Artificial Intelligence in Surgery Revisited: A 2025 Update on Machine Learning for Predicting Complications and...

Area of Science:

• Artificial Intelligence
• Surgical Outcomes Research
• Medical Informatics

Background:

• Machine learning (ML) offers advanced capabilities for predicting surgical complications and outcomes.
• ML models can identify complex, nonlinear relationships in data, outperforming traditional statistical methods.
• Current ML tools show promise across various surgical fields, exceeding conventional risk assessment models.

Purpose of the Study:

• To review the transformative potential of machine learning in surgical prognostication.
• To highlight the advantages of ML over traditional statistical approaches in predicting surgical outcomes.
• To discuss the challenges and considerations for integrating ML into clinical surgical practice.

Main Methods:

• Review of emerging machine learning applications in surgical outcome prediction.
• Analysis of ML's ability to model complex, nonlinear relationships in patient data.
• Examination of challenges such as model transparency, external validation, and rare event modeling.

Main Results:

• ML-based tools demonstrate strong performance in surgical risk prediction, often surpassing traditional models.
• Key challenges include "black box" outputs, performance decay on external validation, and difficulty with rare events.
• Successful integration necessitates rigorous validation, transparency, and addressing methodological considerations for surgeons.

Conclusions:

• Machine learning significantly enhances the precision of surgical risk prediction.
• ML can guide patient selection, optimize perioperative care, and improve shared decision-making.
• Addressing ML's limitations is crucial for its effective and reproducible clinical implementation.