Simulations of obstructive sleep apnoea (OSA) require detailed models of the muscle fibres within the tongue, to correctly capture the motion of the tongue. Diffusion weighted images (DWI) of the oral cavity were collected for 5 healthy controls. Fibre-orientation distributions (FOD) estimated from subject DWI, can resolve intersecting muscles in the tongue. The group averaged FOD reduced the influence of spurious peaks, which gave clearer boundaries between the intrinsic muscles of the tongue. These results may help fully automate the segmentation of the muscles within the tongue.