Diagnostic Prediction of portal vein thrombosis in chronic cirrhosis patients using data-driven precision medicine...

Area of Science:

• Hepatology
• Medical Informatics
• Machine Learning

Background:

• Portal vein thrombosis (PVT) is a critical complication in patients with liver cirrhosis.
• Early and accurate diagnosis of PVT is essential for effective patient management.
• Chronic hepatitis liver cirrhosis patients are particularly susceptible to PVT.

Purpose of the Study:

• To develop and validate a data-driven predictive model for diagnosing PVT in chronic liver cirrhosis patients.
• To identify key clinical and laboratory features predictive of PVT.
• To enhance diagnostic accuracy and clinical decision-making for PVT in this population.

Main Methods:

• Utilized data from 816 chronic cirrhosis patients (Lanzhou and Jilin cohorts) for training and validation.
• Employed a stacking ensemble approach combining Support Vector Machine (SVM), Naïve Bayes, and Quadratic Discriminant Analysis (QDA).
• Identified seven shared predictive features: Portal Velocity (PV), Prothrombin Time (PT), Portal Vein Diameter (PVD), Prothrombin Time Activity (PTA), Activated Partial Thromboplastin Time (APTT), age, and Child-Pugh score (CPS).

Main Results:

• The SVM and Naïve Bayes classifiers identified key features for PVT classification.
• The QDA model, trained on seven shared features, showed significant differentiation between PVT and non-PVT cases (AUROC=0.73 in Lanzhou, 0.86 in Jilin).
• The developed QDA model outperformed other machine learning methods in comparative analysis.

Conclusions:

• A comprehensive data-driven model (SVM-Naïve Bayes-QDA) for PVT diagnosis in cirrhotic patients was successfully developed.
• This model enhances clinical decision-making for managing PVT in patients with chronic liver disease.
• The proposed model offers a precise approach to PVT management in the cirrhotic population.