Presentation Title

Classification analyses of fMRI data can predict perceived color

Faculty Mentor

Stephen A. Engel

Start Date

23-11-2019 8:45 AM

End Date

23-11-2019 9:30 AM

Location

14

Session

poster 2

Type of Presentation

Poster

Subject Area

behavioral_social_sciences

Abstract

The McCollough Effect is an illusion where viewing colorful vertical and horizontal stripes causes black and white stripes to appear colorful. The brain locations that produce this illusion are unknown; the general goal of this project is to reveal them. We measured neural activity using fMRI while three participants viewed colorful and black and white stripes. To test whether visual cortex produces the McCollough effect, we trained classification algorithms to predict what participants saw from activity patterns across voxels. We first tested classification without the illusion, training three different classifiers to distinguish activity patterns associated with colorful stripes from those associated with black and white stripes. We tried different ways of classification to determine the best method. Trained classifiers successfully classified patterns of activity from different scanning sessions and across conditions. Overall, classification using linear discriminant analysis with 10-fold cross-validation produced the best results. These results indicate that classification analysis can predict what colors participants are seeing. Our next step is to test whether the classifiers show the illusion, mistakenly classifying black and white stripes as colorful, when tested on data acquired when the illusion was present. Parts of cortex showing this are likely bases of the McCollough Effect.

This document is currently not available here.

Share

COinS
 
Nov 23rd, 8:45 AM Nov 23rd, 9:30 AM

Classification analyses of fMRI data can predict perceived color

14

The McCollough Effect is an illusion where viewing colorful vertical and horizontal stripes causes black and white stripes to appear colorful. The brain locations that produce this illusion are unknown; the general goal of this project is to reveal them. We measured neural activity using fMRI while three participants viewed colorful and black and white stripes. To test whether visual cortex produces the McCollough effect, we trained classification algorithms to predict what participants saw from activity patterns across voxels. We first tested classification without the illusion, training three different classifiers to distinguish activity patterns associated with colorful stripes from those associated with black and white stripes. We tried different ways of classification to determine the best method. Trained classifiers successfully classified patterns of activity from different scanning sessions and across conditions. Overall, classification using linear discriminant analysis with 10-fold cross-validation produced the best results. These results indicate that classification analysis can predict what colors participants are seeing. Our next step is to test whether the classifiers show the illusion, mistakenly classifying black and white stripes as colorful, when tested on data acquired when the illusion was present. Parts of cortex showing this are likely bases of the McCollough Effect.