How the Brain Responds to Images

Overview

This project explores the relationship between images shown to a human observer, and concurrent 3D fMRI images of the subject's brain while viewing the image. fMRI is capable of measuring the absorption of oxygen from the bloodstream in live patients. Absorption is higher in regions of the brain that are most active, so these images give a volumetric picture of the parts of the brain that are "working hard" during various thinking activities. These data come specifically from the visual areas of the brain, and the measurements are taken while the subject is viewing a series of images. So the fMRI activation patterns can illuminate many aspects of how the visual system interprets images.

Questions

This data is extremely rich and can support a number of basic questions:

* are there areas of visual cortex concerned with recognizing certain *types* of things (e.g. people, animals, buildings)?

* are there areas that respond to attributes of objects (color, size, orientation)?

* are there areas that respond to parts of objects (legs, tail, fur etc)?

* how accurately can we predict the stimulus (the image being viewed) given a brain activation pattern?

There is prior work on some of these questions, but they are far from thoroughly explored. You should pick one or more topics to explore.

The Dataset

Is approximately 1GB (compressed) and contains a number of volume images, and the 2D images that they correspond to. The data are available online and we will give you a link if you choose this project. The data come from Prof. Jack Gallant's lab on campus Links to an external site..

References

This link will take you to a discussion page with a zip file of relevant papers.

Tools

You should pick an appropriate toolset. Its possible to do some analysis in Python, but most questions are compute-intensive so you are likely to get better performance with BIDMach or perhaps Matlab. The techniques be classification from labels (you may provide some) and perhaps clustering.