Quantifying "Normal Pressure Hydrocephalus"-like Brain Morphometry patterns in an Alzheimer's Disease Cohort
Lina Bacha1, Alessandra Griffa2, Felix T Kurz2
1From the Swiss Innovation Hub (L.B., B.M., J.D., T.D.N.), Siemens Healthineers International AG, Lausanne, Switzerland; Department of Radiology (L.B., J.-P.T., B.M., J.D., T.D.N.), Leenaards Memory Center, Department of Clinical Neurosciences (A.G.), Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland; Signal Processing Lab 5 (L.B., J.-P.T., B.M., J.D., T.D.N., G.A.), Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland; Medical Image Processing Laboratory (A.G.), École Polytechnique Fédérale De Lausanne (EPFL), Lausanne, Switzerland; Division of Neuroradiology (F.T.K., I.H., K.B., K.-O.L.), Geneva University Hospitals, Geneva, Switzerland; Faculty of Medicine (F.T.K.), University of Geneva, Geneva, Switzerland and SiemensHealthcare Pvt. Ltd. (P.B.V., K.M.P.), Bangalore, India. lina.bacha@siemens-healthineers.com.
Background And Purpose:
Idiopathic normal pressure hydrocephalus (iNPH) is largely underdiagnosed. It is a potentially treatable contributor to cognitive decline, and cognitive improvement may occur even with comorbid Alzheimer's disease. However, iNPH-related imaging features can overlap with age-related and neurodegenerative atrophy, making differentiation difficult. Automated morphometric profiling of NPH-like patterns in biomarker-defined Alzheimer's disease cohorts has not been systematically explored. This pilot study aimed to quantify iNPH-like brain morphometry in an Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort using deep-learning-based morphometry and automated callosal angle measurement.
Materials And Methods:
We included N=30 amyloid-negative reference iNPH patients and N=1003 ADNI participants with structural T1-weighted MRI and CSF biomarkers. Age- and sex-adjusted volumetric z-scores were obtained using a deep-learning-based morphometry algorithm. We derived an iNPH morphometric profile using principal component analysis and Gaussian modeling of the reference iNPH cohort and applied it to the ADNI participants to compute individual iNPH similarity scores. We also automated callosal angle measurement for further clinical representation. We compared scores across cognitive stages and biological Alzheimer's disease groups and evaluated against expert Radscale review.
Results:
iNPH-like morphometry increased with cognitive impairment, with mean differences of 0.03 between cognitively normal and mild cognitive impairment (p=0.06), 0.06 between mild cognitive impairment and dementia (p<0.001), and 0.09 between cognitively normal and dementia groups (p<0.001). Among cognitively impaired participants, scores were higher in biologically-defined Alzheimer's disease patients than biologically-defined non-Alzheimer's disease patients (mean difference=0.08, p<0.001). The iNPH similarity score correlated most strongly with ventricular enlargement features. Automated callosal angle measurements showed high agreement with expert ratings (Gwet's AC1 = 0.939) and angles below 90° occurred in 1.5% of ADNI participants.
Conclusions:
Automated normative morphometric profiling and callosal angle quantification identified heterogeneous iNPH-like structural patterns in the ADNI cohort. These tools provide a scalable approach for characterizing iNPH-like imaging features in dementia cohorts where identifying a potentially treatable contributor to cognitive impairment remains relevant. Further clinical validation remains necessary.
