Very few longitudinal studies have investigated white matter changes associated with motor-skill learning. Of those reported, most have focussed on visuo-motor regions and utilised voxel-based morphometry or tract-based spatial statistics.¹ Change has been reported near the grey-/white-matter interface, but these analyses are reliant on accurate registration, provide little certainty as to the functional relevance of an area, and can have low statistical power. Parcellation-seeded tractography can provide another way to seed specific tracts, but will include fibres innervating functionally-irrelevant muscles, and so may be too insensitive to detect change. We sought to measure neuroplastic change, in subjects who learned a novel non-visual motor task, by using fMRI-guided diffusion MRI.T1 (MPRAGE), HARDI (64 directions; b=3000s/mm2), and task-based fMRI images were collected from 23 healthy adults immediately before and after four weeks of practicing a (10min/day) finger-thumb opposition task with their non-dominant hand. Subjects were processed individually. For each brain, a mesh of the grey-matter/white-matter interface was created from T1 tissue-segmentations. Functional MRI analyses were performed entirely on the mesh, without reslicing, to avoid implicit or explicit cross-sulcal smoothing that could affect tractography. An 8mm FWHM surface smoothing was used. Learned and unlearned finger-thumb opposition-sequence blocks were pooled and contrasted with rest blocks. T-value meshes were then moved into dMRI space and triangles of three connected statistically-significantly activated (p<0.05 FWE) nodes were used to define seeding regions for tractography (Figure 1 & 2). Activation outside of the primary sensorimotor cortex (S1M1) was discarded. Tractography was restricted to the thalamocortical and corticospinal tracts with manually drawn inclusion masks of the thalamus, posterior limb of the internal capsule, and brain stem. The mesh surface was used to proactively constrain tractography to white matter. As corticomotor and thalamocortical tracts often can pass through the same voxels near the midbrain, k-means clustering was used to classify each track as either corticomotor or thalamocortical, based on node locations near these ROIs. Fractional anisotropy (FA) and mean diffusivity (MD) values were calculated for each voxel. To assess diffusion metrics, for each tract, voxels passed through were classified into seven bins, based on their mean proportional distance along the tracts (Figure 3). Voxels with an FA below 0.2 were excluded. A GLM was then used to determine whether change of FA or MD occurred in each bin between sessions, controlling for subject. Each voxel’s contribution to this linear model was weighted by the proportion of tracks that passed through it.Sequence performance rates were similar before training (p=0.34; Paired TTest). After training, the trained sequence was performed 40% faster than the untrained-sequence (p<0.0001), and performance for both sequences improved (both p<0.01). fMRI activation was consistently in the hand area of S1M1 (Figure 1). Bins were consistently sized and located between sessions and subjects. FA of the corticomotor tract increased significantly for every bin by 1.8 – 4.7% (all p<0.05, Holm-Bonferroni adjusted for multiple comparisons; Figure 4). MD of the corticomotor tract decreased in six of seven bins by -0.15% to -1.75%, increasing only near the internal capsule (0.58%), but these changes did not consistently reach statistically significance. No consistent trend of change in FA or MD was seen for the thalamocortical tract.This study has demonstrated white-matter changes induced by motor learning, by utilising a novel data processing pipeline that allows diffusion metrics of sensorimotor tracts to be compared between both subjects and time points. This method does not rely on voxel-perfect registration. This is important, as both type-I and type-II errors might otherwise be introduced through reslicing and/or registration error, due to the minute changes that are being measured. We found that FA change in motor tracts can take place with motor learning, and found no consistent change in thalamocortical tracks. Unlike previous studies, our fMRI-guided method provides confidence that the areas of change are functionally relevant. We also demonstrated that this change occurs across the full length of the motor tract and so is unlikely to be due to partial volume contamination by grey matter (registration error). This suggests that learning is not purely due to highly-local rewiring of the cortex; connections to the alpha motor neurons themselves may strengthen. It is known that electrical impulses can induce myelination in vivo within a two week period.² Given the pattern of change observed after four weeks of motor training, we cautiously speculate that activity-induced myelination of the white matter tracts may have taken place. Future studies utilising myelin-water imaging may ascertain whether this mechanism plays a role in motor learning.