No sound consensus – FAQ

Our team “No sound consensus” won Hackathon Challenge #1 at the 19th Annual Meeting of the Organization for Human Brain Mapping.

What is the Human Brain Mapping meeting?
It is the largest international meeting dedicated to mapping brain function using neuroimaging. It is attended by around 2800 people. The 2013 meeting was held in Seattle, Washington from June 16-20, 2013.

What was the aim of the challenge?
The Human Brain Mapping Hackathon was an analysis and resource building competition designed to accelerate the connection between open neuroscience and cloud computing. Challenge 1 was to discover the best imaging and gene expression relationship via integration of imaging data with the Allen Human Brain Atlas.

Who sponsored the challenge?
Organization for Human Brain Mapping
Amazon Web Services
Allen Institute for Brain Science

Who were the team members?
From the Cusack Lab at the Brain and Mind Institute of Western University, Canada:
Rhodri Cusack, Charlotte Herzmann, Annika Linke, Conor Wild, Leire Zubiaurre-Elorza

From the Daley Lab at the Brain and Mind Institute of Western University, Canada:
Mark Daley

From the Peelle Lab at Washington University in St Louis, USA:
Jonathan Peelle

What did No Sound Consensus do?
We “mashed up” terabytes of data from two sources: the Human Connectome Project and the Allen Brain Institute.

When we listen to music, speech or any other sound, we use an area of our brain called “auditory cortex.” There is currently no consensus on whether this is one big brain system, or a collection of smaller regions with distinct functions. This contrasts with our understanding of the visual system, where there is broad consensus that there are many regions, each selective for a different visual property.

Why Hackathon?
To conduct the analysis, we needed a multidisciplinary team with the necessary computing, mathematical, neuroscientific and genetic expertise. We used many different software packages and different programming languages.

Why the Cloud?
We needed a high-performance cluster computer to process the images in time for the competition. We estimate that in less than three weeks, we used one-years worth of computing processing time. We followed 3.2 billion tracts through the brain, and created 4.5 Terabytes of results.

What do the results show?
Our results show that auditory cortex is just as modular as visual cortex, and we were able to identify a parcellation for it. You may view these results in our interactive viewer. Finally, we examined genetic expression data from the Allen Brain Institute and found that while it varies from region-to-region within visual cortex, it does not differ across auditory regions.

Why is this important?
It will provide a framework for understanding how the human brain processes sounds, such as speech, music or environmental noises. This in turn will help us understand how this system can become disrupted, in developmental disorders (e.g., specific language impairment or dyslexia), following brain injury (e.g., aphasia, amusia) or psychiatric conditions (e.g., hallucinations).

Tell me more about the data and the analysis.
We used three kinds of data:

  • Diffusion Imaging from the Human Connectome Project. These were acquired on a new generation of MRI scanner at a higher definition, and clearer, than any seen before.
    We can use these images to see many new aspects of brain connectivity. The example image to the right shows the direction of axon connections in a slice through the brain (red=left/right; green=front/back; blue=top/ bottom)
    From these images in a group of people, we performed “tractography”, and traced out, for each part of auditory cortex, what it was connected to. Example tracts radiating out from auditory cortex in a single person are shown in the image to the left and below.
  • Resting state from the Human Connectome Project. Volunteers were asked to rest and relax while lying in an MRI scanner. Even when we’re just daydreaming, regions of the brain that are connected tend to fluctuate together. We can use this pattern of corresponding fluctuations to understand how each part of auditory cortex is connected to the rest of the brain.
  • Gene expression data from the Allen Brain Institute, which shows for 3000 points across the human cortex what genes are expressed.
Once we had established the connectivity of each part of auditory cortex to the rest of the brain, we applied a mathematical technique called “Graph Theory” to group together all the parts with a similar connectivity. The results are shown below, on a 3-dimensional and flat rendering. Each colour shows a module of auditory cortex that has a distinct pattern of connectivity. The top row shows results from the Diffusion data alone, the middle row from the resting state, and the bottom row from putting the two together.

Further images and information
See this presentation, or contact rhodri at or  519 494 52-zero-one




Leave a Reply

Your email address will not be published. Required fields are marked *


You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>