Authors: Kolse Sejal Balasaheb, Prof. R. V. Varhade, Prof. M. C. Kolse, Prof. S. K. Said

Abstract: Bone fractures are common injuries that require accurate and timely diagnosis for appropriate treatment. Conventional detection relies on interpretation of X-ray images by radiologists or emergency physicians, and subtle fractures can be missed, particularly during busy shifts or when specialist reading is unavailable. Artificial Intelligence (AI), especially deep learning and computer vision, offers a promising way to assist fracture detection. This paper reviews AI-based fracture detection on radiographs. It examines datasets such as MURA and GRAZPEDWRI-DX, convolutional neural networks with transfer learning for classifying images as fractured or normal, object-detection models such as Faster R-CNN and YOLO for localizing fracture regions, segmentation networks, and visual explanation methods that highlight suspicious areas. Studies showing that AI assistance improves clinician sensitivity and reading efficiency are discussed. A layered architecture covering image acquisition, preprocessing, a detection engine, explanation and a clinical review interface is proposed, along with the working methodology, evaluation metrics, a comparison of model families and a discussion of challenges including image quality, dataset diversity, fracture type and clinical validation. The review concludes that AI should support rather than replace professional judgement and can help prioritize urgent cases and reduce missed fractures.

DOI: https://doi.org/10.5281/zenodo.24218642