Clinical Report: Development and assessment of machine learning algorithms for predicting postoperative outcomes in biliary tract cancers
Overview
This study developed a machine learning model to predict postoperative outcomes in biliary tract cancer patients using data from the SEER database and a medical center. Key independent prognostic factors were identified.
Background
Biliary tract malignancies are aggressive tumors with poor prognosis, often diagnosed at advanced stages. The lack of effective screening and treatment options is noted.
Data Highlights
Data Source
Patients
SEER Database
6,951
Medical Center
245
Key Findings
Independent risk factors for overall survival included age, sex, marital status, tumor site, tumor differentiation, AJCC stage, regional nodes examined, regional nodes positive, and LN surgery scope.
The random forest model achieved good AUC values across different datasets.
Calibration curves and decision curve analysis indicated strong performance of the random forest model.
SHAP analysis identified regional nodes positive, AJCC stage, tumor site, and age as crucial variables in survival analysis.
Chemotherapy was associated with improved overall survival in patients with positive lymph nodes in both training and test cohorts.
Clinical Implications
Understanding the identified prognostic factors can guide treatment planning.
Conclusion
The study presents a machine learning model that predicts postoperative prognosis in biliary tract malignancies.
