Fully Automated Deep Learning-Based Pipeline for Evans Index Measurement from Raw 3D MRI.

Ventriculomegaly is a key neuroimaging feature of normal pressure hydrocephalus (NPH) and other disorders of cerebrospinal fluid (CSF) dynamics. The Evans Index (EI), defined as the ratio of maximal frontal horn width to maximal inner skull diameter, is a widely used marker of ventricular enlargement. However, manual EI measurement is susceptible to inter-observer variability and depends on accurate alignment to the anterior commissure-posterior commissure (AC-PC) plane, limiting reproducibility in large and multi-center studies. We developed a fully automated deep learning-based pipeline for EI calculation directly from raw T1-weighted MPRAGE MRI scans. The pipeline integrates automated landmark detection using the BrainSignsNet model, rigid AC-PC alignment for orientation standardization, and robust segmentation of the lateral ventricles (LV) and intracranial volume (ICV) using nnU-Net models. Ventricular segmentation is performed with a custom network trained on 1,300 manually annotated scans enriched for hydrocephalus. The Evans Index is then derived automatically from frontal horn width and inner skull diameter measured on the aligned axial slice. Internal validation across the Baltimore Longitudinal Study of Aging, BIOCARD, and Johns Hopkins cohorts demonstrated excellent segmentation performance (Dice coefficient = 0.98). External validation in the PENS trial, including pre- and post-shunt NPH scans, showed strong agreement with expert manual EI measurements (mean bias = 0.0068, mean absolute error = 0.0103, r = 0.96, p < 0.001), with no bias related to age, sex, or ventricular volume. This automated, orientation-standardized approach enables accurate, reproducible, and scalable Evans Index assessment for clinical and research applications, particularly in NPH.