gms | German Medical Science

Objective: Sparse sampling fMRI techniques facilitate language mapping by introducing silent periods for stimulus presentation and response control at the cost of prolonged scanning. Here, clustered acquisition can increase the amount of data collected per time and, thus, sensitivity to activations. We here compared results from an established sparse and a new clustered acquisition method of equal scan durations.

Methods: Fifteen healthy subjects (m=8; mean age 24±3 yrs) underwent task-related language fMRI. MRI data were acquired on a 3 Tesla scanner using a 32 channel head coil (7.4 min session length).

• Sparse Sampling: We applied a sparse acquisition protocol with 12s repetition time (TR) (37 volumes).
• Clustered Sampling: Simultaneous multislice excitation and SENSE were utilised for a total acceleration factor of 3.6 (TR=1.2s). Three volumes were acquired every 12s (111 volumes).
• Functional Task: The picture naming task consisted of overtly naming a picture presented on a screen.
• Analysis: Voxels were considered significant at p≤0.001. Dice coefficient and Simpson coefficient between sparse and clustered data were calculated.

Results: Using clustered acquisition, activation was detected in the inferior frontal gyrus (IFG) in 9 out of 15 subjects (60%). Within the superior temporal gyrus (STG), activation was evident in 14 subjects (93%). Sparse acquisition detected activation in 4 out of 15 subjects within the IFG (27%) and in 6 subjects within the STG (40%). Both methods revealed similar activation patterns. Simpson coefficient between sparse and clustered fMRI was 88% (Dice: 18%).

Conclusion: Clustered acquisition performed better than the sparse protocol when total scan duration was limited. A comparison of the new protocol with the established sparse method in full length and clinical case studies will complement our results.