Quantifying learning-dependent changes in the brain: Single-trial multivoxel pattern analysis requires slow event-related fMRI

Abstract

Single-trial analysis is particularly useful for assessing cognitive processes that are intrinsically dynamic, such as learning. Studying these processes with fMRI is problematic, as the low signal-to-noise ratio of fMRI requires the averaging over multiple trials, obscuring trial-by-trial changes in neural activation. The superior sensitivity of multivoxel pattern analysis over univariate analyses has opened up new possibilities for single-trial analysis, but this may require different fMRI designs. Here, we measured fMRI and pupil dilation responses during discriminant aversive conditioning, to assess associative learning in a trial-by-trial manner. The impact of design choices was examined by varying trial spacing and trial order in a series of five experiments (total n = 66), while keeping stimulus duration constant (4.5 s). Our outcome measure was the change in similarity between neural response patterns related to two consecutive presentations of the same stimulus (within-stimulus) and between patterns related to pairs of different stimuli (between-stimulus) that shared a specific outcome (electric stimulation vs. no consequence). This trial-by-trial similarity analysis revealed clear single-trial learning curves in conditions with intermediate (8.1–12.6 s) and long (16.5–18.4 s) intervals, with effects being strongest in designs with long intervals and counterbalanced stimulus presentation. No learning curves were observed in designs with shorter intervals (1.6–6.1 s), indicating that rapid event-related designs—at present, the most common designs in fMRI research—are not suited for single-trial pattern analysis. These findings emphasize the importance of deciding on the type of analysis prior to data collection.

from #MedicinebyAlexandrosSfakianakis via xlomafota13 on Inoreader http://ift.tt/1TvBcbO
via IFTTT