Machine Learning-Based Prognostic Prediction of Hemorrhagic Stroke Patients and Exploration of Key Factors

Hemorrhagic stroke, a critical medical emergency marked by spontaneous cerebral bleeding, presents substantial challenges in diagnosis, treatment, and prognosis. Representing approximately 23% of all strokes in China, it is associated with a high mortality rate and severe long-term disabilities. This study aims to improve prognosis predictions by integrating patient demographics, clinical data, and advanced imaging features into a comprehensive analysis utilizing machine learning models. A total of 160 patients were enrolled, and their data were processed through decision tree, random forest, XGBoost, and LightGBM regression models. The LightGBM model demonstrated superior performance in predicting the 90-day modified Rankin Scale score, a measure of disability post-stroke, with a high degree of accuracy. Furthermore, principal component analysis and structural equation modeling were applied to identify and validate the relationships between patient outcomes and various clinical and imaging features. The study underscores the significance of early diagnosis, prompt and appropriate treatment, and the potential of machine learning in enhancing patient management and outcomes in hemorrhagic stroke. These findings contribute to the growing body of knowledge on stroke management and suggest pathways for future research in improving stroke care protocols.