Automated Brain and CSF Volume Assessment in Infant Hydrocephalus Using Deep Learning

Abstract Accurate brain and cerebrospinal fluid (CSF) volume assessment is essential for pediatric hydrocephalus management. Current clinical practice relies on linear measurements that fail to capture complex three-dimensional ventricular morphology, while quantitative volumetric assessment remains limited by laborious processing and lack of clinically optimized automated tools. This study developed a rapid, automated AI-based intracranial segmentation model suitable for clinical workflows. We retrospectively analyzed 167 T2-weighted MRI scans from infants with hydrocephalus, randomly split into training (60%), validation (20%), and hold-out test (20%) sets. All scans were manually segmented into CSF, brain parenchyma, and background. Our model integrates DenseNet and U-Net architectures with feature smoothness regularization to enhance generalizability. Performance was evaluated using Dice scores and absolute relative volume error (ARVE) compared with state-of-the-art methods. The AI model achieved Dice scores of 95.7% for CSF and 96.4% for brain parenchyma on the hold-out test set, significantly outperforming FSL FAST (85.0% and 77.9%) and contemporary deep learning approaches (90.4% and 89.7%). Processing time was 0.8 seconds per scan using GPU acceleration. The model demonstrated consistent performance across different hydrocephalus etiologies and effectively handled challenging scenarios including noise, artifacts, and variable resolution. This study successfully developed a robust MRI segmentation model demonstrating superior accuracy and efficiency compared to existing methods. By incorporating domain-specific enhancements, the model enables rapid, clinically viable brain and CSF volume estimation for pediatric hydrocephalus care.