Parametric Comparison of Deep Learning Models for Musculoskeletal Abnormality Detection

Musculoskeletal disorders are conditions that affect joints, bones, muscles, and the spine. The authors of this paper aim to investigate the use of artificial intelligence (AI), particularly deep learning, to assist radiologists in the initial screening of abnormal radiographs. The aim of this study is to provide a comprehensive comparison of the various techniques that can be implemented to enhance the performance of existing models. The authors explore the use of various preprocessing and ensemble techniques on pre-trained models to thoroughly make use of the benefits of transfer learning. The authors of this paper focus on classifying abnormal shoulder radiographs in particular, as a detailed literature survey has pinpointed poorer performance on radiographs of the upper extremities. This paper offers a thorough comparison utilizing parameters such as precision, recall, accuracy, F1-score, and Cohen’s Kappa Score.