This study aims to determine the accuracy and inter-rater reliability of an AI-powered mobile phone health application that estimates curve magnitude from 3D topography captured through a smartphone video. This is a prospective observational validation study of this health application on first use in clinic and at-home. Adolescent idiopathic scoliosis (AIS) patients, 10–18 years old, with Cobb angles ≤45° were recruited. Participants were assisted in downloading the application onto their personal smart phone. Spinal X-rays were taken in the clinic as per standard of care and curve magnitudes were measured by spine specialists blinded to the study. A phone scan was performed in the clinic by a single, trained research assistant and participants were asked to complete a scan themselves at home later that same day, and once a month for the next 6-months. Successful versus failed scans were tracked. Interclass correlation (ICC) analyses were employed to determine (1) the accuracy of the application's Cobb angle predictions against radiographic measures and (2) the interrater reliability of the application by comparing in-clinic and initial at-home scan predictions. There were 54 patients consented into the study. The mean Cobb angle from radiographic meaures was 28.1±10.6°. Forty-four patients (81%) had a successful scan in clinic, and 21 (39%) completed at least one home-scan, either immediately after their clinic visit (12%) or at the one-month mark post-clinic visit (31%). There was moderate agreement between radiographic measurements and scan predictions (ICC=0.71, 95% CI 0.51–0.84, p The application shows potential for remote monitoring of scoliosis. However, the accuracy and number of failed scans suggest it is not yet a reasonable replacement for x-rays and/or in-person clinical evaluation.
Nadler et al. (Wed,) studied this question.