Date of Award

January 2014

Document Type


Degree Name

Medical Doctor (MD)



First Advisor

Jonathan N. Grauer

Subject Area(s)



Magnetic resonance imaging (MRI) is frequently used in the evaluation of degenerative conditions of the cervical spine, with two-dimensional fast spin-echo (2D-FSE) sequences being the accepted standard. Observer variability is an important limitation of MRI, and the relative inter- and intra-rater agreements of MRI findings across different pathologic conditions are underexplored, as most studies are focused on specific findings. A further limitation is that with respect to axial imaging, visualization of pathology may be suboptimal if the image plane is oriented differently than the plane of interest, due to lordosis, kyphosis, or deformity. Three-dimensional turbo spin-echo (3D-TSE) isotropic MRI is a promising novel technology that bypasses this limitation by enabling dynamic image reformation in any desired orientation.

The purpose of this study about the diagnostic capability of MRI in assessing the cervical spine is two-fold. First, to characterize the inter- and intra-rater agreement of 2D-FSE in evaluating seven pathologic cervical spine conditions. Second, to determine the diagnostic performance of 3D-TSE MRI in the assessment of cervical spine pathology relative to the standard 2D-FSE sequences.

Forty-eight patients who underwent 3D-TSE and conventional 2D-FSE T2-weight cervical MRI at our institution were randomly selected. 3D-TSE and 2D-FSE sequences from each subject were independently reviewed by two orthopaedic spine surgeons and four musculoskeletal radiologists. The first ten studies were re-evaluated to determine intra-rater agreement. Images were assessed using specific criteria for seven pathologic findings related to disc degeneration, stenosis, and other degenerative conditions. Inter- method, inter-rater, and intra-rater percent agreements were calculated. Fleiss' kappa coefficients for inter-rater agreement were determined.

In terms of 2D-FSE MRI diagnostic characteristics across different pathologic conditions, the overall inter-rater agreement was 75.7%. The overall intra-rater agreement was 81.6%. Average Fleiss' kappa coefficient was 0.432, suggesting moderate overall agreement. When stratified by condition, however, inter-rater agreement ranged from 54.6% to 95.0%. Disc hydration, central stenosis, and foraminal stenosis had the lowest inter-rater agreements. Intra-rater agreement had a narrower range, from 74.2% to 94.7%.

With respect to the comparison between 3D-TSE and 2D-FSE MRI, the overall inter- method agreement was 80.7%. The inter-rater agreement for 3D-TSE was 75.9%, which was not statistically different from that of 2D-FSE MRI (p = 0.47). The intra-rater agreement was 82.2% for 3D-TSE, which was again similar to that of 2D-FSE (p = 0.71). Fleiss' kappa for 3D-TSE was 0.420, indicating moderate inter-rater reliability, and was

also not statistically different from 2D-FSE (p = 0.62). Lastly, the inter-method agreement and the 2D-FSE intra-rater agreement were statistically similar (p = 0.49).

In conclusion, there was significant variability in the inter- and intra-rater agreement of MRI in assessing different degenerative conditions of the cervical spine, and clinicians should be aware of the condition-specific diagnostic limitations of MRI interpretation. Furthermore, as a promising novel modality in MRI, 3D-TSE exhibited a high degree of agreement with conventional 2D-FSE in assessing the cervical spine. This study demonstrates that 3D-TSE yields at least equivalent diagnostic information as 2D-FSE, with the added benefit of allowing for real-time dynamic image reconstruction.


This thesis is restricted to Yale network users only. This thesis is permanently embargoed from public release.