For good reason, evolution has protected our brains from the exterior with strong barriers. Still, humans have always found ways (not necessarily pleasant) for peeping inside the skull and into the brain. How to link the mess they found inside with the functions of the nervous systems is, essentially, the basic problem of neuroscience.

In earlier times, much of scientific knowledge was conveyed through drawings. Look for example at this drawing by Leonardo da Vinci, which outlines what seems to be the human ventricles. Even though Leonardo eventually abandoned the study of the brain (he deemed it “too complex”), artists and scientists continued to construct more or less faithful representations of the intricate brain anatomy.


Human ventricles, Leonard da Vinci (ca. 1510)


From a medical perspective, the precise location and ramifications of nerves and arteries is invaluable knowledge. Any surgical intervention attempts to minimise damage and therefore it is crucial to know how deeply involved in certain functions (i.e. speech, vision, etc) different brain regions are. Knowledge about the shape of the brain is not only useful from a practical viewpoint, but also tells us about the physical forces in play acting to conform the brain (much in the spirit of D’Arcy Wentworth Thompson’s masterpiece, On Growth And Form). For example, we know that the extensive gyrification of the human cortex is necessary to fit a large area into a very restricted volume. We can suspect that the surface of the brain must have some special status or characteristic when compared to the bulk of it. By opening up a skull and thinking about the forces in play, we might end conjecturing something that resembles the actual cortical organisation into surface layers of information-processing neurons, and a bulk of underlying “cables” connecting different surface elements.

Of course, barriers are there for a reason, and we should open the skull of any living person, unless absolutely necessary. However, in the previous century, physicists developed methods that allow to look into the brain of living, awake subjects and patients. These depictions are closer to reality than the drawings of Leonardo and others, but still, as long as we rely on visual inspection of the data, the principle is exactly the same. Even with functional imaging (such as fMRI), methods by which is possible to map in (more or less) real time neural function and activity, results are often provided in terms of static maps showing the association of different brain areas with different brain functions (“functional cartography”).

In fact, the data yield of modern experimental techniques is so large that it is difficult to draw conclusions by inspecting the data. The ground is prepared, for the first, time go beyond the description of (functional) neuroanatomy. Computational approaches are needed to understand how brain activity patterns emerge in the healthy brain, how different brain areas interact with each other, how these interactions are changed by different diseases, etc.

I am interested in developing techniques (often inspired in other domains of knowledge) to answer increasingly complex questions about the brain based in experimental data that allows to look into the brain. Some of these computational methods include:

  • Machine learning and non-parametric inference.
  • Network modelling of brain interactions, causality.
  • Computational modelling and fitting models to anatomical constraints (structural connectivity) to model functional data.
  • Scaling-up these experimental techniques to facilitate the crunching of large amounts of data with low latency.



Beyond univariate statistics: obtaining and studying bivariate interactions of brain activity, modelled as network. From Tagliazucchi et al., 2013, Neuroimage.