Practical scan-length considerations for mapping upper limb movements to the somatosensory/motor cortex at 7T: A pilot study

Abstract

The relationship between motor cortex (M1) and upper limb movements has been investigated extensively using functional MRI (fMRI). While most research has focused on applications, very few studies have focused on practical aspects of developing the fMRI protocol. Thus, the effect of scan length on M1 activations during various upper limb movements remains unclear. Scan length constraints are important for conducting motor experiments within a 60- or 90-minute scan session. We targeted this gap in the literature in this pilot study by investigating 7T fMRI activations in a male participant while performing eight different upper limb movements (of the fingers, wrist, and elbow) across 16 task runs (8 with the left arm, 8 with the right arm, 88 minutes total fMRI duration). Standard activation analyses were performed (Z > 3.1, p < 0.01, cluster thresholded) independently for 14 different cases (2 runs through 8 runs, left and right arm) and subsequently compared. We found diminishing returns, presented as activations gradually plateauing, with higher number of runs. We observed two broad categories of movements, one with generally higher activation (more activated voxels and higher Z-stats) and the other with lower activation. To achieve similar statistical power, movements with lower activation required longer scanning (more runs). Based on these observations, we propose a 'one size does not fit all' practical protocol within a 60-, 75-, or 90- minute scan session, wherein different numbers of runs are assigned for different movements. We validated the 75-minute protocol using seven separate scans (N = 3). Our study could benefit researchers who are designing upper limb fMRI experiments.

Keywords: Activation; Fingers and thumb; Functional MRI; Somatotopic mapping; Ultra-high field; Wrist and elbow.

Fig. 1

Activations with different upper limb movements (taken across different numbers of runs). (a) Boxplot of number of activated voxels; (b) boxplot of Z-stats in activated voxels; (c) bar graph of Z-stat's 5th percentile, 95th percentile and the median of top 5% values; (d) dice coefficient (similarity) between different limb movements' activations. The diagonal values are forced to zero.

Fig. 2

Activations with different numbers of runs (taken across different limb movements). (a) Boxplot of number of activated voxels; (b) boxplot of Z-stats in activated voxels; (c) bar graph of Z-stat's 5th percentile, 95th percentile and the median of top 5% values; (d) dice coefficient (similarity) between activations with different number of runs. Diagonal values are forced to zero.

Fig. 3

Validation of our primary result using group-level activations in amputees with different upper limb movements. (a) Number of activated voxels; (b) boxplot of Z-stats in activated voxels; (c) bar graph of Z-stat's 5th percentile, 95th percentile and the median of top 5% values; (d) dice coefficient (similarity) between different limb movements' activations.