Utilizing Machine Learning Techniques to Predict Negative Remodeling in Uncomplicated Type B Intramural Hematoma.

Area of Science:

• Cardiovascular Imaging and Intervention
• Artificial Intelligence in Medicine
• Vascular Surgery

Background:

• Uncomplicated Stanford type B intramural hematoma (IMHB) can undergo negative remodeling.
• Predicting this remodeling in the acute phase is crucial for patient management.
• Machine learning (ML) offers potential for predictive modeling in this context.

Purpose of the Study:

• To evaluate the effectiveness of ML techniques in predicting negative remodeling in acute-phase uncomplicated IMHB.
• To identify key clinical features associated with negative remodeling.

Main Methods:

• Compared univariate logistic regression with 8 ML models, including LASSO for feature selection and hyperparameter tuning via randomized and grid searches.
• Model performance was assessed using AUC, F1 score, and Brier score.
• SHAP values were used for feature importance analysis.

Main Results:

• The LASSO method identified 7 significant features: lymphocytes, white blood cells, neutrophil-to-lymphocyte ratio, eosinophils, monocytes, hypertension, and statins.
• The CatBoost model demonstrated superior performance with an AUC of 0.969, F1 score of 0.94, and Brier score of 0.0625.
• SHAP analysis highlighted monocyte count (1.250), lymphocyte count (0.296), and eosinophil count (0.249) as the top predictors in the CatBoost model.

Conclusions:

• The CatBoost ML model effectively predicts negative remodeling in acute-phase uncomplicated IMHB.
• Integrating clinical features like monocyte, lymphocyte, and eosinophil counts improves predictive accuracy.
• This ML approach can serve as a valuable tool for enhancing patient outcomes.