On 12 May in Brussels (Belgium), the final meeting was held including the presentation of the results of the European project BrainScaleS, a project involving Etienne Hugues a researcher at the Center for Brain and Cognition (CBC) directed by Gustavo Deco, ICREA research professor of the Department of Information and Communication Technologies (DTIC) at UPF.

The contribution of UPF's CBC to the BrainScaleS project has involved the development of a computational model of spontaneous brain activity at rest, as well as data analysis and a computational model of neural activity when the brain performs a specific task, such as a perceptual discrimination exercise for which the team worked together with Mexican professor Ranulfo Romo, who has conducted experiments on perceptual discrimination in primates.

In rest state, the neural networks are in equilibrium

With regard to brain activity in rest state, the studies performed with magnetic resonance imaging in humans have shown that fluctuations in the oxygen uptake of the nerve cells point to specific organization in networks. By comparing the computational models constructed with the physiological data obtained it can be seen that neuronal excitement and inhibition are two processes that are kept in equilibrium in a state of rest.

In brain activity, the areas of the brain involved communicate

The team of Ranulfo Romo has carried out stimulation experiments with primates that have received two different tactile vibrations, separated by a time interval. The individuals had to discern which stimulus had a higher vibration frequency. Then, it measured neuronal activity during the performance of the task and was unable to verify that the neuronal activity is time-dependent and distributed in different cortical areas.

In addition, it has seen that neurons can separately give different types of response. In addition to neurons that respond only during stimulation, others have the role of working memory, ignoring the values of the stimulations, and a final category of neurons is involved in decision-making.

Adrià Tauste and Marina Martínez, members of the group of Gustavo Deco have studied the response of neurons, at individual and global level, in different areas of the brain, using information theory techniques. In this way it has been proven that, surprisingly, in the course of a particular task, all brain areas involved communicate with each other. By contrast, when the animal is stimulated, but instructed not to perform the task, communication between all areas disappears. That is to say, as Tauste has witnessed, "it is as if the global state of the brain depended on fulfilling a particular task or not".

A computational model of perceptual discrimination brain activity

With all of this information, the CBC, in the framework of the project BrainScaleS, has for the first time proposed a computational model of the brain's neuronal activity for a perceptual discrimination task. As Hugues states, "the proposed model consists of three neural networks: one sensory, another which acts as a working memory and the last one that would be involved in decision-making".

Continuing her explanation, Hugues says "we have demonstrated that this model not only reproduces the neuronal response categories identified in the brain relating to sensory processes, working memory and decision-making but, and this is even more important, using a mechanism of synaptic plasticity, our model in each successive assay, is able to learn to decide in the same way an experimental subject would do".

The European project BrainScales (Brain-inspired multiscale computation in neuromorphic hybrid systems) is already completed and the overall results of the project can be consulted on the BrainScales website. It has been an ambitious international project of the 7th Framework Programme, within Future Emerging Technologies (FET) for which the European Union awarded 8.6 million euros to carry it out in four years, starting 2011.