A machine learning analysis of difficulty scoring systems for laparoscopic liver surgery.

Area of Science:

• Hepatobiliary Surgery
• Surgical Oncology
• Minimally Invasive Surgery

Background:

• Laparoscopic liver resections (LLR) are increasingly common, necessitating accurate prediction of technical difficulty.
• Several difficulty scoring systems (DSS) have been proposed to stratify risk in LLR.
• Evaluating the predictive accuracy of existing DSS for operative and short-term outcomes is crucial.

Purpose of the Study:

• To assess the ability of four established DSS (Halls, Hasegawa, Kawaguchi, Iwate) to predict operative, short-term, and textbook outcomes in LLR.
• To identify the most effective DSS for predicting specific outcomes in LLR using machine learning.

Main Methods:

• Retrospective analysis of 346 patients undergoing LLR at a tertiary referral center.
• Evaluation of four DSS: Halls, Hasegawa, Kawaguchi, and Iwate.
• Application of machine learning algorithms (random forest) to determine DSS predictive power for operative and short-term outcomes.

Main Results:

• The Kawaguchi DSS was the best predictor of prolonged length of stay, high blood loss, and conversions.
• The Iwate DSS was most associated with operative time, and the Halls score with textbook outcomes.
• None of the investigated DSS demonstrated association with overall or severe postoperative complications.

Conclusions:

• Difficulty scoring systems are significantly related to surgical complexity and short-term outcomes in LLR.
• Kawaguchi and Iwate DSS show strong predictive performance for operative outcomes, while Halls DSS predicts textbook outcomes.
• Existing DSS do not reliably predict postoperative complications in LLR.