Search papers, labs, and topics across Lattice.
This study introduces SpineRankNet, a novel framework that models lumbar spine degeneration as a continuous severity ranking problem rather than a multi-class classification task. By employing a ranking loss to derive scalar severity scores from MRI scans, the method allows for a more nuanced assessment of degeneration, capturing subtle differences in severity that traditional ordinal grading systems overlook. The results demonstrate that the continuous scoring approach not only maintains classification accuracy but also enhances discrimination between more distant degeneration classes.
Continuous severity scoring outperforms traditional multi-class classification in assessing lumbar spine degeneration, revealing finer distinctions in MRI evaluations.
Lumbar spine degeneration is a major contributor to chronic low back pain and is routinely assessed on MRI using ordinal grading systems, e.g. normal, mild, moderate, severe. Consequently, most approaches to train models to grade these MRIs formulate grading as a multi-class classification problem, treating ordinal grades as categorical, ignoring differences in misclassification severity, and imposing hard decision boundaries on a continuous disease process. This work explores modeling spinal degeneration as a continuous severity ranking problem. We introduce SpineRankNet, a framework that learns scalar severity scores from lumbar spinal MRI, and compare it against multi-class classification and ordinal regression. Using multiple degeneration measures from the Genodisc dataset, we show that a model trained using a ranking loss to produce a continuous score enables fine-grained ordering of MRI scans. Furthermore, the ordinal grading classes can be recovered from the score with comparable accuracy to those from a model trained directly for classification. The score learned by ranking even improves discrimination between more distant classes. Source code is available at https://github.com/spinetools/spineranknet.
ETH