Improved myelin water imaging using B correction and data-driven global feature extraction: Application on people with MS.

The predominant technique for quantifying myelin content in the white matter is multicompartment analysis of MRI's Trelaxation times ( analysis). The process of resolving the Tspectrum at each voxel, however, is highly ill-posed and remarkably susceptible to noise and to inhomogeneities of the transmit field ( ). To address these challenges, we employ a preprocessing stage wherein a spatially global data-driven analysis of the tissue is performed to identify a set of mcTconfigurations () that best describe the tissue under investigation, followed by using this basis set to analyze the signal in each voxel. This procedure is complemented by a new algorithm for correcting B inhomogeneities, lending the overall fitting process with improved robustness and reproducibility. Successful validations are presented using numerical and physical phantoms vs. ground truth, showcasing superior fitting accuracy and precision compared with conventional (non-data-driven) fitting.application of the technique is presented on 26 healthy subjects and 29 people living with multiple sclerosis (), revealing substantial reduction in myelin content within normal-appearing white matter regions of people with MS (i.e., outside obvious lesions), and confirming the potential of data-driven myelin values as a radiological biomarker for MS.