Publication extender

Farewell to "Blind Exploration" Treatment: New AI Tool Achieves Intelligent Prediction of Spontaneous Gallbladder Stone Passage

Journal of Clinical Hepatology

What is it about?

This article aims to construct and verify a machine learning model for predicting the spontaneous passage of gallbladder stones in patients with choledocholithiasis and develop a corresponding web application to assist in clinical decision-making. The study retrospectively collected data from two centers, using LASSO regression to screen out five key predictive factors: solitary stones, unexpanded common bile duct, stone diameter <8 mm, decreased serum ALP, and decreased GGT. The model demonstrated good predictive performance in the training and validation sets, with the ROC curve showing high discriminability. The innovation lies in combining AutoML technology to improve the efficiency of model construction and using SHAP plots to intuitively explain individual prediction results, enhancing the model's interpretability. In addition, the study developed a user-friendly web application to facilitate clinicians in quickly assessing the probability of spontaneous stone passage in patients, avoiding unnecessary ERCP and its complications. External validation further proves the generalization ability of the model. This tool combines high predictive accuracy and operational convenience, and has strong clinical transformation value.

Why is it important?

The main significance of this study lies in constructing and verifying an interpretable machine learning model for predicting the possibility of spontaneous stone passage in patients with choledocholithiasis, thereby reducing unnecessary ERCP procedures. Although ERCP is an important treatment method for choledocholithiasis, it carries risks of complications such as pancreatitis, bleeding, and perforation. For patients who have spontaneously passed stones, the operation is unnecessary and may pose additional risks. The study screens out five key predictive factors: solitary stones, unexpanded common bile duct, stone diameter <8 mm, decreased ALP, and decreased GGT, using LASSO regression, which improves the simplicity and clinical applicability of the model. Furthermore, the study uses GBM model combined with SHAP analysis to enhance the model's interpretability, allowing clinicians to understand the prediction logic and overcome the "black box" problem. In addition, the study developed a web application based on R and Python, using the Dash library to implement an interactive interface, facilitating real-time clinical application and promoting the application of artificial intelligence tools in medical scenarios. Overall, this study not only improves the ability to identify spontaneous stone passage but also provides technical support for individualized diagnosis and treatment, and has important clinical transformation value.

Resources1 total

Who is involved?