Developing deep learning-based cerebral ventricle auto-segmentation system and clinical application for the evaluation of ventriculomegaly.
Authors
Affiliations (4)
Affiliations (4)
- Department of Neurosurgery, Seoul National University College of Medicine, Seoul National University Hospital, Seoul 03080, Republic of Korea.
- Research and Science Division, Research and Development Center, MEDICALIP Co. Ltd, Seoul 03129, Republic of Korea.
- Neuro-Oncology Clinic, National Cancer Center, Goyang 10408, Republic of Korea. Electronic address: [email protected].
- Department of Neurosurgery, Seoul National University College of Medicine, Seoul National University Hospital, Seoul 03080, Republic of Korea. Electronic address: [email protected].
Abstract
Current methods for evaluating ventriculomegaly, particularly Evans' Index (EI), fail to accurately assess three-dimensional ventricular changes. We developed and validated an automated multi-class segmentation system for precise volumetric assessment, simultaneously segmenting five anatomical classes (ventricles, parenchyma, skull, skin, and hemorrhage) to support future augmented reality (AR)-guided external ventricular drainage (EVD) systems. Using the nnUNet architecture, we trained our model on 288 brain CT scans with diverse pathological conditions and validated it using internal (n=10),external (n=43) and public (n=192) datasets. Clinical validation involved 227 patients who underwent CSF drainage procedures. We compared automated volumetric measurements against traditional EI measurements and actual CSF drainage volumes in surgical cases. The model achieved exceptional performance with a mean Dice similarity coefficient of 93.0% across all five classes, demonstrating consistent performance across institutional and public datasets, with particularly robust ventricle segmentation (92.5%). Clinical validation revealed EI was the strongest single predictor of ventricular volume (adjusted R<sup>2</sup> = 0.430, p < 0.001), though influenced by age, sex, and diagnosis type. Most significantly, in EVD cases, automated volume differences showed remarkable correlation with actual CSF drainage amounts (β = 0.956, adjusted R<sup>2</sup> = 0.936, p < 0.001), validating the system's accuracy in measuring real CSF volume changes. Our comprehensive multi-class segmentation system offers a superior alternative to traditional measurements with potential for non-invasive CSF dynamics monitoring and AR-guided EVD placement.