Brain white matter (WM) could be generally categorized into two types, deep and superficial WM. Studies combining these two types WM are important for a better understanding of brain plasticity induced by motor training. In this study, we applied both univariate and multivariate approaches to study gymnastic training-induced plasticity in brain WM. Specifically, we acquired diffusion tensor imaging data from 13 world class gymnasts and 14 non-athlete normal controls, reconstructed brain deep and superficial WM tracts, estimated and compared their fractional anisotropy (FA) difference between the two groups. Taking FA values as the features, we applied logistic regression and support vector machine to distinguish the gymnasts from the controls. Compared to the controls, the gymnasts showed lower FA in four regional deep WM tracts, including the occipital lobe portion of left inferior fronto-occipital fasciculus (IFOF.L), occipital and temporal lobe portion of right inferior longitudinal fasciculus (ILF.R), insular cortex portion of right uncinate fasciculus (UF.R), and parietal lobe portion of right arcuate fasciculus (AF.R). Meanwhile, we found lower FA in the superficial U-shaped tracts within the frontal lobe in the gymnasts compared to the controls. In addition, we detected that mean FA in either the AF.R or the U-shaped tracts connecting the left pars triangularis and superior frontal gyrus was negatively correlated with years of training in the gymnasts. Classification analyses indicated FA in deep WM hold higher potential to distinguish the gymnasts from the controls. Overall, our findings provide a more complete picture of training-induced plasticity in brain WM.
Keywords: Logistic regression; Neuroplasticity; Support vector machine (SVM); Tractography.