Applications of Ensemble learning techniques for Knee Arthroplasty surgery: a systematic review

Background and aims: Ensemble models consist of multiple algorithms that individual resultsare aggregated in various ways. Compared to individual predictors, such as a neural network, theyare more stable and provide better predictions. usage of artificial intelligence in the clinical arearecommends the outstanding accuracy of these hybrid models. In knee arthroplasty, single algorithmshave been widely used to improve the decision-making process, surgical planning, accuracy,and repeatability of surgical procedure, however, there is no evidence to review applicationsand capability of ensemble methods for knee arthroplasty. This study aims to provide a systematicoverview of the ensemble learning techniques were used for Knee Arthroplasty surgery.Method: A systematic search of PubMed, Embase, the Cochrane Library, ERIC database, andgoogle scholar was conducted for articles published in English between ۲۰۱۰ and ۲۰۲۳. Thesearch terms used related to “ensemble learning” and “stacking”, “boosting”, “bagging”, “artificialintelligence”, and “machine learning” algorithms in” knee arthroplasty” “TKA” and “UKAsurgery. since this paper focuses on machine learning techniques applied in medicine, a combinationof strategies from two related guidelines was implemented to conduct the systematicliterature review (SLR) which includes ۱۷-item checklist of the Preferred Reporting Items forclinical Systematic reviews and Meta-Analyses (PRISMA) statement and Software EngineeringGuidelines for Conducting Systematic reviews.Results: After applying inclusion and exclusion criteria, we identified ۳۶ studies of which ۲۳were boosting and ۱۰ were stacking designs, and three used bagging methods. The review showedthat the most common classes of combining different models were boosting and stacking. wealso found that stacking with neural networks was mostly recommended. The development of anEnsemble deep learning technique for TKA/UKA using Fluoroscopy, x-ray, and MRI images performwell with an average of accuracy ۹۰.۰۵% to build a ۳D model of customized implants andpredicting preferred intraoperative sagittal alignment of the implant during arthroplasty.Conclusion: Applications of meta-approach to machine learning for knee arthroplasty surgery areexpanding rapidly and offer significant improvement in a forecast of the final model. This studyprovides evidence that ensemble algorithms have the potential for better strategies with morecapabilities than single classifiers to predict accurately.

کلیدواژه ها:

Ensemble learning ، Knee arthroplasty ، Machine learning