Disproportionately elevated sulcal index (DESI): An automatically driven index representing disproportionate subarachnoid space enlargement in brain MRI scans

Background: Idiopathic normal pressure hydrocephalus (iNPH) is a reversible cause of dementia in older adults yet remains underdiagnosed due to nonspecific clinical presentation and reliance on invasive confirmatory tests. Conventional MRI biomarkers, including the Evans index and callosal angle, capture ventricular enlargement but incompletely characterize cerebrospinal fluid (CSF) redistribution. Disproportionately enlarged subarachnoid-space hydrocephalus (DESH) features, tight high-convexity sulci and enlarged Sylvian fissures, are more disease-specific but lack standardized quantitative measures. Methods: We developed the Disproportionately Elevated Sulcal Index (DESI), an automated volumetric biomarker quantifying the ratio of Sylvian fissure volume to superior sulcal space volume. T1-weighted MRI scans were acquired from three independent cohorts: the Baltimore Longitudinal Study of Aging (n = 1,032), Johns Hopkins CSF Disorders Clinic (n = 216), and the Placebo-controlled Efficacy of iNPH Shunting trial (n = 94). Preprocessing included N4 bias correction, standard-space registration, and alignment. A U-Net segmentation model with an EfficientNet-B0 encoder delineated Sylvian fissures and sulcal compartments. DESI was computed from 3D reconstructions and evaluated using Dice similarity coefficients, landmark detection error, and classification performance across diagnostic groups. Results: Segmentation achieved ~0.80 DICE for Sylvian fissures; ~0.74 for superior sulci. On external validation, DESI discriminated NPH with DESH from non-DESH NPH with AUC = 0.99, accuracy 97.9%, and from all other groups (healthy controls, Alzheimer's disease, vascular dementia) with perfect classification (AUC = 1.00, 100% sensitivity and specificity at threshold = 4.83). In cognitively normal adults, DESI showed only a modest age-related increase and no sex differences, confirming specificity for hydrocephalic pathology. Conclusion: DESI is a robust, fully automated MRI biomarker that sensitively detects CSF redistribution characteristic of DESH in iNPH. By outperforming traditional indices and achieving near-perfect diagnostic accuracy across independent cohorts, DESI offers a noninvasive tool for early detection, patient selection for shunting, and longitudinal monitoring of disease progression.