目的:腿长差异(LLD)和下肢不对齐可导致疼痛和骨关节炎。各种射线照相参数用于评估LLD和对准。FDA批准的510(k)人工智能(AI)软件在全腿站立射线照片上定位地标并执行多项测量。这项研究的目的是评估与三个手动阅读器相比,该AI工具的可靠性。
方法:使用320条腿的样品。将三个阅读器的测量值与髋膝角度(HKA)的AI输出进行了比较,解剖-胫股角(aTFA),解剖-机械-轴角(AMA),联合线会聚角(JLCA),机械-外侧-股骨近端-角度(mLPFA),机械外侧-股骨远端角度(mLDFA),机械-内侧-近端-胫骨角(mMPTA),机械-外侧-远端-胫骨角(mLDTA),股骨长度,胫骨长度,腿全长,腿长差异(LLD),和机械轴偏差(MAD)。使用组内相关系数(ICCs)和Bland-Altman分析来跟踪性能。
结果:在研究中成功地产生了272/320腿的AI输出。读者与AI配对ICC大多处于优异范围:读者1、2和3的12/13、12/13和9/13变量分别处于优异范围(ICC>0.75)。对腿长有更好的约定,股骨长度,胫骨长度,LLD,和HKA比其他变量。三位读者和人工智能的平均阅读时间分别为250、282、236和38秒,分别。
结论:这项研究表明,基于AI的软件提供了对LLD和下肢对准的可靠评估,并节省了大量时间。
OBJECTIVE: Leg length discrepancy (LLD) and lower extremity malalignment can lead to pain and osteoarthritis. A variety of radiographic parameters are used to assess LLD and alignment. A 510(k) FDA approved artificial intelligence (AI) software locates landmarks on full leg standing radiographs and performs several measurements. The objective of this study was to assess the reliability of this AI tool compared to three manual readers.
METHODS: A sample of 320 legs was used. Three readers\' measurements were compared to AI output for 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). Intraclass correlation coefficients (ICCs) and Bland-Altman analysis were used to track performance.
RESULTS: AI output was successfully produced for 272/320 legs in the study. The reader versus AI pairwise ICCs were mostly in the excellent range: 12/13, 12/13, and 9/13 variables were in the excellent range (ICC > 0.75) for readers 1, 2, and 3, respectively. There was better agreement for leg length, femur length, tibia length, LLD, and HKA than for other variables. The median reading times for the three readers and AI were 250, 282, 236, and 38 s, respectively.
CONCLUSIONS: This study showed that AI-based software provides reliable assessment of LLD and lower extremity alignment with substantial time savings.