eXplainable Artificial Intelligence for Hip Fracture Recognition

Abstract

Detecting hip fractures from X-rays is a critical area where artificial intelligence can significantly reduce diagnostic errors, minimize reliance on advanced imaging techniques, and expedite the diagnostic process and subsequent surgical interventions. In this paper, we present an approach of eXplainable Artificial Intelligence, which focuses not only on the accuracy of models but also on their interpretability and the ability of users to understand and trust the decisions made by the automatic system. We present a model for the automatic classification of hip fractures in radiographs based on Convolutional Neural Networks, which is enhanced by a twin Case-Based Reasoning methodology for acquiring explanatory experiences and learning how to generate textual explanations. These findings underscore the practical benefits of incorporating explanations into medical diagnostics, paving the way for improved patient outcomes and more reliable diagnostic processes.

Ähnliche Arbeiten