Performance Analysis of Artificial Intelligence-Based Posture Estimation Software for Lumbar Posture Assessment: A Pilot Comparison with Plain Radiographs

Study Design: Retrospective cross-sectional pilot studyObjectives: This study aimed to assess the correlation between digital lumbar lordotic alignment (DLA), measured from digital photographs using artificial intelligence (AI)-based posture estimation software, and the lumbar lordotic angle (LLA), measured using whole-spine lateral radiographs.Summary of Literature Review: Periodic postural evaluations play a crucial role in preventing potential health problems.Although radiographs have been considered the gold-standard method for postural evaluation, their costs and radiation exposure limit the ability to conduct periodic assessments.In contrast, digital photographs may be an alternative tool without the risk of radiation exposure.Materials and Methods: DLA was measured using AI-based posture estimation software with a lateral two-dimensional image captured by a mobile device.Each participant was photographed once by three healthcare professionals, for a total of three sessions.The correlation between DLA and LLA, along with the inter-rater reliability of DLA measurements, was analyzed.Results: A total of 34 participants were enrolled.A weak positive correlation was found between DLA and LLA (Spearman's rho=0.364,p=0.034).The inter-rater reliability of DLA measurements was moderate.Conclusions: AI-based, non-invasive posture estimation software may have potential as a screening tool for periodic lumbar spine alignment assessments.