This prospective study was approved by the ethical commission of the University of Luebeck, Germany (vote: 2023–416). Thirteen volunteers (8 women, 5 men; aged 22.8 ± 1.4 years) were informed about the project and possible risks and provided consent before the examination. The sample size was calculated based on the findings from Taoka et al. [17], who found significant inter-method differences with 7 volunteers. Under the assumptions of a large size effect (r = 0.82, estimated from Fig. 4a in Ref. [17]), significance level α = 0.05 and 80% power (1-β = 0.8), the required sample size was 11.65 participants. Adding 10% to account for non-normal distributed data, approximately 13 volunteers were considered sufficient to confirm significant inter-method differences.

The examination was performed on a 3T MRI scanner (MAGNETOM Vida, Siemens Healthineers, Germany), using a 20-channel head coil in neutral position (without tilt). The imaging parameters for DTI and DWI were: TE/TR = 92/4400 ms, FOV = 200 × 200 mm2, voxel size = 2 × 2 × 2 mm3, and b-values = 0 and 1000 s/mm2. The DTI measurement comprised 64 diffusion directions and DWI had three orthogonal directions in the readout, phase-encoding, and slice-encoding direction. Simultaneous multi-slice imaging with a total acceleration factor of 4 was used. The total durations of the DTI and DWI scans were 5:30 and 1:30 min, respectively.

The DTI images were analyzed using DSI Studio [40], version Apr. 23 2023 (http://dsi-studio.labsolver.org), and the DWI images were analyzed in ImageJ [41], version 1.54g (U.S. National institutes of health, Bethesda, Maryland, USA, https://imagej.nih.gov/ij/). In ImageJ, an image composite was generated with three channels: red, green, and blue for diffusivity along x-, y- and z-directions (in the readout, phase- and slice-selection coordinate system), as described by Taoka et al. [17].

The data were evaluated by a senior scientist (PU), placing the ROIs in a single slice per hemisphere, with ROIs aligned parallel within each hemisphere. Eleven ROIs were placed next to the lateral ventricle in the projection (superior corona radiata) and association (longitudinal fasciculus) areas according to a color-coded fractional anisotropy (FA) map by referring to susceptibility-weighted images to ensure the periventricular vessels were perpendicular to the lateral ventricle.

The ROIs consisted of a single voxel (v1), 2, 3, 4, 6 and 9 voxels (with volumes of 8, 16, 24, 32, 48 and 72 mm3, respectively) and were placed parallel (v2p, v3p, v4p, v6p) and orthogonal (v2o, v3o, v4o, v6o) to the lateral ventricle, and as a 4-voxel and 9-voxel ROIs in a square configuration (v4s and v9s, respectively), as shown in Fig. 1. The 6-voxel ROIs, as they did not fit as a single row or column within the projection and association areas, were arranged in 3 × 2 or 2 × 3 voxel grids for parallel and orthogonal configurations, respectively. For DWI ALPS, we aimed to place the ROIs in the same slice (±1) as for DTI, with visualization of the corpus callosum, superior corona radiata, and longitudinal fasciculus based on the color-coded composite in ImageJ.

A Susceptibility-weighted image showing the position of the projection and association ROIs (in blue and green, respectively) using DSI Studio. B, C ROI configurations placed parallel and orthogonal to the lateral ventricle, respectively. D Squared ROI configurations. For simplicity, only the ROI placement on the left hemisphere is shown

The projection ROIs were placed at least one voxel away from the lateral ventricle to avoid partial volume effect from the cerebrospinal fluid, which could artificially increase the ALPS index, and at least one voxel away from the association areas. Likewise, the association ROIs are at least one voxel away from the projection and subcortical regions.

From the tensor matrix, x-, y-, and z-axis diffusivities were obtained, and the DTI ALPS index was calculated as mean(DxProj,DxAssoc)/mean(DyProj,DzAssoc) [4], where Dx, Dy, and Dz correspond to the diagonal elements of the diffusion tensor. For the DWI ALPS index calculation, diffusivities along the readout (x–), phase- (y–), and slice-encoding (z–) directions were obtained directly from the scan and calculated using the same formula [17].

Statistical analysis was performed using Matlab R2022a with a significance level of α = 0.05. Normality was determined using Lilliefors test. Differences were evaluated using the Wilcoxon signed-rank test.

We investigated interhemispheric differences in the DTI and DWI ALPS indices and between ROI configurations. In addition, we investigated the inter-method agreement (between DTI and DWI ALPS) using intraclass correlation coefficient (ICC) and Bland–Altman analysis. ICC estimates and their 95% confidence intervals (CI) for inter-method agreement were calculated using a two-way random-effects model to assess the agreement in a single rating. ICC values above 0.75 are considered good, values between 0.5 and 0.75 are considered moderate agreement, and values below 0.5 indicate poor reliability. Values above 0.9 are considered excellent agreement.

Bland–Altman analysis was performed to determine the average bias and the 95% limits of agreement between DTI and DWI. Limits of agreement (LOA) were calculated as bias ± 1.96 * SD, where SD is the standard deviation of the difference between paired samples.