Enhancing Osteoporosis and Osteopenia Diagnosis from Knee X-Rays with Attention-based Deep Learning

Abstract

Osteoporosis and osteopenia are prevalent bone disorders that have a profound effect on mobility and overall well-being, especially in older adults. Timely detection is vital to avoid severe complications, but existing diagnostic techniques, like bone densitometry, can be costly and difficult to access. This research proposes a deep learning-based approach for detecting osteopenia and osteoporosis in the knee by analysing X-ray images. The suggested model employs the effective channel attention (ECA) mechanism to improve feature extraction, ultimately leading to increased classification accuracy. The dataset employed in this research consists of labelled knee X-Ray images, divided into three groups: normal, osteopenia, and osteoporosis. To guarantee optimal performance, the model underwent rigorous training and validation using a comprehensive pipeline that incorporated data augmentation and adaptive learning techniques. The evaluation results clearly indicate that the proposed model outperforms existing methods in terms of accuracy and reliability. The findings from the experiments suggest that the ECA-based model greatly enhances the accuracy of diagnosing bone density problems, providing a more affordable and easily accessible option for early detection.