Introduction:

We have previously shown that the representation of the fingers in an individual subject's contralateral primary somatosensory cortex (S1) can be mapped using a travelling wave paradigm with vibrotactile stimuli^1,2. Here, we generate a probabilistic atlas of digit representations of both hands in standard volumetric and surface space, to provide a reference for future studies in patient groups with disorders such as hand dystonia.

Methods:

Twenty-one right-handed healthy subjects were scanned at 7T (Philips Achieva) with a 32-channel receive coil. fMRI data were acquired using multi-slice gradient echo–echo planar imaging (GE-EPI) with either 1.5 or 1.25mm isotropic resolution (TE 25ms, TR 2s). A ‘travelling wave’ paradigm was used to sequentially stimulate each fingertip of a given hand (vibrotactile stimuli, 30Hz, 4s period). Functional scans consisted of 8-12 cycles, alternating between forward (thumb to little finger) and backward (little finger to thumb) ordering. In each subject, digit somatotopy was repeated for each hand. Additionally, a whole head (1mm isotropic) T1-weighted structural scan was acquired. fMRI data from the forward and backward scans were combined (to cancel the haemodynamic delay) prior to Fourier analysis (mrTools http://gru.stanford.edu/mrTools). Standard phase and coherence maps were generated and coherence maps converted to p-values and used to threshold (p≤0.05, uncorrected) the phase maps (Figure 1). Cortical surfaces of the left and right hemisphere were generated in Freesurfer for each subject’s T1-weighted structural images. Phase maps were projected to individual flattened representations of their sensorimotor cortex for a manual approach to delineate the hand area ROI (Figure 2A). For an automated approach (Figure 2B), Freesurfer labels of Brodmann areas 1, 2, 3a and 3b were projected to individual space, and combined to form an automated somatosensory (S1) ROI mask. For both approaches, individual digit ROIs were formed by dividing the phase map into 5 equally spaced bins. These data in individual subject space were then transformed into standard volumetric (MNI152 2 mm) and surface (fsaverage MNI305) space (Figure 1). Full-Probabilistic-Maps (FPMs) for each digit were generated from the number of times a location belonged to a digit-specific ROI, normalized by the total number of subjects. Maximum-Probabilistic-Maps (MPMs) were generated using a winner-takes-all approach (assigning the digit with highest probability to a given location)3. Group mean and variance phase maps were created by calculating the circular mean and standard-deviation of the phase across subjects. To validate each atlas, a leave-one-out cross validation procedure was performed by computing the central tendency metric³ between digit ROIs in the leave-one-out FPM and digit ROIs defined in individual subjects, this was performed for all pairwise digit comparisons.

Results:

The manual and automated hand ROIs produced comparable probabilistic maps (Figure 2C, 2D). The FPMs show the mediolateral organization of digits D1-D5 in contralateral S1 for stimulation of both hands. The FPMs show the variability of individual digit representations across subjects. Sharp boundaries between digits are more clearly seen in the MPMs (Figure 3A), which are similar to the group averaged phase map (Figure 3B). In these maps, digits D2, D3, D4 show the largest representations, and D1 and D5 the smallest, quantified in Figure 4 by the mean digit ROI size across subjects and the MPM digit ROI size. Figure 4 shows the blurring metric³ (which is inversely proportional to how much ROIs for a given digit overlap across subjects) is higher for D1-D4 of the left hand than right hand, suggesting a higher digit specificity for the dominant hand. For both volumetric and surface based atlases, Figure 5 shows higher central tendency values lie along the diagonal of the matrix, as expected, except for D2 which shows large overlap with D1.

Discussion:

We present a volume and surface based probabilistic atlas of digit somatotopy for both the right (dominant) and left (non-dominant) hand in contralateral S1. This probabilistic atlas is generated from travelling wave phase maps, and thus represents maps according to the most dominant digit across S1. In previous work⁴, we have used an event-related design to reveal the spatial overlap of the digit representations in an individual subjects. Incorporating this overlap into a probabilistic atlas would provide additional useful information about variability of S1 across individuals, and likely higher overlap between subjects.

Conclusion:

We present a probabilistic atlas in standard volume and surface space of right and left hand individual finger representations in contralateral S1 from a large population of right handed subjects. This probabilistic atlas will be used in future investigations to assess how digit representations are altered in patient populations. This digit somatotopy probabilistic atlas will be made freely available.

Acknowledgements

This work was funded by a MRC grant MR/M022722/1 and a Leverhulme fellowship.