Predicting intraventricular hemorrhage growth with a machine learning-based, radiomics-clinical model