Written informed consent was obtained from all participants before examinations. The study protocol was approved by the Hospital Bioethics Committee. Notably, 41 patients with iNPH who were referred to the Department of Neurology at our institution between January 2019 and March 2022, as well as 40 healthy controls, were recruited to the study. All patients fulfilled the criteria of the updated Japanese guidelines for iNPH [13, 14] and were eligible for cerebral spinal fluid (CSF) shunt operation. Figure 1 shows the flowchart of the selection process for the iNPH and NC groups. A neurologist with 15 years of experience in movement disorders performed the clinical assessment according to the guidelines for iNPH. Initially, 98 patients were included in this study based on the following criteria: age > 60 years, presenting with one or more typical triad symptoms (gait disturbance, cognitive disorder, and urinary incontinence), and presence of recent radiological evidence of ventricular enlargement. However, 57 participants were excluded due to secondary or congenital/developmental hydrocephalus (n = 21), absence of CSF shunt operation (n = 7) or CSF pressure exceeding the normal threshold (n = 3), absence of 3-T brain MRI including DTI (n = 23), or image artefacts (n = 3).

Flowchart of selection for the iNPH and NC groups in this study. iNPH, idiopathic normal pressure hydrocephalus; NC, normal control

During the same period, an age- and sex-matched NC group was recruited by advertising in several health management centres, targeting older individuals (age > 60 years) without a definite history of significant medical conditions, such as ischemic or haemorrhagic stroke, trauma, or cognitive disorder. Therefore, 40 age- and sex-matched healthy volunteers were selected for the NC group following a thorough medical history review. Notably, all recruited participants were right-handed and underwent clinical assessments.

All MRI data were obtained using a 3-T MRI scanner (Siemens Prisma). The participants’ heads were immobilised with foam pillows inside the coil to reduce motion artefacts. DTI sets with b = 1000 (echo planar, repetition time (TR) = 9500 ms, echo time (TE) = 90 ms, motion probing gradients = 64 directions, field of view (FOV) = 240 × 240 mm2, matrix = 128 × 128 mm2, slice thickness = 3 mm) were acquired alongside conventional morphological images. The following sequences were obtained using routine protocols: sagittal three-dimensional (3D) T1-weighted imaging (T1-WI) magnetisation-prepared rapid acquisition gradient-echo (MPRAGE) sequence, sagittal 3D fluid-attenuated inversion recovery (FLAIR) imaging with axial reconstruction, axial T2-WI, and susceptibility weighted imaging (SWI). Imaging parameters were as follows: (1) sagittal 3D T1-WI: TR = 2300 ms; TE = 3.55 ms; slice thickness, 0.9 mm; flip angle, 8°; FOV = 240 × 240 mm2; acquisition matrix = 256 × 256 mm2; (2) 3D FLAIR: TR = 4800 ms; TE = 274 ms; slice thickness, 1.5 mm; flip angle, 90°; FOV = 240 × 240 mm2; acquisition matrix = 240 × 240 mm2; (3) axial T2-WI: TR = 3000 ms; TE = 80 ms; slice thickness, 5 mm; flip angle, 90°; FOV = 180 × 230 mm2; acquisition matrix = 420 × 375 mm2; (4) SWI: multi-echo fast-field-echo sequence; TR = 41 ms; total 4 echoes; first TE = 7.2 ms; echo interval, 6.2 ms; slice thickness, 2 mm; flip angle, 20°; FOV = 200 × 220 mm2; acquisition matrix = 384 × 384 mm2.

Diffusion metric images from DTI were processed using FMRIB Software Library (FSL) version 5.0.9 (https://fsl.fmrib.ox.ac.uk/). We computed diffusivity maps in the directions of the x-axis (right-left, Dx), y-axis (anterior-posterior, Dy), and z-axis (inferior-superior, Dz). Fractional anisotropy (FA) maps were created and aligned into the FMRIB58_FA standard space using FSL’s linear and nonlinear registration tools.

Initially, referencing SWI, an axial slice was selected at the level of the lateral ventricular body, where the trans-medullary vessels passed perpendicularly to the ventricle. At this level, the direction of the perivascular space was perpendicular to the ventricular wall (mostly along the x-axis). This direction was also perpendicular to the direction of the projection (mostly along the z-axis) and association fibres (mostly along the y-axis) (Fig. 2). Therefore, the diffusivity along the x-axis in the regions with projection/association fibres at least partly represents the diffusivity along the perivascular space. The areas of these neural fibres were identified on a colour-coded FA map using different FA values. Similar to Taoka et al all measurements were performed in the left hemisphere since all participants were right-handed. However, in contrast to Taoka et al we were concerned about more crossed nerve fibres in the area of the subcortical fibres; therefore, we chose to place two 5-mm-diameter spherical regions of interest (ROIs) in the areas of projection and association neural fibres (Fig. 2). These ROIs were registered using the same FA template. Diffusivity in the direction of the x-axis (Dx), y-axis (Dy), and z-axis (Dz) within the ROIs was obtained for each participant.

Schematic drawing of the diffusivity measurement using the DTI-ALPS methods. a Axial SWI on the slice at the level of the lateral ventricle body indicating that parenchymal vessels run laterally (x-axis). b Colour-coded FA map of DTI showing the distributions of the projection neural fibres running along the z-axis (blue colour) and the association neural fibres along the y-axis (green colour) areas. Two ROIs were placed in the areas with the projection (red circle) and association areas (yellow circle) to measure diffusivities of the three directions (x, y, z) on a colour-coded FA map. c Schematic diagram indicating the relationship between the direction of the perivascular space (grey cylinders) and the directions of the fibres. Note that the direction of the perivascular space is perpendicular to both projection and association fibres. d Two ROIs on colour-coded FA maps were then copied and pasted onto the three diffusivity maps to measure diffusivities along the x-, y-, and z-axes. DTI-ALPS, diffusion tensor image analysis along the perivascular space; SWI, susceptibility-weighted imaging; ROIs, regions of interest; FA, fractional anisotropy

Finally, we calculated the ALPS index using the algorithm by Taoka et al [5]: ALPS index = (mean [Dxproj] + mean [Dxassoc])/(mean [Dyproj] + mean [Dzassoc]). An ALPS index close to 1.0 reflects minimal diffusion along the perivascular space, whereas higher values indicate greater diffusivity. Two trained neuroradiologists blinded to the clinical information of participants independently assessed all diffusion metrics. Interobserver consistency of diffusivities and ALPS indexes was assessed, and all diffusivities and ALPS indexes were then averaged and used for further analysis when consistency was good.

Our research team chose MPRAGE data for manual segmentation for ventricular volumetry. The specific manual delineation of the ventricle volume (VV) and total intracranial volume (ICV) process is as follows: (1) a radiologist with > 5 years of clinical work experience manually marked the VV and ICV used the ITK software (v3.8.0-RC1; http://www.itksnap.org) to label the ventricles; (2) Furthermore, a neurosurgeon with > 10 years of clinical work experience reviewed and adjusted the manual annotation results. The total VV was divided by ICV to obtain normalised VV (VV/ICV). The WMH lesion volume was calculated for each patient from the generated lesion mask using FSLstats, and WMH lesion masks were created on FLAIR images using the lesion prediction algorithm [15] and the Statistical Parametric Mapping–Lesion segmentation toolbox subsequent to lesion identification by an experienced neuroradiologist. After lesion filling using the FSL lesion filling toolbox [16], normalised brain and grey matter volumes were assessed from T1-weighted images using Structural Image Evaluation and Normalization of Atrophy for cross-sectional data, which is part of the FSL [17]. The WMH lesion masks and lesion-filled T1-weighted images were visually checked for accuracy. Similarly, considering individualised differences in ICV, the total WMH volume was divided by ICV to obtain normalised WMH volume (WMH volume/ICV). Furthermore, periventricular white matter hyperintensity (pWMH) and deep white matter hyperintensity (dWMH) were evaluated by an experienced neuroradiologist using the Fazekas scale [18] based on axial FLAIR imaging. pWMH was graded as follows: grade 0, absence; grade 1, caps or pencil-thin lining; grade 2, smooth halo; and grade 3, extension into the deep white matter. dWMH was graded as follows: grade 0, absence; grade 1, punctate foci; grade 2, beginning confluence; and grade 3, large confluence.

All statistical analyses were performed using SPSS software (version 25.0; IBM Corporation). The Shapiro–Wilk test was used to assess normality. Clinical and structural MRI characteristics of patients with iNPH and controls were compared using unpaired t-tests (for normally distributed continuous variables) or the Mann–Whitney U test (for non-normally distributed variables), and differences in frequencies were compared using Chi-square tests. The consistency of the diffusivities and ALPS indexes between two observers was tested with intraclass correlation coefficients.

Receiver operating characteristic (ROC) curve analysis was employed to compare the diagnostic performance of the ALPS index between control and iNPH groups. The correlation between the diffusivities, ALPS index, WMH volume, ventricular volume, and the degree of pWMH and dWMH was determined using Spearman’s correlation coefficient. We performed multivariate linear regression analyses on normalised ventricular and WMH volumes. For all regression analyses, critical confounding factors, such as age and sex, were adjusted. The number of covariates was limited to three due to the small sample size. Statistical significance was set at p < 0.05.