Segmentation and linear registration for TSC. (A) The low-resolution MPRAGE images with a voxel size of 1.5 mm³ were segmented into GM, WM and CSF masks with SPM. Our in-house software regridded these masks to the ²³Na image resolution of 6 mm isotropic and convolved them with the point spread function. (B) For the whole brain tissue, we used the measured TSC map and the tissue volume fraction to solve the overdetermined equation using a linear regression approach as described in the Materials and methods section. (C) Top: the modelled image intensity distribution in the same slice as the measured TSC map, reconstructed using the global least-squares WM, GM and CSF concentrations. Bottom: a residual error map of the absolute value difference between measured image intensity and modelled TSC map. Note the small residual values for WM and GM demonstrating the quality of the global concentration assumption. (D) A normal probability plot of the error residuals from the entire brain tissue of the same subject. Most voxels lie on the red dashed line indicating a normal distribution, which points to a normal distribution of the residuals.