Learning sculpts microstructure in real time: evidence from dense temporal sampling of diffusion MRI

Description: Learning induces rapid microstructural plasticity in the human brain, yet the precise temporal dynamics of these changes remain unclear. Using dense temporal sampling of diffusion-weighted MRI (DW-MRI) combined with task-based fMRI, we assessed microstructural changes throughout a declarative learning paradigm and subsequent rest. Seventy-four participants (36 females) learned image-location associations across four encoding-retrieval repetitions while undergoing interleaved functional and DW-MRI acquisitions. A matched control group (N=37, 21 females) underwent a similar imaging protocol without learning. Dense sampling of DW-MRI acquisitions (k=2146 across 22 time points in 127 min) revealed that learning-induced mean diffusivity (MD) decreases emerged shortly after learning onset and continued to develop during post-learning rest. The most robust and spatially consistent change was localized to the left middle occipital/temporal gyrus, a region also showing functional activation during encoding and retrieval. Linear mixed-effects modeling further confirmed a significant group-by-time interaction, with MD reductions in the left middle occipital/temporal gyrus emerging as early as ≈7 min after learning onset, becoming robust by ≈35-40 min, and persisting throughout the extended post-learning period, while controls showed no changes. Our findings demonstrate that learning-related microstructural plasticity unfolds continuously from encoding to offline consolidation, with learning-induced structural changes emerging in functionally engaged regions. Dense temporal sampling of DW-MRI offers a powerful approach to bridge functional activation and structural remodeling, providing evidence of when and where experience-dependent plasticity occurs during memory formation.