Longo Umile Giuseppe, De Salvatore Sergio, Piccolomini Alice, Ullman Nathan Samuel, Salvatore Giuseppe, D'Hooghe Margaux, Saccomanno Maristella, Samuelsson Kristian, Papalia Rocco, Pareek Ayoosh
Fondazione Policlinico Universitario Campus Bio-Medico Roma Italy.
Research Unit of Orthopaedic and Trauma Surgery, Department of Medicine and Surgery Università Campus Bio-Medico di Roma Roma Italy.
J Exp Orthop. 2025 Mar 22;12(1):e70195. doi: 10.1002/jeo2.70195. eCollection 2025 Jan.
There has been substantial growth in the literature describing the effectiveness of artificial intelligence (AI) and machine learning (ML) applications in total hip arthroplasty (THA); these models have shown the potential to predict post-operative outcomes using algorithmic analysis of acquired data and can ultimately optimize clinical decision-making while reducing time, cost and complexity. The aim of this review is to analyze the most updated articles on AI/ML applications in THA as well as present the potential of these tools in optimizing patient care and THA outcomes.
A comprehensive search was completed through August 2024, according to the PRISMA guidelines. Publications were searched using the Scopus, Medline, EMBASE, CENTRAL and CINAHL databases. Pertinent findings and patterns in AI/ML methods utilization, as well as their applications, were quantitatively summarized and described using frequencies, averages and proportions. This study used a modified eight-item Methodological Index for Non-Randomized Studies (MINORS) checklist for quality assessment.
Nineteen articles were eligible for this study. The selected studies were published between 2016 and 2024. Out of the various ML algorithms, four models have proven to be particularly significant and were used in almost 20% of the studies, including elastic net penalized logistic regression, artificial neural network, convolutional neural network (CNN) and multiple linear regression. The highest area under the curve (=1) was reported in the preoperative planning outcome variable and utilized CNN. All 20 studies demonstrated a high level of quality and low risk of bias, with a modified MINORS score of at least 7/8 (88%).
Developments in AI/ML prediction models in THA are rapidly increasing. There is clear potential for these tools to assist in all stages of surgical care as well as in challenges at the broader hospital administrative level and patient-specific level.
Level III.
描述人工智能（AI）和机器学习（ML）应用于全髋关节置换术（THA）有效性的文献大量增加；这些模型已显示出通过对获取的数据进行算法分析来预测术后结果的潜力，并最终在减少时间、成本和复杂性的同时优化临床决策。本综述的目的是分析关于AI/ML在THA中应用的最新文章，并展示这些工具在优化患者护理和THA结果方面的潜力。
根据PRISMA指南，截至2024年8月完成了全面检索。使用Scopus、Medline、EMBASE、CENTRAL和CINAHL数据库搜索出版物。使用频率、平均值和比例对AI/ML方法使用中的相关发现和模式及其应用进行了定量总结和描述。本研究使用了经过修改的八项非随机研究方法学指数（MINORS）清单进行质量评估。
19篇文章符合本研究要求。所选研究发表于2016年至2024年之间。在各种ML算法中，有四种模型已被证明特别重要，几乎在20%的研究中被使用，包括弹性网惩罚逻辑回归、人工神经网络、卷积神经网络（CNN）和多元线性回归。术前规划结果变量中报告的曲线下面积最高（=1），并使用了CNN。所有20项研究均显示出高质量和低偏倚风险，修改后的MINORS评分为至少7/8（88%）。
THA中AI/ML预测模型的发展正在迅速增加。这些工具在手术护理的各个阶段以及更广泛的医院管理层面和患者特定层面的挑战中都有明显的辅助潜力。
三级。
