Diagnostic Performance of Artificial Intelligence for Detection of Anterior Cruciate Ligament and Meniscus Tears: A Systematic Review
PMID: 32956803
Imaging-based detection of anterior cruciate ligament (ACL) injuries use magnetic resonance imaging (MRI) as its current gold standard for diagnosis. However, the accuracy of these MRI readings may decrease due to observer inexperience, presence of multiple injuries, and small or incomplete tears that are difficult to detect. Artificial intelligence (AI) technology has helped in diagnosing diseases in other fields of medicine, such as detecting diabetic retinopathy and skin cancer. Despite the use of AI in other branches of medicine, its performance and clinical utility in sports medicine has remained poorly defined. This article evaluates the diagnostic accuracy of AI technology in diagnosing ACL and meniscus tears compared to human clinical experts.
Although findings suggest that AI technology did not outperform clinical experts in diagnosing ACL and meniscus tears, the current study found that a combination of AI and human outperformed human experts or AI alone when diagnosing ACL and meniscus tears. It is possible that as algorithms used in artificial intelligence are further trained with more data, that prediction accuracies and image recognition improves and may eventually outperform these experts. Ensuring clinical relevance at every step in algorithm design and development will lead to true progress for AI in orthopedic sports medicine.
