Precision Medicine in ICH Unveiling the Superior Predictive Power of a Joint Model.

Abstract

This study examined the effectiveness of a combined model using non-contrast computed tomography (NCCT) imaging, clinical data, and radiomics for predicting early hematoma enlargement in patients with spontaneous intracerebral hemorrhage. The study involved 232 patients with primary cerebral hemorrhage who met the inclusion criteria at the General Hospital of the Western Theater Command, PLA, between January 2018 and December 2023. Imaging and clinical features were compared, radiomic features were extracted from head CT scans, and a multivariate logistic regression model identified key imaging markers and clinical features. Univariate and multivariate logistic regression models were used for dimensionality reduction of radiomic features and to develop a radiomic signature/model. Patients were split into training and validation sets in a 7:3 ratio. Then, NCCT, clinical, radiomics, and combined NCCT-clinical-radiomics models were built, along with a nomogram. The AUC values for hematoma expansion prediction were as follows in the training set: NCCT model (0.758), clinical model (0.742), radiomics model (0.779), and combined model (0.872). In the validation set, the AUCs were: NCCT model (0.853), clinical model (0.754), radiomics model (0.778), and combined model (0.905). Calibration and decision curve analysis further confirmed the superior clinical utility of the combined model over the individual models. In conclusion, the combined NCCT-clinical-radiomics model significantly outperformed the individual models, leading to improved predictive accuracy, stability, and generalizability.