Effects of Inter-stimulus Interval on the Analysis of Color Representation in Human Visual Cortex in Functional MRI

Abstract

Multi-variate pattern analysis (MVPA) of fMRI signal has been used to investigate cortical mechanisms of human vision. The performance of MVPA is generally more accurate when the number of trials is larger, but the use of a short inter-stimulus interval (ISI) to obtain more trials would cause significant overlap of fMRI responses to multiple events and the isolation of each response could worsen. We investigated this issue using color stimuli (12 hues) presented under shorter- (around 6 s) and longer- (around 15 s) ISI conditions under the same total imaging time. In the waveform analysis using general linear model and deconvolution methods, the goodness-of-fit of mean fMRI response was better in long ISI condition. The positive response trials were found more under long ISI, while negative response trials were found more under short-ISI condition. Although performances of classification into 12 classes with a linear support vector machine were statistically significant above the chance level for both ISI conditions, short ISI condition showed higher performance in early visual areas; this difference disappeared when classified into 5 (i.e., categorical) classes. It implies the presence of different cortical representations across visual areas under same task, together with the advantage of short ISI in the MVPA studies.