| 引用本文: | 康运泽,毛谷平,潘柏祺,吴小宇,姚泽扬,涂玉成,韩铁玲,盛璞义,张紫机,李志文.人工智能术前规划系统预测全膝关节置换术假体型号的准确性研究 -基于AIJOINT系统的412例回顾性分析[J].中国临床新医学,,():-. |
| Yunze Kang,Guping Mao,Baiqi Pan,Xiaoyu Wu,Zeyang Yao,Yucheng Tu,Tieling Han,Puyi Sheng,Ziji Zhang.人工智能术前规划系统预测全膝关节置换术假体型号的准确性研究 -基于AIJOINT系统的412例回顾性分析[J].中国临床新医学,,():-. |
|
| |
|
|
| 本文已被：浏览 15次 下载 0次 |
|
|
|
| 人工智能术前规划系统预测全膝关节置换术假体型号的准确性研究 -基于AIJOINT系统的412例回顾性分析 |
|
康运泽, 毛谷平, 潘柏祺, 吴小宇, 姚泽扬, 涂玉成, 韩铁玲, 盛璞义, 张紫机, 李志文
|
|
中山大学附属第一医院
|
|
| 摘要: |
| 目的 评估AIJOINT系统预测全膝关节置换术（TKA）假体型号的准确性，并分析影响因素。 方法 回顾性纳入2025年1月至12月于中山大学附属第一医院关节外科接受初次TKA的连续病例，术前采用AIJOINT系统进行AI规划。主要结局指标为预测准确性（精确匹配或±1号内），采用Cohen加权Kappa（κw）评估一致性，bootstrap法计算95%CI。分别以股骨和胫骨假体精确匹配为因变量，将单因素分析P<0.10的变量纳入多因素logistic回归模型，采用向后逐步回归法识别独立影响因素，结果以OR及95%CI表示。 结果 股骨假体精确预测率为72.8%（300/412），κw=0.751（95%CI: 0.710～0.792）；胫骨假体精确预测率为70.1%（289/412），κw=0.723（95%CI: 0.681～0.765）。±1号内预测率股骨92.0%、胫骨91.0%。多因素分析显示，股骨假体方面，轻度（OR=6.00，95%CI: 3.11～11.57，P<0.001）和中度畸形（OR=2.42，95%CI: 1.24～4.72，P=0.010）较重度畸形预测准确性更高，女性（OR=1.57，95%CI: 1.02～2.42，P=0.041）亦为独立影响因素；胫骨假体方面，轻度（OR=2.53，95%CI: 1.40～4.56，P=0.002）、中度畸形（OR=1.86，95%CI: 1.02～3.40，P=0.045）及女性（OR=1.47，95%CI: 1.01～2.14，P=0.047）同样为独立影响因素，均呈剂量-效应趋势。 结论 AI术前规划预测国人TKA假体型号具有较高的准确性，±1号内预测率股骨92.0%、胫骨91.0%。畸形程度是预测准确性最强的独立影响因素，轻度（股骨OR=6.00，胫骨OR=2.53）和中度畸形（股骨OR=2.42，胫骨OR=1.86）均优于重度，呈剂量-效应趋势；女性亦为独立影响因素（股骨OR=1.57，胫骨OR=1.47），严重畸形患者的预测准确性显著降低，临床应用时应更谨慎参考AI预测结果。 |
| 关键词: 人工智能 关节成形术,置换,膝 假体 术前规划 准确性 |
| DOI： |
| 分类号: |
| 基金项目:科技部国家重点研发计划子课题(编号:2022YFC2407505) |
|
| Accuracy of AI-based Preoperative Planning System in Predicting Prosthesis Size for Total Knee Arthroplasty: -A Retrospective Study of 412 Cases Using AIJOINT System |
|
Yunze Kang, Guping Mao, Baiqi Pan, Xiaoyu Wu, Zeyang Yao, Yucheng Tu, Tieling Han, Puyi Sheng, Ziji Zhang
|
|
The First Affiliated Hospital, Sun Yat-sen University
|
| Abstract: |
| Objective To evaluate the accuracy of the AIJOINT system in predicting implant size for total knee arthroplasty (TKA) and to analyze the influencing factors. Methods Consecutive patients undergoing primary TKA at the Department of Joint Surgery, First Affiliated Hospital of Sun Yat-sen University between January and December 2025 were retrospectively included. Preoperative planning was performed using the AIJOINT system. The primary outcome was prediction accuracy (exact match or within ±1 size). Agreement was assessed using Cohen's weighted Kappa (κw), with 95% CIs calculated by bootstrap resampling. Multivariate logistic regression models were constructed separately for femoral and tibial component exact matching as dependent variables, with variables showing P<0.10 in univariate analysis entered into the model. Independent predictors were identified by backward stepwise selection, and results were expressed as odds ratios (OR) with 95% confidence intervals (CI). Results A total of 412 patients were included, with a mean age of 67.5±7.9 years and 67.9% female. The AI system achieved exact prediction accuracy of 72.8% (κw=0.751, 95%CI: 0.710–0.792) for femoral components and 70.1% (κw=0.723, 95%CI: 0.681–0.765) for tibial components; within ±1 size, the accuracy was 92.0% for femur and 91.0% for tibia. Multivariate analysis showed that mild deformity (OR=6.00, 95%CI: 3.11–11.57, P<0.001) and moderate deformity (OR=2.42, 95%CI: 1.24–4.72, P=0.010) compared with severe deformity, as well as female sex (OR=1.57, 95%CI: 1.02–2.42, P=0.041), were independent predictors for femoral component accuracy; for tibial components, mild deformity (OR=2.53, 95%CI: 1.40–4.56, P=0.002), moderate deformity (OR=1.86, 95%CI: 1.02–3.40, P=0.045), and female sex (OR=1.47, 95%CI: 1.01–2.14, P=0.047) were likewise independent predictors, all showing a dose-response trend. Conclusions AI-based preoperative planning demonstrates high accuracy in predicting implant size for TKA in a Chinese population, with within ±1 size accuracy of 92.0% for femoral components and 91.0% for tibial components. Deformity severity is the strongest independent factor associated with prediction accuracy, with both mild (femoral OR=6.00, tibial OR=2.53) and moderate deformity (femoral OR=2.42, tibial OR=1.86) outperforming severe deformity in a dose-response pattern. Female sex is likewise an independent factor (femoral OR=1.57, tibial OR=1.47). Prediction accuracy is significantly reduced in patients with severe deformity, warranting more cautious interpretation of AI predictions in this population. |
| Key words: Artificial intelligence Arthroplasty, replacement, knee Prosthesis Preoperative planning Accuracy |
|
|
|
|
