Are artificial intelligence generated lower extremity radiographic measurements accurate in a cohort with implants?

Abstract

Leg length discrepancy (LLD) and malalignment of the lower extremity can lead to pain and increased risk of osteoarthritis. Radiographic measurements on anteroposterior (AP) full-length radiographs can be used to assess LLD and lower extremity alignment. The primary aim of this study was to evaluate the accuracy of AI software in performing lower extremity radiographic measurements in patients with implants. The secondary aim was to compare its efficiency to that of radiologists. This study used the following eight angles and five lengths: hip-knee-angle (HKA), anatomical-tibiofemoral angle (aTFA), anatomical-mechanical-axis angle (AMA), joint-line-convergence angle (JLCA), mechanical-lateral-proximal-femur-angle (mLPFA), mechanical-lateral-distal-femur-angle (mLDFA), mechanical-medial-proximal-tibia-angle (mMPTA), mechanical-lateral-distal-tibia- angle (mLDTA), femur length, tibia length, full leg length, leg-length-discrepancy (LLD), and mechanical-axis-deviation (MAD). Two radiologists and AI software independently performed these measurements on 156 legs. The statistical methods used to assess AI performance were intraclass correlation coefficient (ICC) and Bland-Altman analysis. The AI generated output for 129/156 legs. 11/13 of the variables showed excellent agreement (ICC ≥ 0.75) between AI and the readers. Bland Altman performance targets were met for 5/13 variables. The mean (standard deviation) reading time for the AI and two readers, respectively, was 38 (6) seconds, 181 (41) seconds, and 214 (77) seconds. In a cohort with lower extremity metal implants, AI-based leg length measurements were fast and accurate although most angular measurements were not.