Performance and clinical implications of machine learning models for detecting cervical ossification of the posterior longitudinal ligament: a systematic review.

Abstract

Ossification of the posterior longitudinal ligament (OPLL) is a significant spinal condition that can lead to severe neurological deficits. Recent advancements in machine learning (ML) and deep learning (DL) have led to the development of promising tools for the early detection and diagnosis of OPLL. This systematic review evaluated the diagnostic performance of ML and DL models and clinical implications in OPLL detection. A systematic review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. PubMed/Medline and Scopus databases were searched for studies published between January 2000 and September 2024. Eligible studies included those utilizing ML or DL models for OPLL detection using imaging data. All studies were assessed for the risk of bias using appropriate tools. The key performance metrics, including accuracy, sensitivity, specificity, and area under the curve (AUC), were analyzed. Eleven studies, comprising a total of 6,031 patients, were included. The ML and DL models demonstrated high diagnostic performance, with accuracy rates ranging from 69.6% to 98.9% and AUC values up to 0.99. Convolutional neural networks and random forest models were the most used approaches. The overall risk of bias was moderate, and concerns were primarily related to participant selection and missing data. In conclusion, ML and DL models show great potential for accurate detection of OPLL, particularly when integrated with imaging techniques. However, to ensure clinical applicability, further research is warranted to validate these findings in more extensive and diverse populations.