Machine learning to predict the decision to perform surgery in hepatic echinococcosis

Area of Science:

• Hepatology
• Parasitology
• Surgical Decision Support

Background:

• Cystic echinococcosis (CE) is a major public health concern, predominantly impacting the liver.
• Current management of hepatic CE lacks precise tools for guiding surgical decisions.
• This study addresses the need for predictive models to optimize patient stratification for surgical or non-surgical treatment.

Purpose of the Study:

• To develop and validate predictive models for surgical decision-making in hepatic cystic echinococcosis.
• To enhance the accuracy of patient allocation to surgical versus non-surgical management pathways.
• To provide a dynamic tool for clinical practice that improves upon static guidelines.

Main Methods:

• Retrospective analysis of 406 hepatic CE patients (2009-2021).
• Development of Cox regression and decision tree models using clinical, imaging, and treatment data.
• Model performance validated through K-fold cross-validation, train/test split, and bootstrapping.

Main Results:

• Imaging findings and symptomatology were identified as key predictors for surgical intervention.
• The Cox model achieved a concordance index of 0.94 and an AUC of 0.96.
• The decision tree model, incorporating imaging, cyst stage, and symptoms, demonstrated strong performance (mean AUC 0.950).

Conclusions:

• Validated predictive models for assessing surgical risk in hepatic CE have been developed.
• These models offer a dynamic clinical tool to optimize patient management pathways.
• Integration into practice can lead to improved outcomes by enhancing precision in treatment allocation.