The Neural Dynamics of Visual Processing in Monkey Extrastriate Cortex: A Comparison between Univariate and Multivariate Techniques

Abstract

Abstract. Understanding the brain mechanisms underlying invariant visual recognition has remained a central tenet of cognitive neuroscience. Much of our current understanding of this process is based on knowledge gained from visual areas studied individually. Previous electrophysiology studies have emphasized the role of the ventral stream of the visual cortex in shape processing and, in particular, of higher level visual areas in encoding abstract category information. Surprisingly, relatively little is known about the precise dynamics of visual processing along the ventral stream of the visual cortex. Here we recorded intracranial field potentials (IFPs) from multiple intermediate areas of the ventral stream of the visual cortex in two behaving monkeys engaged in a rapid face categorization task. Using multivariate pattern analysis (MVPA) techniques, we quantified at millisecond precision the face category information conveyed by IFPs in areas of the ventral stream. We further investigate the relationship between the selectivity and latency of individual electrodes as estimated with classical univariate vs. multivariate techniques and conclude on the similarity and differences between the two approaches.

How emotional expressions motivate action

“…In accordance with studies in monkeys (Cauchoix et al, 2012), it has been shown that faces can be discriminated from scrambled images within 80ms in magnetoencephalography sensors over the occipito-temporal cortex in humans (Dima et al, 2018). Angry versus neutral expressions are differentiated soon after (at 93ms) in the same sensor space and localized to the calcarine, lingual and inferior occipital cortices (Dima et al, 2018).…”

Fast ventral stream neural activity enables rapid visual categorization

“…We trained and tested a linear Support Vector Machine (SVM) classifier directly on pooled electrode potentials for every time point independently. This type of MVPA typically provides higher statistical power over univariate methods by pooling information across all electrodes, enabling reliable estimates of latencies from single trials [24]. Here we found that, indeed, reliable decoding of superordinate category information (animal vs. non animal) was possible from single trials under 100 ms post stimulus onset (M1: 92 ms, M2: 96 ms; Fig.…”

Decoding Dynamic Brain Patterns from Evoked Responses: A Tutorial on Multivariate Pattern Analysis Applied to Time Series Neuroimaging Data

“…For example, decoding analysis can result in earlier detection of differences in the signals (Cauchoix, Arslan, Fize, & Serre, 2012;Cauchoix et al, 2014), and the differences found by classifiers can differ from those found in components (Ritchie, Tovar, & Carlson, 2015).…”