A data-driven paradigm for myelin water imaging – validation on multicompartment phantom and multiple sclerosis patient

Multicomponent (MC) T2 analysis is a common technique for probing sub-voxel compartmentation in myelinated tissues. However, the task of resolving the T2 spectra from a single voxel is ill-posed and highly sensitive to noise. Applying a spatially-global MC analysis of the tissue prior to the voxel-wise analysis promotes more stable solutions. Specifically, this approach identifies a specific set of MC T2 features which are then used for fitting the signal in each voxel. Preliminary results suggest that the new data-driven approach can accurately estimate myelin content and correctly estimate myelin content in healthy tissue and in multiple sclerosis lesions.