A Simple and Quick Screening Method for Intrapulmonary Vascular Dilation in Cirrhotic Patients Based on Machine...

Area of Science:

• Hepatology
• Medical Imaging
• Machine Learning

Background:

• Screening for hepatopulmonary syndrome (HPS) in cirrhotic patients is challenging due to reliance on contrast-enhanced echocardiography (CEE) and arterial blood gas (ABG) analysis.
• Intrapulmonary vascular dilation (IPVD) is a key indicator of HPS, but its detection methods are invasive and time-consuming.

Purpose of the Study:

• To develop a simple, quick, and noninvasive screening method for IPVD in cirrhotic patients.
• To utilize machine learning (ML) algorithms with easily available variables for IPVD detection.

Main Methods:

• A two-step machine learning model was developed using adaptive boosting, gradient boosting decision tree, and eXtreme gradient boosting algorithms.
• The first step (NI model) used noninvasive variables, while the second step (NIBG model) incorporated both noninvasive variables and ABG results.
• Model performance was evaluated using AUCROCs, precision, recall, F1-score, and accuracy.

Main Results:

• The NI and NIBG models achieved AUCROCs of 0.850 and 0.867, respectively, with an accuracy of 87.2% for both.
• Recall values were 0.867 for the NI model and 0.875 for the NIBG model.
• Precision was 0.813 for the NI model and 0.913 for the NIBG model.

Conclusions:

• A two-step ML model effectively screens for IPVD in cirrhotic patients using noninvasive variables and ABG results.
• This ML-based approach offers a potential solution to the limitations of CEE and ABG analysis for widespread HPS screening.