Magnetic resonance imaging in the evaluation of cervical foraminal stenosis: comparison of 3D T2 SPACE with sagittal oblique 2D T2 TSE

The local ethics committee approved this retrospective study. Sixty consecutive patients who underwent MRI of the cervical spine at our institution were enrolled with the following inclusion criteria: (i) MRI of the cervical spine at our institution including sagittal oblique 2D T2-weighted and sagittal 3D T2-weighted sequences; (ii) referral by the institution’s orthopedic surgeons for cervical pain, cervical radiculopathy, or degenerative disease of the spine; and (iii) age > 18 years.

Examinations were performed on a 1.5 Tesla or a 3 Tesla-MR scanner (Avanto-fit and Skyra-fit, respectively, Siemens Healthineers) with a 20-channel head/neck coil. The protocol included sagittal T2w and T1w TSE sequences, an axial T2w TSE sequence, sagittal oblique T2w TSE sequences angled and centered on the neuroforamina of both sides (“sagittal oblique T2w”), and a sagittal 3D T2w sequence (SPACE, Sampling Perfection with Application optimized Contrasts using different flip angle Evolutions). Details of the sequences are listed in Table 1.

Differences between 2D T2w TSE and 3D T2w SPACE sequences concerning non-uniformity values and the results of the qualitative analysis (artifacts, structure visibility, foraminal stenosis) were tested for statistical significance by use of a Wilcoxon matched-pairs signed-ranks test (using the mean values of the two readers for the qualitative evaluation). A p-value of < 0.05 was considered statistically significant. Interreader agreement was assessed by means of a weighted κ correlation coefficient, interpreted as poor (κ < 0), slight (κ 0–0.2), fair (κ 0.21–0.4), moderate (κ 0.41–0.6), substantial (κ 0.61–0.8), or almost perfect (κ > 0.8) [21].

Statistical analyses were performed using SPSS version 24.0 (IBM Corp.) and MedCalc version 17.6 (MedCalc Software bvba), the latter for calculation of the weighted kappa.

Non-uniformity values of 2D T2w TSE and 3D T2w SPACE sequences at 1.5 T and 3 T are presented in Table 2: Significant differences between the sequences were found at 1.5 T, the non-uniformity of CSF being lower for 3D T2w SPACE (4.77 (2.32 SD) vs. 5.72 (2.39 SD), p = 0.022) and higher for bone (13.22 (3.63 SD) vs. 11.25 (2.74 SD), p < 0.01) and muscle (34.2 (9.15 SD) vs. 23.75 (4.38 SD), p < 0.01). Differences between 3D T2w SPACE and 2D T2w TSE were not significant at 3 T except for muscle, with a higher non-uniformity for the 3D T2w SPACE sequence (31.9 (14.57 SD) vs. 21.11 (5.31 SD), p < 0.01).

Artifacts of the investigated sequences were rated as non-existent to mild (Table 3), without a significant difference between 3D T2w SPACE and 2D T2w TSE when considering all cases (average score of 1.5 for each, p = 0.745), but with higher scores for the 2D T2w TSE sequence at 3 T (1.7 vs. 1.4 p = 0.023) and for the 3D T2w SPACE sequence at 1.5 T (1.6 vs. 1.5, p = 0.030).

Visibility scores of the evaluated anatomical structures on 3D T2w SPACE and 2D T2w TSE are presented in Table 4. Visibility of most structures was rated as good to excellent on both sequences, with the exception of the foraminal vessel and the nerve root, whose visibility was rated as adequate on 3D T2w SPACE at 1.5 T and 3 T for the former and at 3 T for the latter (with good visibility at 1.5 T). Significant differences between the two sequences were found for neuroforaminal fat (mean score of 4.6 vs. 4.1), nerve root (4.3 vs. 3.7), and foraminal vessel (4.1 vs. 3.1), with a better visibility on 2D T2w TSE, as well as facet joint (4.6 vs. 4.7), with a better visibility on 3D T2w SPACE. The highest difference in mean visibility scores between sequences was 1 point for the foraminal vessel (3.1 on 3D T2w SPACE vs. 4.1 on 2D T2w TSE, p < 0.01).

A little over half of the evaluated segments showed a stenosis: 53% of foramina based on the 2D T2W TSE sequence and 56% based on the 3D T2w SPACE sequence, with a minority of severe stenoses (Table 5).

Results of foraminal stenosis grading are shown in Table 6. Mean stenosis grades were low, ranging from 1.1 to 2.6. Significant differences between gradings based on 3D T2w SPACE and 2D T2w TSE were found at five locations: C3/C4 and C4/C5 on both sides and C5/C6 on the left, with higher grades on 3D T2w SPACE (e.g., C3/C4 on the left with scores of 1.7 vs. 2.0, p < 0.01, mean scores of the two readers).

Interreader agreement was moderate to substantial for both 3D T2w SPACE and 2D T2w TSE (Table 7), except for one level with fair agreement on 3D T2w SPACE (C7/T1 right), without consistently higher agreement for any of the sequences.

Quantitative analysis revealed differences between the sequences mainly at 1.5 T, in favor of 3D T2w SPACE concerning CSF non-uniformity values and in favor of T2w TSE concerning bone and muscle non-uniformity values. Non-uniformity values match those found in previous studies [13, 15], though the differences between 2 and 3D T2w sequences were not statistically different in a study at 1.5 T [13], while they were in a 3 T cohort [15], in favor of T2w TSE for CSF and muscle. Discrepancy of these findings may be due to technical reasons (different machines and sequence parameters between studies) and differing composition of the study cohorts, with younger volunteers in the mentioned studies and a patient cohort including older subjects in the present study. Overall, the differences in image heterogeneity appear acceptable for clinical use of the sequences.

Neither 3D T2w SPACE nor 2D T2w TSE was limited by artifacts, and there was no difference between the sequences concerning artifact grading. This is in concordance with the previous results [13]. In contrast, other studies that analyzed different artifact types showed significantly less CSF flow artifact on 3D T2w sequences [14, 15]. As the current study focused on analysis of the neural foramina, we did not examine artifacts mainly concerning the intraspinal compartment such as CSF flow.

Anatomical structures were well visible on both sequences, with slightly better visibility scores for neuroforaminal contents on 2D T2w TSE, especially for foraminal vessels, whereas previous studies showed better delineation of foraminal contents on 3D T2w sequences [13, 14] or no statistical significance between sequences [15]. This may be due to differences in technique including utilization of strictly sagittal planes in those studies, as opposed to the sagittal oblique planes dedicated to foraminal analysis in the present study, allowing optimized foraminal visualization on 2D T2w TSE and potentially blurring the advantages of the 3D sequence (including reduced slice thickness with the potential of better visualization of the foramina even in a strictly sagittal plane). However, the differences between sequences were limited, visibility of foraminal contents and other structures being overall graded as “good” on both sequences with the exception of the foraminal vessel (the latter not representing the main structure of interest in foraminal imaging), and consequently do not appear to be significant for daily clinical practice.

Prior studies on the use of 3D T2w SPACE in the cervical spine mostly assessed the visibility of anatomical structures [13‐15], mainly in smaller cohorts of healthy volunteers [13, 15] and in comparison to standard 2D T2w TSE. The present study also analyzed its use for a clinical indication, namely foraminal stenosis grading in a patient population, comparing it to a currently used dedicated sequence for foraminal analysis, sagittal oblique T2w TSE.

Stenosis grades were not significantly different between sequences at most levels, while higher grades were attributed at certain mid-cervical levels based on 3D T2w SPACE at 1.5 T. Underestimation and overestimation of stenosis on the 2D T2w TSE sequence seems possible due to its inherently greater slice thickness and spacing, with the potential of volume averaging. The single plane coverage of all neuroforamina of one side, while they may have different orientations, may contribute to difficulties in evaluation. It remains unclear which sequence more closely reflected true foraminal stenosis, as there was no reference standard for the evaluation. Prior studies on the grading of cervical foraminal stenosis focused mainly on the comparison of sagittal oblique 2D T2w sequences [7, 16] or 3D T2w sequences [22, 23] with conventional, i.e., sagittal and axial planes.

Stenosis prevalence and mean stenosis grades in the present study were low, ranging from 1.1 to 2.6. Additional investigations in cohorts with more prevalent significant foraminal stenosis as well as correlation with symptoms and surgical findings could clarify the respective sensitivity and specificity of 3D T2w and 2D T2w sequences.

Compared to the acquisition time of bilateral sagittal oblique 2D T2w TSE sequences, acquisition time of the 3D T2w SPACE sequence, which was accelerated by parallel imaging (iPAT of 2–3), was about 30% longer at 3 T (04:31 min vs. 03:06 min), and slightly shorter at 1.5 T (04:28 min vs. 04:54 min), respectively. Within these acquisition times, 3D T2w SPACE also offers increased anatomical coverage in comparison to sagittal oblique 2D T2w TSE sequences, including the neuroforamina of both sides as well as the entire spinal canal, thereby offering an additional possibility to analyze structures in the spinal canal (particularly the nerve rootlets, better delineated than on 2D T2w according to prior studies [14]) and the potential to follow rootlets and proximal nerve roots on one single sequence. The possibility to reformat the sequence in arbitrary planes is of potential additional benefit, as the foramina of different levels may not lie in the same plane, particularly in patients with scoliosis or vertebral segmentation anomalies (Fig. 5). In this context, the longer time necessary for grading the foraminal stenosis compared to 2D T2w TSE—less than 1 min, including the time needed to reformat the images—appears acceptable for clinical practice.

In light of the relatively low additional examination and interpretation time and its added benefits, 3D T2w SPACE could supplant sagittal oblique 2D T2w TSE sequences. Replacement of conventional sagittal and axial T2w TSE and/or spoiled T2*w sequences (such as multi-echo data image combination, MEDIC) seems less practicable unless for indications where the spinal cord itself is not of interest, as spinal cord anatomy and pathology are superiorly depicted on the latter sequence [24‐28].

This study is limited by its retrospective design and the absence of a surgical reference standard. Further, the prevalence of significant foraminal stenosis of the cervical spine was low; however, this reflects the prevalence in a typical clinical setting.

3D T2w SPACE is comparable with sagittal oblique 2D T2w TSE in the evaluation of cervical neural foramina and can be acquired and analyzed in an acceptable time, furthering its potential to replace sagittal oblique sequences.