| Popis: |
Background: Currently, no large dataset-derived standard has been established for normal or pathologic human cerebral ventricular and cranial vault volumes. Manual image segmentation is impractical and laborious, and most efforts utilize linear or two-dimensional approaches. Automated volumetric measurements could be used to construct normal reference tables and assist in diagnosis and follow-up of hydrocephalus or craniofacial syndromes. In this work we utilize big data and deep learning algorithms to measure ventricular and cranial vault volumes in head computed tomography (CT) scans. Methods: A dataset comprising 13,851 scans was utilized to deploy U-net deep learning networks to segment and quantify lateral cerebral ventricular and cranial vault volumes in relation to age and sex. The models were validated against manual segmentations. The radiological reports of the dataset were annotated using a rule-based natural language processing (NLP) framework to identify normal scans, cerebral atrophy, or hydrocephalus. Findings: U-net models were fast with high fidelity when compared to manual segmentations for both lateral ventricular and cranial vault volume measurements (DICE 0·909 and 0·983, respectively). The NLP identified 6,239 (44·7%) normal radiological reports, 1,827 (13·1%) with cerebral atrophy and 1,185 (8·5%) with hydrocephalus. Age- and sex-based refence tables with medians, 25th and 75th percentiles for scans classified as normal, atrophy and hydrocephalus were constructed. The median lateral ventricular volume in normal scans was statistically significantly smaller compared to hydrocephalus (15·7 mL [11·1, 22·2] vs 82·0 mL [51·1, 126] respectively, P |
