Establishing a preoperative predictive model for gallbladder adenoma and cholesterol polyps based on machine...

Area of Science:

• Hepatobiliary surgery
• Medical imaging
• Machine learning in medicine

Background:

• Gallbladder polypoid lesions (GPLs) require accurate preoperative differentiation.
• Distinguishing benign cholesterol polyps from potentially malignant gallbladder adenomas is critical for patient management.
• Current diagnostic methods may not always provide sufficient preoperative accuracy.

Purpose of the Study:

• To develop and validate a machine learning-based preoperative prediction model.
• To accurately differentiate between gallbladder adenomas and cholesterol polyps.
• To improve risk stratification and guide personalized treatment strategies for GPLs.

Main Methods:

• Retrospective analysis of clinical, serological, and ultrasound imaging data.
• Construction and evaluation of 110 combination predictive models using 12 machine learning algorithms.
• Validation using internal and external cohorts, ROC curves, AUC, calibration, and decision curves.

Main Results:

• The Support Vector Machine + Random Forest (SVM + RF) model achieved the highest AUC (0.972 training, 0.922 internal validation).
• Key predictive features included gallbladder wall thickness, polyp size, echo, and pedicle.
• External validation confirmed the model's excellent predictive performance in distinguishing adenomas from cholesterol polyps.

Conclusions:

• An SVM + RF predictive model effectively differentiates gallbladder adenomas from cholesterol polyps preoperatively.
• This machine learning approach aids clinicians in assessing GPL risk.
• Personalized treatment strategies can be enhanced by accurate preoperative risk stratification.