ECOLOGICAL VALIDATION AND RELIABILITY OF AN AI-BASED MECHANICAL DIAGNOSIS AND THERAPY (MDT) SYSTEM FOR CHRONIC MECHANICAL LOW BACK PAIN

Although mechanical diagnosis and therapy (MDT) is generally used for its accurate “hands-on manual or face-to-face diagnostic” classification and therapeutic efficacy in treating low back pain (LBP), the validity and test–retest reliability of an AI-based MDT platform (“Dr. AI”) for chronic LBP diagnosis remain unestablished. This research aimed to assess the validity and reliability of Dr. AI compared to the conventional face-to-face diagnosis by physical therapists (“Dr. PT”). Fifty participants ([Formula: see text] years, eight females) with chronic LBP were classified into postural, derangement, and other syndromes using MDT guidelines based on the signs, posture, and repeated movement tests. Convolutional neural networks (CNNs) were employed to compare the classification accuracies of Dr. PT and Dr. AI. The significance value for linear regression and intraclass correlation coefficient (ICC) analyses was set at [Formula: see text]. Linear regression analysis demonstrated excellent validity ([Formula: see text], [Formula: see text]), indicating the outstanding accuracy of Dr. AI. The test–retest reliability data demonstrated good-to-excellent ICC (0.82–0.94; [Formula: see text]), indicating consistent measurements by Dr. AI. These findings highlight the important clinical application of the Dr. AI MDT system, which accurately and consistently identifies and classifies syndromes in chronic LBP in a remote or ecological environment. Our novel study has demonstrated more than 10% higher reliability and validity than previous studies.