Automated Ventricular and Midline Segmentation in Cranial Ultrasound with Metrology

Real-time ultrasound imaging through sonolucent cranial implants is an emerging modality for post-neurosurgical monitoring of the adult brain, but quantitative interpretation remains challenging due to speckle, attenuation, shadowing, and the difficulty of consistently delineating thin anatomical landmarks. We present a deep learning system developed at Longeviti Neuro Solutions for segmenting key intracranial structures–the ipsilateral and contralateral lateral ventricles and the cranial midline–in coronal-plane adult cranial ultrasound images from patients with Longeviti ClearFit® Acoustic Brain Interface (ABI)TM implants. Our dataset comprises 456 proprietary, de-identified ultrasound frames (JPEG with known pixel spacing) annotated in CVAT with ventricle and midline labels. We benchmark multiple encoder–decoder segmentation architectures and address severe class imbalance via class-weighted optimization, test-time augmentation (horizontal flip with left–right label swapping), and class-specific post-processing to reduce spurious components and improve mask coherence. The best-performing configuration achieves a foreground macro Dice of 0.869 on a held-out test set, with ventricle Dice values above 0.92 and midline Dice of approximately 0.75. Finally, we transform predicted masks into geometry-based metrology by estimating maximal perpendicular ventricle spans and ventricle-to-midline distances, producing standardized measurement overlays suitable for downstream review.