Developing an automated tractography model for mapping the Superior Longitudinal Fasciculus: a diffusion MRI study

Background & Aims

Diffusion Magnetic Resonance Imaging ( ) is a magnetic resonance imaging technique which uses the random motion of water molecules to create contrast between tissues. Structures such as axons permit water to diffuse along the axon more readily than across it. This allows us to map white matter connections in the living human brain using a computational method called probabilistic tractography. However, many tractography methods, such as of interest (ROI) analysis, are time-consuming and advanced neuroanatomical knowledge is required. We aimed to reduce these limitations by developing and validating an automated model for mapping the superior longitudinal fasciculus (SLF). SLF is a bundle of fibres in each hemisphere of the brain. It connects the ipsilateral frontal cortex with the parietal, occipital and temporal regions, and is involved in core cognitive processes including language, attention, memory and emotion.


MRI data from young healthy participants (N = 30, age ~24) were taken from a large longitudinal study, the Avon Longitudinal Study of Parents and Children. MRI data acquired using a GE HDx 3T scanner at Cardiff University Brain Research Imaging Centre. Data preparation and quality assessment followed in-house procedures. We used ExploreDTI to manually dissect SLF bundles. Two ROIs were applied to delimit each fibre tract: a “SEED” and an “AND” operation to include fibres passing through both ROIs. The resulting manual tractography was then used as training data for an in-house automated tractography programme able to streamline patterns through principal component analysis. The diffusion metrics from the in-house tractography model (using 15 and 30 examples) was compared to manual tractography using a Wilcox Signed Rank test and a Spearman’s Rho correlation. The spatial pattern of the tract images was assessed using a Dice coefficient score

Result & Discussion

Results will be presented and discussed.