Bone X-Ray Classification For Upper Extremity Radiographs

Abstract

Accurate identification of abnormalities in the upper limb is crucial for effective diagnosis and treatment in musculoskeletal conditions. This research paper presents a deep learningbased approach aimed at surpassing human performance in detecting abnormalities in radiographs of the upper limb. The dataset used for training comprises 40,561 multi-view radiographic images from approximately 14,000 musculoskeletal investigations, with expert radiologists manually labeling each study as normal or abnormal. Through extensive testing and meticulous model tuning, remarkable results are achieved, with an accuracy of 78.57% and an F1 score of 78.43%, rivaling the kappa score reported by the MURA dataset producer. This study contributes to advancing the field by demonstrating the potential of deep learning techniques to outperform humans in identifying abnormalities in upper limb radiographs. The practical applicability of deep learning models is highlighted, showcasing their exceptional accuracy in supporting medical professionals in
diagnosing and managing musculoskeletal disorders. This research represents a testament to the ongoing advancements in artificial intelligence, benefiting patient care and outcomes in the medical domain