A machine learning-based predictive model for biliary stricture attributable to malignant tumors: a dual-center
Qifan Yang1, Lu Nie2, Jian Xu1
1Department of Gastroenterology, The Affiliated People's Hospital of Jiangsu University, Zhenjiang, Jiangsu, China.
Frontiers in Oncology
|August 13, 2024
Summary
This study identified key risk factors for malignant biliary stricture (MBS) using machine learning. The developed model accurately predicts MBS, aiding early diagnosis and improving patient outcomes.
Area of Science:
- Gastroenterology and Hepatology
- Medical Informatics
- Oncology
Background:
- Malignant Biliary Stricture (MBS) presents diagnostic challenges, impacting patient prognosis and treatment efficacy.
- Accurate differentiation between benign and malignant biliary strictures is critical for effective patient management.
Purpose of the Study:
- To identify risk factors associated with malignancy in patients diagnosed with biliary stricture via Endoscopic Retrograde Cholangiopancreatography (ERCP).
- To develop and validate a predictive clinical model for enhancing the diagnostic accuracy of Malignant Biliary Stricture.
Main Methods:
- Retrospective analysis of 398 patients with biliary stricture diagnosed by ERCP.
- Utilized univariate and Lasso regression for risk factor screening and selection.
- Developed and evaluated predictive models using seven machine learning algorithms, selecting the best based on AUROC and calibration.
Main Results:
- The Random Forest (RF) model demonstrated superior performance with an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.988.
- Key risk factors identified include age, stricture location/length, CA199, TBil, ALP, and DBil/TBil ratio.
- C-Reactive Protein (CRP) was identified as a protective factor against MBS.
Conclusions:
- The validated predictive model is effective and stable for identifying patients at high risk for MBS.
- This tool can guide clinical decisions, leading to improved diagnostic outcomes and patient prognosis.
