Supplemental Figure 1. Architecture used for each neural network inversion.
Supplemental Figure 2. Flowchart summarizing the processing steps for in vivo data.
Supplemental Figure 3. Example brain simulation before adding noise.
Supplemental Figure 4. Summary test set results. Each plot shows the correlation between the estimated stiffness (left column) or damping ratio (right column), either by direct inversion (DI, top row) or neural network inversion (NNI, bottom row), and the true stiffness in a test set of 100,000 examples. The color of each point in the scatter plot indicates the mean signal-to-noise ratio (SNR) for that example.
Supplemental Figure 5. Maps of the estimated change in stiffness (Δµ) due to NPH with respect to both the CN and AD groups. The difference map is overlaid on the mean CN stiffness map, with color representing the magnitude of the stiffness change. Only clusters meeting a family- wise error corrected P<0.025 are shown (primary thresholds of P<0.01 on F-test for group and P<0.01 on T-test for pair-wise comparison of interest with minimum cluster size of 351 voxels).
Supplemental Figure 6. Maps of the estimated change in damping ratio (Δζ) due to NPH with respect to both the CN and AD groups. The difference map is overlaid on the mean CN stiffness map, with color representing the magnitude of the stiffness change. Only clusters meeting a family-wise error corrected P<0.025 are shown (primary thresholds of P<0.01 on F-test for group and P<0.01 on T-test for pair-wise comparison of interest with minimum cluster size of 263 voxels).
Supplemental Figure 7. Simulation-based estimate of the potential confounding effect of group- wise differences in brain morphology on estimated differences in stiffness using the direct inversion algorithm. The top 3 rows show the mean stiffness maps after age- and sex-correction (ideal solution is 2.33 kPa at all voxels). The bottom 2 rows show the estimated group-wise effects comparing NPH to each of the other two groups (ideal solution is 0 kPa at all voxels).
Supplemental Figure 8. Simulation-based estimate of the potential confounding effect of group- wise differences in brain morphology on estimated differences in damping ratio using the direct inversion algorithm. The top 3 rows show the mean stiffness maps after age- and sex-correction (ideal solution is 0.22 at all voxels). The bottom 2 rows show the estimated group-wise effects comparing NPH to each of the other two groups (ideal solution is 0 at all voxels).
Supplemental Figure 9. Simulation-based estimate of the potential confounding effect of group- wise differences in brain morphology on estimated differences in stiffness using the neural network inversion algorithm. The top 3 rows show the mean stiffness maps after age- and sex- correction (ideal solution is 2.33 kPa at all voxels). The bottom 2 rows show the estimated group-wise effects comparing NPH to each of the other two groups (ideal solution is 0 kPa at all voxels).
Supplemental Figure 10. Simulation-based estimate of the potential confounding effect of group-wise differences in brain morphology on estimated differences in damping ratio using the neural network inversion algorithm. The top 3 rows show the mean stiffness maps after age- and sex-correction (ideal solution is 0.22 at all voxels). The bottom 2 rows show the estimated group- wise effects comparing NPH to each of the other two groups (ideal solution is 0 at all voxels).
Supplemental Figure 11. Histograms summarizing the voxel-wise distribution of values in each of the group-wise difference maps shown in Supplementary Figures 6-9. The ideal solution is 0