کنفرانس مهندسی برق ایران

صفحه اصلی / سی و سومین کنفرانس بین المللی مهندسی برق

Deep Learning Meets Explainable AI: A Robust Framework for X-Ray Fracture Detection

نویسندگان :

Ali Tamizifar¹ Shakiba Berenjkoub² Mina Amiri³

1- Isfahan university of technology 2- Isfahan university of technology 3- Isfahan university of technology

کلمات کلیدی :

Bone Fracture،FracAtlas،Explainable AI

چکیده :

Bone fractures, commonly diagnosed by X-ray imaging, often pose challenges for accurate detection due to variability in appearance and image quality. This study proposes a three-stage deep learning framework to address these challenges, namely preprocessing, binary classification, and localization. The preprocessing stage enhances image quality and removes extraneous parts of the X-ray images to isolate the primary regions of interest. The Swin Transformer V2 model achieves a classification accuracy of 95.41%, while YOLOv11 attains a mAP50 of 0.606 for fracture localization. To ensure explainability, Gradient-weighted Class Activation Mapping (Grad-CAM) is utilized to visualize the outputs from the last layer before the classifier and highlight the regions most relevant to the model’s decisions. This robust framework not only achieves high accuracy but also offers transparency, making it suitable for clinical applications in fracture detection.