Automated classification of intervertebral disc degeneration using Pyradiomics features: XGBoost versus OnevsRest

Motivation: Disc degeneration is the leading cause of low back pain. However, there are no objective measures of the disc degeneration. Currently disc degeneration is graded by visual assessment of MRI, which often leads to uncertainty and disagreements. Goal(s): Our goal was to develop an automated, efficient, accurate and objective diagnostic tool to assess disc degeneration. Approach: Binary disc masks are generated using nnU-Net. Radiomics features are extracted from T2w MRI and XGBoost and OnevsRest classification methods were tested. Results: XGBoost was in good agreement with the reader and the important features used in classification were in accord with expected changes in discs. Impact: Lack of objective measures of disc degeneration may cause uncertainties in treatment decisions. Automated evaluation of disc degeneration streamlines the physician's workflow and reduce uncertainties. Using radiomics features enables explainability and provides simple and robust training for machine learning approaches.