Prediction of Basal Ganglia Hematoma Expansion Based on Radiomics and Clinical Characteristics: A Novel Multivariate Predictive Nomogram

Background. This study is aimed at formulating and authenticating a pioneering nomogram integrating noncontrast computed tomography (NCCT) mean CT densities (m-CTD) of hematoma, morphological indicators from NCCT hematoma, and clinical manifestations to foresee hematoma expansion (HE) in patients suffering from spontaneous basal ganglia hemorrhage (BGH). Methods. A predictive model was constructed by retrospectively evaluating the data from 406 patients. This model was externally validated using an independent dataset of 174 patients. Multivariate logistic regression analysis was deployed to discern independent prognostic indicators and to generate a nomogram for HE prediction. Model calibration was examined using 1000 bootstrap samples for internal validation. Results. Multivariate logistic regression disclosed that m-CTD (odds ratio (OR) 0.846, 95% confidence interval (CI) 0.782-0.909), baseline hematoma volume (BHV) (OR 1.055, 95% CI 1.017-1.095), NCCT blend sign (BS) (OR 3.320, 95% CI 1.704-6.534), NCCT black hole sign (BHS) (OR 2.468, 95% CI 1.293-4.729), systolic blood pressure (SBP) (OR 1.027, 95% CI 1.014-1.040), and homocysteine (Hcy) (OR 1.075, 95% CI 1.038-1.114) were independent predictors of HE. The area under the curve (AUC) for the training and validation datasets yielded 0.874 and 0.883, respectively. The calibration curve for the nomogram closely approximated the optimal diagonal. The decision curve analysis (DCA) indicated that the prediction model offers substantial net benefits. Conclusions. The innovative predictive nomogram, leveraging radiomics and clinical traits of hematoma, presents a potent and noninvasive tool for HE risk stratification. The method of quantifying mean hematoma density holds significant prognostic value in forecasting HE.