Abstract
Our goal was to assess the increase in sensitivity afforded by a new Philips 32-channel head coil for decoding brain states associated with viewing natural images. Data collected with the 32-channel coil were compared to a dataset collected previously using an 8-channel coil. The stimuli for the study included images of 6 animal species including 2 primates, 2 birds, and 2 insects. We tested three scanning protocols to assess trade offs between signal-to-noise ratio (SNR), voxel size, and brain coverage using multivariate pattern (MVP) classification and similarity analysis as dependent measures. Condition 1 (3 mm isotropic voxels, SENSE 2) provided the largest SNR and the most brain coverage (105 mm inferior to superior, I-S), but the lowest resolution. Condition 2 (2 mm, SENSE 2) provided better resolution, but smaller SNR and the least brain coverage (50 mm I-S). Condition 3 (2 mm, SENSE 3) provided better brain coverage (66 mm I-S) than condition 2, but further decreased the SNR. All three protocols used TR of 2 sec, which matched that in the original study (8-channel coil; 3 mm, SENSE 2). We assess results within anatomical masks in ventral temporal (VT) and early visual cortex (EV). Classification results in both VT and EV show that condition 3 performs worse than the other conditions, most likely due to decreased SNR associated with higher SENSE. Classification measures in VT indicate comparable performance for conditions 1 and 2, while in EV there is an appreciable boost in performance for condition 2, which is maximized at lower levels of spatial smoothing. This finding is consistent with the characteristic of the 32-channel coil that it provides higher signal especially at the periphery of the scanning volume. Thus studies targeting peripheral brain structures stand to benefit most from the increased signal provided by the 32-channel coil.