was identified on MR. DISCUSSION Magnetic resonance has major advantages in spinal imaging when compared with other modalities. Direct sagittal images can be obtained without the necessity of reformatting images from axial data with the resultant loss of resolution. The spinal cord and subarachnoid space are visualised without intrathecal contrast, and specific tissue characteristics such as haemorrhage o r oedema can be demonstrated (10). In the acute spinal trauma setting, the major disadvantages are the prolonged imaging time for a full MR examination and the difficulty in examining a patient connected to monitoring equipment and head tongs ( 1 1). In our patients a typical examination was completed within 30-45 minutes and no major difficulties were encountered. One patient was excluded from the study on the basis of artefact from gunshot pellets resulting in unsatisfactory imaging of the spinal cord on MR. This also hampered CT assessment. Australasian Radiology, Vol. 36, No. 3, August, 1992
M.R.I. AND C.T. IN ACUTE SPINAL TRAUMA
FIGURE 3A - Spinal cord compression due to osteophytes and subluxation. Sagittal midline reformatted CT image shows C b c 7 subluxation with osteophytes projecting into spirial canal at C-C7 and C7-TI.
FIGURE 3B - Sagittal T1 weighted MR image clearly demonstrates cord compression as well as anterior epidural haernatoma. Note disruption of anterior longitudinal ligament at
FIGURE 3C - T2 weighted image. Osteophytes are visualised as low signal due to cortical bone. Note associated cord abnormality.
C5-C6.
Our results indicate that MR is an adequate method for assessing vertebral body fractures in the acute trauma setting. This was a better result than that of Flanders et a1 (3) who had a 20% false negative rate, in comparison to our 10% at 0.3T. When a fracture was identified on MR, this finding had 94% specificity in Flanders series which is similar to our result. In the series of Tracy et a1 (7) at OST, all 10 fractures in 13 patients with acute spinal trauma were identified on MR Australasian Radiology. Vol. 36, No. 3 , August, 1992
where CT was regarded as the standard. In another report of 14 patients with vertebral body fracture examined at OST, MR identified all cases (6). In view of the ability of MR to demonstrate bone bruises as marrow signal intensity changes even when plain radiography and CT are normal (12), this modality should be considered complementary to CT in the investigation of vertebral body fractures, although our false positive example of a vertebral haemangioma should be
kept in mind when making this diagnosis. Vertebral haemangioma is usually well defined with high signal intensity on both T1 and T2 weighted images and is frequently an incidental finding (13). Our results support those of Flanders et a1 (3) who found a 75% false negative rate for identification of posterior element fractures on MR. We similarly had a high positive predictive value (100%) but both series suggest that posterior element fractures are best evaluated by CT. This is less optimistic than the results of Tarr et a1 ( 6 ) who identified 57% of posterior element fractures in 14 patients, but the relevance of identification of these fractures to procedures implies that any false negative rate is unacceptable in the acute trauma setting-( 14). Although the ability of MR to demonstrate cortical bone was in doubt in the early days of imaging (15), we had no difficulty in identifying posterior osteophytes in patients with spondylosis and our 100% sensitivity and specificity compare favourably with the 95% recorded by Flanders et a1 (3). In contrast to Flanders, we used CT as the standard of reference for subluxation with 100% accuracy. In the series of Flanders, CT failed to identify 28% of patients with subluxations, and had a false positive rate of 13% using MR as standard. Tracy et a1 (7) had 6 patients with subluxations, with CT and MR providing equivalent estimates of displacement. Theorectically, CT might be deficient in imaging sagittal displacements because of the axial plane of section and loss of resolution in sagittal reconstructions (16), a problem not found in MR which has direct sagittal imaging. It may be that our 0.3T system failed to identify subtle subluxations in the sagittal plane because of its reduced inherent resolution, and that MR at 1.5T would have identified other patients with subluxations missed at 0.3T. However, none of the 3 patients with scans at 1.5T had subluxations, and review of the plain films performed at the conclusion of our study failed to reveal any subluxations not previously visualized. CT was accurate in identifying soft tissue swelling in the prevertebral region in our study. We did not find the high (65%) false negative rate of Flanders er a1 (3). These authors do not state the time interval between CT and MR, which may have influenced detection rate, as their specificity (97%) is similar to ours. The ability of MR to image liagment injury non195
M. SILBERSTEIN et a1
FIGURE 4A - Spinal cord compression by haematoma. Axial CT with intrathecal contrast demonstrates anterior epidural lesion.
invasively has made it the first risk free technique able to do this (8). Our technique for identification of this parameter on CT was relatively crude, involving inference of subluxation on the basis of subjective assessment of degree of displacement, and although we failed to identify 73% of ligament injuries, we had 100% specificity with our technique. Clearly, however, MR is the method of choice and may make late functional views redundant. We failed to identify any of the patients with disc herniation, extramedullary haematoma, or cord compression with unenhanced CT. This is similar to the results of Tracy et a1 (7) but differs from the experience of Flanders et a1 (93) who identified 44% of disc protrusions and 77% of patients with cord compression using CT. We did not routinely photograph CT images of the spine on narrow soft tissue windows, and the analyses in this study involved retrospective review of previously photographed images, which is the likely reason for our poor rate of detection. Nevertheless, MR is likely to be the method of choice for identification of these findings (3). In the patients who had CT with intrathecal contrast, there was no difficulty in identifying extramedullary haematoma, cord compression, swelling or transection, although 1 of the 2 patients with disc herniation who 196
FIGURE 4B - Axial T1 weighted MR image yields similar findings. Acute epidural haematoma was suspected clinically.
had this study was not identified. The non-invasive nature of M R , however, will almost certainly make installation of intrathecal contrast redundant in the management of acute spinal trauma. In addition to the parameters assessed in this study, comparing CT and MR, specific identification of spinal cord lesions such as haemorrhage and oedema is possible with MR, a feature not offered by any other non-invasive imaging modality (1,2). Comparison of our results with previous studies at 1.5T, comparing CT and MR in acute spinal trauma suggests that M R at 0.3T yields a similar accuracy to previously reported series. Therefore, for the parameters assessed by this study, MR at 0.3T is a valid method of imaging in acute spinal trauma. One area in which low field MRI may be less than optimal is in the demonstration of the time-related changes in haemorrhage. Spin-echo sequences, as used in this study, frequently fail to demonstrate classical T2 shortening induced by intracellular deoxyhaemoglobin, which may only be visible using gradient recalled sequences (17). Evaluation of acute spinal trauma at lower field strengths should, therefore, include images obtained with gradient-echo techniques, as well as standard spin-echo sequences.
Rapid advances in high field strength MR are occurring, and with the increasing use of volume acquisitions allowing contiguous slice reconstructions, fat saturation techniques, and gradient echo techniques, resulting in high contrast between CSF and surrounding tissues, the ability of h4R to image subtle fractures and neural arch fractures is likely to improve to a stage where CT will no longer be required in the acute trauma setting (18, 19). CONCLUSION MR at 0.3T is highly accurate in the imaging of vertebral body fractures and spondylosis in the acute trauma setting. For ligament injury, prevertebral swelling, disc herniation and cord compression, it is the method of choice, out-performing CT. However, CT will still be required for demonstration of posterior element fractures. Although CT and MR were equally accurate in the demonstration of subluxations in our study, the results of previous studies at 1.5T suggest that an element of caution be maintained when excluding subtle subluxations using CT alone, and the theoretically poorer resolution of 0.3T MR should be kept in mind when attempting to make this diagnosis at lower field strengths.
Australasian Radiology, Val. 36, NO. 3 , August, 1992
M.R.I. AND C.T. IN ACUTE SPINAL TRAUMA
REFERENCES
FIGURE 5A - Spinal cord transection at T3. Axial CT images with intrathecal contrast demonstrate contrast within the cord superiorly, with cord discontinuity below.
1. Kulkarni MV, Bondurant FJ, Rose SL, Narayana PA. 1.5 Tesla magnetic resonance imaging of acute spinal trauma. Radiographics 1988; 8: 1059-1082. 2. Mirvis SE, Geisler FH, Jelinek JJ, Joslyn JN, Gellad F. Acute cervical spine trauma: evaluation with 1.5T MR imaging. Radiology 1988; 166: 807-816. 3. Flanders AE, Schaefer DM, Huynh TD, Mishkin MM, Gonzalez CF, Northrup BE. Acute cervical spine trauma: correlation of MR imaging findings with degree of neurologic deficit. Radiology 1990; 177: 25-33. 4. Hyman RA, Gorey MT. Imaging strategies for MR of the brain. Rad Clin North Am 1988; 26: 471-503. 5 . Hyman RA, Gorey MT. Imaging strategies for MR of the spine. Rad Clin North Am 1988; 26: 505-533. 6. Tarr RW, Drolshagen LF, Kerner TC, Allen JH, Partain CL, James AE Jr. MR imaging of recent spinal trauma. J Comput Assist Tomog 1987; 11: 4 12-417. 7. Tracy PT, Wright RM, Hanigan WC. Magnetic resonance imaging of spinal injury. Spine 1989; 14: 292-301. 8. Goldberg AL, Rothfus WE, Deeb ZL et al. The impact of magnetic resonance on the diagnostic evaluation of acute cervicothoracic spinal trauma. Skelet Radio1 1988; 17: 89-95. 9. Cooper PR, Cohen W. Evaluation of cervical spinal cord injuries with metrizamide mydography - CT scanning. J Neurosurg 1984; 61: 281-288. 10. Han JS, Benson JE, Yoon YS. Magnetic resonance imaging in the spinal column and craniovertebral junction. Rad Clin North Am 1984; 22: 805-827. 11. McArdle CB, Wright JW, Provest WS, Dornfest DJ, Amparo EG. MR imaging of the acutely injured patient with cervical traction. Radiology 1986; 159: 273-274. 12. Deutsch AL, Mink JH. Magnetic resonance imaging of musculoskeletal injuries. Rad Clin North Am 1989; 27: 983-1001. 13. New PFJ. Shoukimas GM. Thoracic s h e and spinal cord. In: Stark DD, Bradley WG, (eds). Magnetic resonance imaging. St Louis; CV Mosby: 1988: 662. 14. Jacobs RR, Casey MP. Surgical management of thoraco-lumbar spinal injuries: general principles and controversial considerations. Clin Orthop Re1 Res 1984; 189: 22-35. 15. Harms SE, Greenway G. The musculoskeletal system. In: Stark DD, Bradley WG (eds.). Magnetic resonance imaging. St Louis, CV Mosby; 1988: 1323. 16. Murphy MD, Batnitzky S, Bramble JM. Diagnostic imaging of spinal trauma. Rad Clin North Am 1989; 27: 855-871. 17. Zimmerman RD, Heier LA, Snow RB, Liu DPC, Kelly AB, Deck MDF. Acute intracranial haemorrhage: intensity changes on sequential MR scans at 0.5T. AJNR 1988; 9: 47-47. 18. Burke DL Jr, Dalinka MK, Schiebler ML, Cohen EK, Kressel HY. Strategies for musculoskeletal magnetic resonance imaging. Rad Clin North Am 1988; 26: 653-672. 19. Kricun R, Kricun ME. Advances in spinal imaging. Rad Clin North Am 1990; 28: 321-339.
FIGURE 5B -Axial TI weighted MR image confirms CT finding.
Australasian Radiology, Vol. 36, No. 3 , August, 1992
197
