bc047286b224b7bfa73d4cb02de1238d-AuthorFeedback.pdf

The9 model is not trained in an online manner. We train one regressor per held-out volume ( 2.2). BothPatch2SelfandMarchenko-12 Pastur (MP) are patch based algorithms, assembling voxels from a patch across volumes into a matrix. Values close to zero are not plausible in a healthy brain and can28 therefore confound DKI analyses, classifying that voxelasdisease/ abnormal (asinthecase oftissue degeneration).29 As pointed out by the arrows, the number of these degenerate (black) voxels havebeen reduced inthe data.