Assessing CT-based volumetric analysis via deep learning for idiopathic normal pressure hydrocephalus.

Initially, a 2D U-Net model was trained to predict ventricular cerebrospinal fluid segmentations from computed tomography scans, using paired MR-ventricular cerebrospinal fluid labels from healthy controls.

The training dataset included 734 computed tomography datasets from healthy controls paired with T1-weighted magnetic resonance imaging scans from the Gothenburg H70 Birth Cohort Studies and 62 computed tomography scans from iNPH patients at Uppsala University Hospital.