Redefining Idiopathic Normal Pressure Hydrocephalus Using AI-Driven Brain Volumetry

Idiopathic normal pressure hydrocephalus (iNPH) is a potentially reversible cause of gait disturbance and cognitive impairment in older adults, yet its diagnosis remains challenging and controversial. The core difficulty lies in distinguishing true hydrocephalus from ventricular enlargement secondary to cerebral atrophy or neurodegenerative disease, a distinction now recognized as non-binary. In many patients, ventricular enlargement reflects a continuum ranging from predominantly hydrocephalic iNPH to mixed pathological states combining impaired cerebrospinal fluid (CSF) dynamics and neurodegeneration. Conventional neuroradiological markers, including the Evans Index, the callosal angle, and the disproportionately enlarged subarachnoid-space hydrocephalus (DESH) pattern, provide useful qualitative guidance but are limited by their two-dimensional nature, interobserver variability, and poor sensitivity for differential diagnosis and outcome prediction. Over the past decade, advances in artificial intelligence-based brain volumetry (AI-BrV) have introduced a new paradigm for quantitative structural assessment. By enabling automated, anatomically precise, and reproducible three-dimensional quantification of ventricular and extraventricular CSF, cortical and subcortical gray matter, deep gray matter nuclei, and periventricular white matter, AI-BrV addresses many limitations of traditional imaging approaches. Beyond absolute volume measurements, AI-BrV enables the derivation of composite indices and ratios that may capture disease-specific structural phenotypes and better reflect the underlying pathophysiology of ventricular enlargement. Importantly, AI-BrV pipelines can be applied retrospectively to large legacy neuroimaging datasets and compared with extensive publicly available repositories, facilitating normative modeling, cross-disease analyses, and external validation of volumetric biomarkers. When integrated with clinical data and multivariable statistical or machine-learning frameworks, these approaches hold promise for improving patient selection, refining disease categorization, and supporting more rational decision-making regarding CSF diversion. In this context, AI-BrV offers a unifying framework for reconciling divergent clinical perspectives and advancing iNPH toward a more precise, reproducible, and evidence-based diagnostic and therapeutic paradigm.