A Predictive Model Integrating Coagulation Profiles for Preoperative Risk Stratification of Difficult Single-Incision Laparoscopic Cholecystectomy: A Retrospective Cohort Study Based on Explainable Machine Learning

Background Accurate preoperative prediction of surgical difficulty in single-incision laparoscopic cholecystectomy (SILC) remains challenging, as models for multiport surgery are not directly applicable. Methods This retrospective single-center study included 520 patients, who were randomly divided into a training cohort (n=364) and an internal validation cohort (n=156). Six predictors were identified, guiding 8 machine learning models. Performance was assessed using the area under the curve (AUC), calibration,…

By the Truvace deskOctober 15, 20262 min read1 sourcePeer-reviewedTRV-2026-1469· cite & verify

In brief

Researchers built and tested machine learning models to predict which patients would be harder to operate on with single-incision laparoscopic cholecystectomy, since prediction tools designed for multiport surgery do not transfer directly. They used a retrospective single-center cohort of 520 patients split into training and internal validation groups.

The work matters because surgeons could use an interpretable, openly available calculator to anticipate difficulty before operating. The evidence is retrospective and from one center with only internal validation, so performance in other hospitals and in routine practice is not yet shown.

Main points

The study included 520 patients, divided into a training cohort of 364 and an internal validation cohort of 156.
Six predictors guided eight machine learning models, and a neural network performed best with an AUC of 0.867.
Excluding coagulation parameters lowered the AUC by 0.043, suggesting they add predictive value.
A sensitivity analysis using a hard-event endpoint gave an AUC of 0.877, and the model is available as an online calculator.

The gain

A neural network built on six predictors including coagulation parameters discriminated difficult from non-difficult single-incision cholecystectomy cases in internal validation, potentially aiding patient selection and surgical planning.

The rundown

The authors addressed the gap that preoperative difficulty prediction models for multiport laparoscopic surgery are not directly applicable to the single-incision approach. They selected six predictors and trained eight machine learning models on the training cohort.

Performance was judged with AUC, calibration, decision curve analysis, and classification metrics. The best neural network reached accuracy of 0.833, specificity of 0.941, NPV of 0.855, and a Brier score of 0.118, with net benefit across clinical thresholds.

The model was interpreted with SHapley Additive exPlanations and deployed as a public online calculator. Ablation and correlation analyses indicated coagulation parameters provide complementary predictive value.