Explainable Machine Learning Models for Dual Diagnosis of Osteoporosis and Osteoarthritis from Knee X-Rays

Abstract

This study presents an explainable machine learningbased framework for the dual diagnosis of osteoporosis and osteoarthritis in knee X-ray images by identifying specific visual indicators, such as trabecular bone patterns, osteophyte formation, and joint space reduction, providing an accessible and inexpensive assessment option. A unified pipeline integrating image processing and deep learning techniques is developed, where each component corresponds to a specific radiographic feature. The framework links each diagnosis to the appropriate radiographic markers and facilitates the identification of instances in which osteoporosis and osteoarthritis coexist within the same radiograph. By emphasizing visual explanation, the proposed approach enhances interpretability and supports clinically meaningful decision-making.

Ähnliche Arbeiten