Neural Connectomics Challenge #1: From Imaging to Connectivity
The challenge is over. 143 teams entered. Congratulations to the winners!
- 1st prize (1st place): Antonio Sutera, Arnaud Joly, Aaron Qiu, Vincent François-Lavet , Gilles Louppe (team AAAGV)
- 2nd prize (3rd place): Ildefons Magrans
- 3rd prize (4th place): Lukasz Romaszko (team: Lukasz 8000)
The second place participants decided not to reveal their code. The results of verifications are available for review. Information on the methods can be found in the fact sheets [pdf][form][summary].
The Kaggle platform accepts post-challenge submissions.
Understanding the brain structure and some of its alterations caused by disease, is key to accompany research on the treatment of epilepsy and Alzheimer's disease and other neuropathologies, as well as gaining understanding of the general functioning of the brain and its learning capabilities. At the neural level, recovering the exact wiring of the brain (connectome) including nearly 100 billion neurons, having on average 7000 synaptic connections to other neurons, is a daunting task. No traditional neuroanatomic methods of axonal tracing cannot scale up to very large networks. Could there be alternative methods recovering neural network structures from patterns of neural activity? [Learn more ...]
Today's cutting edge optical imaging of neural activity (using fluorescent calcium indicator proteins) provides a tool to monitor the activity of tens of thousands of neurons simultaneously. Mathematical algorithms capable of discovering network structures are faced with the challenge of solving a new inverse problem: recover the neural network structure of a living system given the observation of a very large population of neurons. To achieve reconstruction of a detailed connectome, we need excellent performance. Even tiny changes in connectivity can dramatically affect the way in which a neural circuit process information (like e.g. storing or reading out memories). Monitoring changes in effective connectivity patterns of a network in action during behavior promises to advance our understanding of learning and intelligence. This challenge will stimulate the advancement of research on network structure learning algorithms from neurophysiological data, including causal discovery methods. We need your help! [Learn more...]
Brain of the zebrafish in action. Today's cutting edge neurophysiology multi-electrode recording tools are capable of recording (and even stimulating) of the order of 100 neurons. Optical imaging of neural activity using fluorescent calcium indicator proteins (calcium imaging) provide a tool to increase the number of neurons recorded by three orders of magnitude. Recently, researchers have been able to record activity of the brain of a zebrafish embryo in 80% of its 100,000 neurons.
This video comes from the work of Ahrens et al. Nature 485, 471–477 (May 2012).
Please register to our Newsgroup to receive all notices.