A Systematic Review on Deep Learning Model in Computer-aided Diagnosis for Anterior Cruciate Ligament Injury.

Abstract

Introduction: In developing Computer-Aided Diagnosis (CAD), a Convolutional Neural Network (CNN) has been commonly used as a Deep Learning (DL) model. Although it is still early, DL has excellent potential in implementing computers in medical diagnosis.

Methods: This study reviews the use of DL for Anterior Cruciate Ligament (ACL) tear diagnosis. A comprehensive search was performed in PubMed, Embase, and Web of Science databases from 2018 to 2024. The included study criteria used MRI images to evaluate ACL tears, and the diagnosis of ACL tears was performed using the DL model. We summarized the paper by reporting their model accuracy, model comparison with arthroscopy, and explainable.

Results: AI implementation in tabular format; we conclude that many medical professionals believe that arthroscopic diagnosis is the most reliable method for diagnosing ACL tears. However, due to its intrusive treatment, CAD is projected to be able to produce similar outcomes from MRI scan results. To gain the trust of physicians and meet the demand for reliable knee injury detection systems, an algorithm for CAD should also meet several criteria, such as being transparent, interpretable, explainable, and easy to use. Therefore, future works should consider creating an Explainable DL model for ACL tear diagnosis. It is also essential to evaluate the performance of this Explainable DL model compared to the gold standard of arthroscopy diagnosis.

Conclusion: There are issues regarding the need for Explainable DL in CAD to increase confidence in its result while also highlighting the importance of the involvement of medical practitioners in system design. There is no funding for this work.