Explainable vertebral fracture analysis with uncertainty estimation using differentiable rule-based classification
This addresses the need for reliable and interpretable fracture diagnosis in clinical settings, offering a tool that matches human annotator reliability.
The paper tackles vertebral fracture assessment in low-dose radiographs by introducing a method that provides explainable classifications and uncertainty estimates, achieving a vertebra-level sensitivity of 93% and an end-to-end AUC of 97%.
We present a novel method for explainable vertebral fracture assessment (XVFA) in low-dose radiographs using deep neural networks, incorporating vertebra detection and keypoint localization with uncertainty estimates. We incorporate Genant's semi-quantitative criteria as a differentiable rule-based means of classifying both vertebra fracture grade and morphology. Unlike previous work, XVFA provides explainable classifications relatable to current clinical methodology, as well as uncertainty estimations, while at the same time surpassing state-of-the art methods with a vertebra-level sensitivity of 93% and end-to-end AUC of 97% in a challenging setting. Moreover, we compare intra-reader agreement with model uncertainty estimates, with model reliability on par with human annotators.