British Journal of Radiology (2004) 77, 189-196
© 2004 British Institute of Radiology
doi: 10.1259/bjr/97837637
Are T1 weighted images helpful in MRI of cervical radiculopathy?
A G Ryan, FRCR1,
B M Morrissey, FRCR1,
R G Newcombe, PhD2,
S F S Halpin, FRCR1 and
M D Hourihan, FRCR1
Departments of 1 Neuroradiology and 2 Epidemiology, Statistics and Public Health, University Hospital of Wales, Heath Park, Cardiff 1 CF14 4XW and 2 CF14 4XN, UK
 |
Abstract
|
|---|
MRI in patients with cervical myelopathy or radiculopathy usually includes T1 weighted (T1W) and T2 weighted (T2W) images. We prospectively examined a hypothesis that T2W alone is sufficient to diagnose the cause of cervical myelopathy and radiculopathy and that the T1W sagittal images do not provide additional useful information. 30 patients presenting with a history of cervical radiculopathy with or without myelopathy were prospectively assessed by MRI. Those with a history suggestive of intrinsic primary cord disease or who had previously had surgery were excluded. Two neuroradiologists, blinded to the clinical information, separately viewed the sagittal and axial T2W images alone, and at a later time, the full set of T1W and T2W images. Image quality, location and severity of disease and confidence of diagnosis at each level were scored on 4- or 5-point scales. The T1 sequences did not demonstrate any significant lesions not already seen on the T2W images alone. The T1W sequence may safely be omitted in patients with radiculopathy.
|
Introduction
|
|---|
Historically, a number of imaging options have been available to assess cervical myelopathy and radiculopathy, and CT myelography (CTM) has been considered to be the standard examination [1]. Rapid refinement of MRI technology has enabled this non-invasive procedure to assume a primary role in the assessment of spinal radicular symptoms. Both T1 weighted (T1W) and T2 weighted (T2W) sagittal images are usually used to assess the cervical spine and the contents of the spinal canal. T1W images have traditionally been used to demonstrate anatomy, with the added advantage of showing marrow fat as bright signal, while T2W images give a myelographic effect, demonstrating intraspinal nerve roots, and allowing assessment of the internal structure of the spinal cord: these latter features are not reliably seen on T1W scans. Image quality has improved to such an extent in the newer high field scanners that we can now ask whether the T2W images alone provide as much diagnostic information as the T1W and T2W images combined.
|
Methods and materials
|
|---|
A prospective study of 30 patients referred for MRI of the cervical spine because of radiculopathy with or without myelopathic symptoms was conducted.
Patients
The patients were 18 men (ranging from 31 years to 80 years, mean=47.4 years) and 12 women (ranging from 25 years to 73 years, mean=45.1 years). Patients who had undergone spinal surgery, those with predominantly intrinsic cord symptoms and those thought to have a systemic illness associated with the radiculopathy, e.g. rheumatoid arthritis or cancer, were excluded.
All images were obtained on a 1.5 T Signa (GE Medical Systems, Milwaukee, WI) with a phased array coil. Each patient had a T1W sagittal spin echo (typical values: repetition time (TR)=600, echo time (TE)=12, 2 excitations, matrix=256 x 256, field of view (FOV)=28 cm, slice thickness=3 mm, time=4.30 min, with saturation pulses), T2W sagittal fast spin echo (TR=4500, TE=90 [ef], matrix=256 x 256, 4 excitations, FOV=28–32 cm, slice thickness=3 mm, time=4.20 min, with saturation pulses), T2W* axial gradient recalled echo (GRE), (TR=600, TE=17, flip angle=20 degrees, matrix=256 x 192, 4 excitations, FOV=20 cm, slice=3 mm, time=7.40 min with flow compensation, and saturation pulse). 3D volume gradient recalled acquisition in the steady state (GRASS) blocks were also obtained when further information on the neural foramina was required (TR=33, TE=17, flip angle=5 degrees, matrix=256 x 160, 1.5 excitations, FOV=20 cm, slice thickness=1.2 mm, time=8.20 min, flow compensation and saturation pulses).
Two neuroradiologists (SH, MDH), blinded to the clinical information, independently reviewed the images in random order on two separate occasions. Initially the T2W axial and sagittal images were assessed alone, and at a different sitting, at least a week later, the full set of T1W and T2W images was reviewed. The images were viewed on cut film, without access to a diagnostic workstation.
The parameters assessed, derived from the work of Ross et al [2], were:
- overall quality;
- location of disease;
- severity of disease;
- whether impingement was due to disc, osteophyte or a combination of both;
- overall confidence of diagnosis.
Allowance was made for comments if necessary, such as the influence of the T1W sequence on the overall diagnosis.
The quality of the images was graded on a 5-point scale, 0 being uninterpretable, 1–4 indicating poor, adequate, good and excellent quality, respectively. Location of disease was assessed at each of the 5 levels from C3/4 to C7/T1. At each level, the presence of a disc and/or osteophyte was noted, as was its position (central, right lateral, left lateral). Severity of disease was graded on a 4 point scale, 0 being normal and grades 1–3 indicating mild, moderate and severe neural impingement (1
33%, 2=34–66%, 3
66% canal or neural foraminal encroachment). A minor disc bulge would be graded 1 (Figure 1
), while a large disc prolapse would be graded 3 (Figure 2). Confidence of diagnosis was scored on a 5-point scale, 0 = not at all confident; 1=a little confident; 2=moderate confidence; 3=a high degree of confidence; 4=certain confidence (Figure 3).
View larger version (181K):
[in this window]
[in a new window]
|
Figure 2. T2 weighted sagittal image demonstrating Grade 3 disc bulge at C4/5 and Grade 1 disc bulge at C7/T1.
|
|
View larger version (129K):
[in this window]
[in a new window]
|
Figure 3. This T2 star weighted (T2W*) axial gradient recalled echo image is poor, and was scored Grade 2 in quality but there is clearly a right lateral disc protrusion which is of grade 2 severity. The examination was scored 4 (certain) in confidence of diagnosis.
|
|
Statistical analysis
Analysis was carried out on data supplied via an ExcelTM spreadsheet and transferred to both SPSS (Statistical Package for the Social Sciences, Chicago, IL) version 9 for windows and an ASCII mainframe. Sensitivity and specificity could not be assessed as no gold standard was available within the study, and thus receiver–operator curves could not be constructed. The statistical analysis was directed towards assessing the tendency for one method to over or under-read the other systematically. In particular, the question was asked: does the T2 set undercall the combined study? As the number of patients is small, it is more appropriate to seek to estimate the differences on moving from one set to the other, rather than relying on simple tests of statistical significance [3]. The differences between the image sets were assessed initially on a site by site basis at each level for each possible lesion. The same analysis was repeated after combining the data for all discs, all osteophytes and all lesions combined.
After binary collapse of the four-point rated severity scale, (using two alternative divisions within the data; primarily regarding 0 and 1 as negative and 2 and 3 as positive as the initial division, and secondarily assigning negativity to 0 and positivity to 1, 2 and 3), the data were analysed to estimate the degree of difference (with respect to severity) between the image sets using a method for comparing the difference in the proportion of positive readings between methods, with 95% confidence intervals calculated by method 10 of Newcombe [4]. This same method was used to assess interobserver differences. The 
statistic was calculated for both intermethod and interobserver agreement with 95% confidence limits calculated by the method of Donner and Eliasziw [5]. The differences in quality and confidence between methods for each observer were assessed by applying the paired t-test.
|
Results
|
|---|
Image quality
The majority of images were graded 2–3 on the 5-point quality scale on each set by both observers (Table 1). Each observer gave a significantly higher rating on the full set compared with the T2 images alone.
Lesion severity
Of 900 possible lesions (30 patients, five disc levels, three possible sites: left lateral, central and right lateral, two possible diagnoses: disc and/or osteophyte), observer 1 identified 197 abnormalities on the T2W sequences alone, with 162 identified when the full set was reviewed (Table 2). Observer 2 initially detected 132 abnormalities with 112 identified when the T1W sagittal sequence was added. In grading the severity of disease both observers scored most at grade 1 (Table 2) with the addition of the T1W images resulting in downgrading of a proportion of scores.
Intermethod comparison
The T2W images identified a higher degree of abnormality than the full set on 81 occasions (Observer 1, Tables 3 and 3a) and 53 occasions (Observer 2, Tables 4 and 4a). On 12 occasions (for each observer) there was a significant change in diagnosis, i.e. a change in score of two points or more, moving from the T2 set to the full set. Seven times (Observer 1) and nine times (Observer 2) there was a 2-point difference, while on one occasion (Observer 1, Figure 4) a 3-point difference was noted. Even with retrospective manipulation of the window settings the impingement could not be demonstrated on the T1W images.
View larger version (139K):
[in this window]
[in a new window]
|
Figure 4. (a) The T1 weighted sagittal image fails to show a large osteophyte bar, which is clearly demonstrated on (b) the T2 weighted sagittal image.
|
|
Conversely, the full set outperformed the T2W images on 40 occasions (Observer 1) and 26 occasions (Observer 2), but in each of these there was only a 1-point change (from 0 to 1), none of which was considered to be of clinical significance. No significant lesion was demonstrated on the T1W images not seen on the T2W images alone.
On a site by site basis there was no significant difference between methods, however, the effect of aggregating the data for discs and osteophytes at the five levels was to produce a significant difference in a number of instances. There was no significant difference between methods for Observer 1 with regard to discs or osteophytes alone but when the two were combined the difference became significant. Similarly, for Observer 2, there was no significant difference with respect to osteophytes on their own, whereas such a difference did exist in relation to discs considered on their own and the combined total. The kappa statistics (Tables 3a and 4a
) demonstrate good agreement between the methods. Shifting the division within the binary collapse i.e. when 0=negative and 1, 2 and 3 are positive, had little effect on these results.
Interobserver agreement
Tables 5 and 6 demonstrate the degree of interobserver agreement (with statistical analysis tabulated separately in Tables 5a and 6a). Agreement occurred on 732 occasions on T2 alone and on 761 occasions on the full set, out of 900 possible outcomes (450 locations, two possible lesions). Most (87.5%, T2 alone; 91.36%, T1+T2) of the disagreements were of one point only, the majority of which (87.07%, T2 alone; 88.18%, T1+T2) were at the grade 1 vs 0 level, where either a small disc or osteophyte was accorded a grade 1 score by one observer but disregarded as insignificant by the other. Observer 2 tended to call higher than Observer 1, except at the 0 vs 1 where this trend was reversed. Moving from T2 to T1+T2 made no difference to these relationships. Shifting the division within the binary collapse as above alters the relationship slightly with Observer 1 calling higher with respect to Observer 2, reflecting the former's tendency to call small indentations as positive (grade 1) which the second observer tended to ignore. The chance-corrected statistics suggest fair to good agreement between the observers.
Confidence
Both observers had a high or certain degree of confidence in diagnosis in the majority of cases (Table 7) with a statistically significant higher rating on the full set. For Observer 1, on 90 occasions the full set allowed a one-point increase and on one occasion, a two-point increase in the degree of confidence. A drop in confidence by one point occurred on 301 occasions. For Observer 2 there was an increase in confidence by one point on 180 occasions when the full set was viewed. A drop in confidence by one point occurred on 91 occasions.
|
Discussion
|
|---|
Imaging options in the radiological assessment of cervical radiculopathy have included plain radiography, myelography, contrast-enhanced CT (CECT), CTM, radioisotope studies, discography and MRI [6]. CTM has been considered the gold standard [1], but more recently, MRI has assumed a primary role in both pre-operative screening and post-operative assessment and several studies have demonstrated accuracy equal to or better than myelography or CTM [1, 7, 8]. Larsson et al [8] concluded that MRI can replace these more invasive procedures in the pre-operative assessment of radicular symptoms due to either disc or osteophyte.
Traditionally, both T1W and T2W sagittal images have been required to assess the spinal canal and its contents fully [6, 9, 10]. The T1W sagittal image provided better anatomical definition, showing soft disc herniation well, and marrow infiltration, if present. There was poor soft tissue differentiation, however. The T2W images were obtained primarily for the myelographic effect due to bright cerebrospinal fluid (CSF), but also to demonstrate myelomalacia within the cord and vertebral osteomyelitis [6]. On first and second generation machines, poor cord/CSF differentiation was a problem and the images generally suffered from a long acquisition time, CSF artefacts, dark and bright artefacts and inherent low signal-to-noise and contrast-to-noise ratios [11].
Once MRI was accepted as an important imaging modality in assessing cervical radiculopathy [6], much work went into developing the most efficient sequences to enable the exact site and extent of disc or osteophytic impingement to be assessed with maximum accuracy and confidence. From the initial classic spin-echo and fast spin-echo sequences, Enzmann and Rubin [11] assessed GRASS (with short TR, short TE and small flip angle), Czervionke et al [12] compared T2W gradient echo with gadolinium diethylene triamine pentacetic acid (Gd-DTPA) enhancement and Ross and his co-workers [2, 9, 13] compared 2D and 3D techniques such as low flip angle fast imaging with steady state free precession (FISP) and Gd-DTPA-enhanced "turboflash" sequences. This latter technique provides detailed imaging of the cervical spinal canal in a little over 5 min and the authors have suggested it may be the single best study for assessing foraminal disease, but provides less information on internal cord structure or bone marrow. Similarly, Georgy et al [10] have proposed that a single contrast-enhanced 3D sequence might be sufficient for screening for degenerative disc disease [14]. Jones has suggested that T2W FSE images alone may be adequate for routine spine imaging [15], however this has raised some concerns in the literature, in particular with regard to the perceived danger of missing intrinsic cord lesions. Miaux [16] reported a case in which an intramedullary lesion seen on proton density weighted CSE sagittal images was not detected on FSE T2W images. This patient presented with symptoms suggestive of a cord lesion and so would not have been deemed a suitable candidate for a single weighting study in our cohort.
Our study demonstrated that on our 1.5 T images, standard T2W sagittal and T2W* axial sequences are of sufficient quality for full assessment of the pre-operative patient, with only occasional recourse to a 3D GRASS block for better definition of the neural foramina. The reasons for the improved quality of the T2W sagittal images (Figure 5) include use of a high field magnet, a larger matrix, up to four excitations and also the introduction of new gradient echo sequences, 3D acquisitions, saturation pulses, flow compensation and pulse gating [9, 13, 14]. It is this considerable improvement in quality coupled with the intrinsic myelographic effect of the white CSF on T2W images that enables the radiologist to obtain all the necessary information from these sequences alone [17]. Ross has proposed that the ideal of spinal MR is the use of a "very few robust sequences for complete evaluation of the spinal axis" with "T2W sequences generally to be preferred" [18]. There is no doubt that T2W and T2W* images on the latest generation scanners are of such high quality that MRI can now compete with CTM as the imaging tool of choice in cervical myelopathy and radiculopathy [7, 17].
View larger version (80K):
[in this window]
[in a new window]
|
Figure 5. (a) T2 weighted sagittal image demonstrates the myelographic effect of the bright cerebrospinal fluid. Increased signal indicating the myelomalacia within the cord is also seen. (b) T2 star weighted (T2W*) axial image through neural foramen clearly shows internal structure of the spinal cord and the exiting nerve roots.
|
|
Our results suggest that the full set may undercall the T2W images (although in the absence of a gold standard, it is impossible to objectively verify or refute this). We would postulate that this may be due to a moderating influence of the T1W images on the T2W impression, resulting in the (incorrect) downgrading of a lesion perceived on T2, e.g. the osteophyte bar not seen on T1. Without a gold standard we cannot ascertain the false-positive rate of the T2W images in isolation. Even if it is the case that the T2W images falsely upgrade a demonstrated lesion, the surgeons at our institution will operate on the basis of the signs and symptoms in conjunction with the imaging findings (after double reading) i.e. rather than on the imaging alone, thus patients will not receive inappropriate treatment. If multiple lesions are demonstrated and a choice of operating level must be made, the T2W images will certainly aid in the determination of the most severe lesion. On the other hand, with the undoubted greater sensitivity of T2W images, we can be assured that significant lesions will not be missed.
Our interpretation of the results is that the T1W images have downgraded the reviewers interpretation of the T2W images inappropriately. However, an alternative interpretation might be that the T2W images inappropriately magnify the imaging findings, and that the T1W images tend to correct this supposed tendency of T2W scans to overcall abnormality. However, on the majority of occasions, the use of T1W images upgraded the severity of diagnosis, from normal to minor disease, and the reverse effect occurred only 12 times. Although in these 12 patients, the diagnosis was changed upwards two points in the severity scale by considering the T2W images alone—seven patients for Observer 1, and nine for Observer 2. The fact that these 2-point variations were not always in the same patient between observers suggests that one set of observations may have been made in error. Certainly Observer 1, who scored the lesion in Figure 4 as "severe" when looking at the T2W images alone and "normal" when looking at T1 and T2 weighted images together (a three point variation) cannot now understand how the "normal" diagnosis was reached.
It is a weakness of this paper that we did not assess intraobserver variation in this study, but we feel that most of the 12 two-point alterations in severity are likely due to erroneous interpretation on one or other occasion; such a discrepancy is to be expected among 900 observations, and it might be said that 12 two-point discrepancies is acceptably low intraobserver error.
We have looked at the notes of these 12 patients. In eight, the level in question was inappropriate for the patients symptoms and signs, or these settled spontaneously and the surgeon decided on conservative treatment; and in four the patient underwent surgery, of whom three experienced relief of symptoms post-operatively. These clinical data are unhelpful in deciding between the two competing interpretations of the results, as the numbers are too small. However in the light of our finding that no significant lesion seen on T1W scans was not seen on T2W images, and that by far the most common variation between the two image sets was a one-point variation in the severity scale, we do not think that that the T1W sagittal scans provide clinically useful information in these patients.
Although the differences between confidence levels are highly significant, they are in opposite directions. We believe the results reflect individual variation and that we cannot extrapolate this finding to radiologists in general. The interobserver differences in severity rating were for the most part due to the threshold at which tiny radiographic abnormalities merge with the changes of normal ageing. If alterations between grade 1 and grade 0 disease are ignored, excellent overall agreement was reached. The relatively large number of discordant observations was either of a minor degree (one grade difference) or occurred in images that were of poor quality. In practice, as all neurological examinations are double reported in this department, such differences do not arise.
In conclusion, it has been shown that while the T1W sagittal sequence may increase the perceived quality of a study, it does not contribute clinically significant useful information in addition to that provided by the T2W images alone.
Specific indications remain for use of the T1W sagittal sequence: if there is uncertainty after the T2W images have been obtained (this did not happen in our study); if cord haemorrhage or bone marrow abnormality is suspected: if gadolinium contrast enhancement is required [10], or perhaps in the search for vertebral metastases (we would prefer T2W fat suppression techniques here). We no longer routinely perform sagittal T1W images in the assessment of myelopathy or radiculopathy and believe that in the routine assessment of cervical radiculopathy the T1W sagittal sequence can be safely omitted.
Received for publication June 25, 2002.
Revision received July 6, 2003.
Accepted for publication November 19, 2003.
|
References
|
|---|
- Yousem DM, Atlas SW, Hackney DB. Cervical spine disc herniation: comparison of CT and 3DFT gradient echo MR scans. J Comput Assist Tomogr 1992;16:345–51.[Medline]
- Ross JS, Ruggieri PM, Glicklish M, et al. 3D MRI of the cervical spine: low flip angle FISP versus GdDTPA TurboFLASH in degenerative disc disease. J Comput Assist Tomogr 1993;17:26–33.[Medline]
- Gardner MJ, Altman DG. Confidence intervals rather than P values: estimation rather than hypothesis testing. Br Med J 1986;292:746–50.
- Newcombe RG. Improved confidence intervals for the difference between proportions based on paired data. Stat Med 1998;17:2635–50.[CrossRef][Medline]
- Donner A, Eliasziw M. A goodness-of-fit approach to inference procedures for the kappa statistic: confidence interval construction, significance testing and sample size estimation. Stat Med 1992;11:1511–9.[Medline]
- Modic MJT, Masaryk TJ, Ross JS, et al. Imaging of degenerative disc disease (1). Radiology 1988;168:177–86.[Free Full Text]
- Masaryk TJ, Modic MT, Geisinger MA, et al. Cervical myelopathy: a comparison of MR and myelography. J Comput Assist Tomogr 1986;10:184–94.[Medline]
- Larsson EM, Holtass S, Cronquest S, et al. Comparison of myelography, CT myelography and magnetic resonance imaging in cervical spondylosis and disc herniation: pre and postoperative findings. Acta Radiol 1989;30:233–9.[Medline]
- Ross JS. MR imaging of the cervical spine: techniques for two and three-dimensional imaging. AJR Am J Roentgenol 1992;159:779–86.[Abstract/Free Full Text]
- Georgy BA, Snow RD, Hesselink JR. MR imaging of spinal nerve roots: techniques, enhancement patterns and imaging findings. AJR Am J Roentgenol 1996;166:173–9.[Abstract/Free Full Text]
- Enzmann DR, Rubin JB. Cervical spine: MR imaging with a partial flip angle gradient-refocused pulse sequence. Part 1. General considerations and disc disease. Radiology 1988;166:467–72.[Abstract/Free Full Text]
- Czervionke LF, Daniels DL, Ho PS, et al. Cervical neural foramina: correlative anatomic and MR imaging study. Radiology 1998;169:753–9.
- Ross JS, Tkach J, VanDyke C, et al. Clinical MR imaging of degenerative spinal disease: pulse sequences, gradient echo techniques and contrast agents. J Magn Reson Imaging 1991;1:29–37.[Medline]
- Georgy BA, Hesselink JR. MR imaging of the spine: recent advances in pulse sequences and special techniques. AJR Am J Roentgenol 1994;162:923–34.[Abstract/Free Full Text]
- Jones KM, Mulkern RV, Schwartz RB, et al. Fast spin-echo MR imaging of the brain and spine: current concepts. AJR Am J Roentgenol 1992;158:1313–20.[Free Full Text]
- Miaux Y, Feldman L, Ducerveau MN, et al. Are sagittal T2 weighted fast spin-echo MR images sufficient for routine spine imaging? AJNR Am J Neuroradiol 1995;16:1558–60.[Medline]
- Yousem DM, Atlas SA, Goldberg HI, Grossman RI. Degenerative cervical spine foraminal narrowing; ultra-thin, high resolution 3DFT gradient-echo MR imaging. AJNR Am J Neuroradiol 1991;12:226–36.
- Ross JS. Newer sequences for spinal MR imaging: smorgasbord or succotash of acronyms? AJNR Am J Neuroradiol 1999;20:361–73.[Free Full Text]
Top
Abstract
Introduction
