Systematic Review of Digital Innovations in Surgery: Machine Learning Applications and Implementation Guidelines

Abstract

Digital technologies, particularly machine learning (ML), are increasingly integrated into contemporary surgical practice, though implementation barriers remain. This systematic review examined the current evidence on digital innovations in surgery, focusing on ML applications, implementation challenges, and evidence-based implementation strategies. A comprehensive search of seven electronic databases identified 87 studies published between January 2018 and December 2024, following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses framework. Evidence synthesis encompassed six domains: artificial intelligence and ML applications, extended reality (XR) technologies, clinician-led innovation, sustainable surgical practices, specialized training models, and nontechnical skills development. ML models demonstrated improved performance in preoperative risk stratification compared with conventional statistical approaches. Receiver operating characteristic analysis showed ML models, including deep neural networks (area under the curve [AUC] ≈ 0.82–0.96), random forests (AUC ≈ 0.86–0.93), and support vector machines (AUC ≈ 0.86), outperformed traditional logistic regression (AUC ≈ 0.68–0.74) for predicting postoperative complications. Computer vision algorithms improved procedural precision, while XR technologies (virtual reality/augmented reality) enhanced surgical training, showing skill acquisition comparable or superior to traditional methods. However, substantial implementation barriers were identified, including algorithmic bias, insufficient training in digital competencies, regulatory constraints, and documented concerns regarding bias in nonrandomized studies of XR technologies. Environmental impact assessments revealed that telemedicine applications reduced carbon emissions, whereas robotic surgical systems demonstrated higher resource consumption. The successful integration of digital innovations requires a phased implementation approach, multidisciplinary collaboration, comprehensive competency development, and systematic evaluation of both clinical and operational outcomes. This review provides recommendations for translating digital innovations, addressing technical, regulatory, and human factors.

Ähnliche Arbeiten