A ventriculomegaly feature computational pipeline to improve the screening of normal pressure hydrocephalus on CT.

Objective: The objective of this study was to develop a computational pipeline that extracts objective features of ventriculomegaly from non-contrast CT (NCCT) for the accurate classification of idiopathic normal pressure hydrocephalus (NPH) from headache controls (HCs), Alzheimer's dementia (AD), and posttraumatic encephalomalacia (PTE).

Methods: Patients with possible NPH (n = 79) and a subset with definite NPH (DefNPH; n = 29) were retrospectively identified in the Veterans Affairs Informatics and Computing Infrastructure system, along with the AD (n = 62), PTE (n = 53), and HC (n = 59) cohorts. Image-processing pipelines were developed to extract a novel feature capturing the maximum eccentricity of the lateral ventricles (MaxEccLV), a proxy splenial angle (p-SA), the Evans indices (EI-x, -y, and -z), callosal angle, normalized maximum third-ventricle width, and CSF to brain volume ratio from their NCCT scans. The authors used t-tests to examine group differences in the features and multivariate logistic regression models for classification. Additionally, the NPH versus HC classifier was validated on external data.

Results: When NPH and DefNPH were compared with HC, AD, and PTE, significant differences were found in all features except the p-SA, which only significantly differed between NPH and PTE. The test-set area under the receiver operating characteristic curve (AUC), sensitivity, and specificity were 0.98, 100%, and 98.3% for NPH versus HC classification; 0.94, 87.3%, and 85.5% for NPH versus AD; 0.96, 92.4%, and 90.6% for NPH versus PTE; and 0.96, 94%, and 88% for NPH versus the other groups using logistic regression under five-fold cross-validation. Consistently high performance was noted for DefNPH. The NPH versus HC classifier provided an AUC of 0.84, sensitivity of 76.9%, and specificity of 90% when assessed on external data.

Conclusions: Including the novel MaxEccLV, this framework computes useful features of ventriculomegaly, which had not previously been algorithmically assessed on NCCT. This framework successfully classified possible and definite NPH from HC, AD, and PTE. Following validation on larger representative cohorts, this objective and accessible tool may aid in screening for NPH and differentiating it from symptomatic mimics such as AD and PTE.