A Comparative Analysis of Deep Learning Models and Conventional Approaches for Osteoporosis Detection in Hip X-Ray Images
Virja Kawade, Vedant Naikwade, Vibha Bora, Sharda Chhabria
Abstract
Osteoporosis is a bone disease marked by reduced bone mineral density (BMD) and a higher risk of fracture. The DEXA or DXA (Dual-Energy X-Ray Absorptiometry) scan is one of the modern methods used for osteoporosis detection and diagnosis and it is currently the benchmark for accurate osteoporosis detection. This method’s only drawback is that it is expensive and unaffordable for those from lower socioeconomic strata. Thus, by conducting this study, the authors propose the use of deep learning techniques on X-Ray images which may be used as an alternative to DEXA. The study summarizes and compares four deep learning algorithms employed for the same – ResNet-50, Inception Net, YOLOv7, and Ultralytics YOLO v8 models on an X-Ray dataset of hips consisting of 117 images after augmentation. Out of the four algorithms, the best accuracy was obtained for the YOLO v8 model.