Development and validation of a machine learning-based predictive model for prognosis in cerebral hemorrhage patients...

JOURNAL/mgres/04.03/01612956-202701000-00003/figure1/v/2026-09-13T085902Z/r/image-tiff Intracerebral hemorrhage is a common and critical neurological emergency, and the inconsistent results of hyperbaric oxygen therapy make predicting outcomes challenging. Therefore, a reliable predictive model for outcomes is needed to improve treatment plans. This study used interpretable machine learning to determine key risk factors linked to adverse outcomes in cerebral hemorrhage patient post-hyperbaric oxygen therapy. To improve prognostic accuracy and support clinical decision-making, we analyzed the data from 401 patients with intracerebral hemorrhage who received hyperbaric oxygen therapy at the First Affiliated Hospital of Jiaxing, China, using multivariate logistic regression and machine learning to develop predictive models, with model performance assessed via 10-fold cross-validation. SHapley Additive exPlanations values aided in model interpretation and in the development of a logistic regression-based nomogram. Based on the modified Rankin scale, patients were stratified into two prognostic groups: a good prognosis group (modified Rankin scale < 3, n = 265) and a poor prognosis group (modified Rankin scale ≥ 3, n = 136). Significant differences in various factors were observed between the groups. Independent risk factors identified included the Glasgow coma scale score, computed tomography hemorrhage volume, activities of daily living score, treatment modality, intraventricular bleeding, and rebleeding. Among the models tested, the random forest algorithm yielded the highest predictive performance (the area under the receiver operating characteristic curve = 0.99). The study identified key predictors, including activities of daily living score, computed tomography hemorrhage volume, and Glasgow coma scale score, as factors influencing hyperbaric oxygen therapy outcomes in intracerebral hemorrhage patients. Random Forest model demonstrated strong predictive capability, with SHapley Additive exPlanations analysis providing insights into individual risk factors. Overall, these findings are crucial for improving clinical decision-making and personalized treatment.