Andrew Thwaites, Psychology Dept, Cambridge University & MRC-CBSU

Mapping the information processing pathways of the cortex: challenges and opportunities

Andrew Thwaites¹, Eric Wieser², Andrew Soltan³, Ian Nimmo-Smith⁴, William D. Marslen-Wilson¹

¹Psychology Department, University of Cambridge

²Department of Engineering, University of Cambridge

³Department of Pathology, University of Cambridge

⁴MRC-CBSU, Cambridge

Abstract

 The human brain is one of the most complicated information-processing entities known to mankind. Yet understanding the paths information takes as it moves around the brain - and how these paths differ between clinical groups – is a core component of clinical neurology. Furthering this understanding not only benefits basic clinical research (for instance by identifying processing deficiencies in patients with neurodegenerative disease and by revealing the restorative mechanisms of forms of treatment) but also underpins new, technologically innovative research such as brain-computer interfacing.

 With the advance of high-speed neuroimaging imaging techniques and high-performance computing comes the ability to map these pathways with increasing accuracy (Thwaites et al., 2014). Although the amount of brain imaging data needed to extract these pathways is large (upwards of 7000TB for a group of pathways), the resulting ‘map’ is able to trace the processing of information as it moves through the brain.

 The nature of the data involved raises two key challenges. First, there is the visualisation of the data. Pathway data has five major dimensions (three spatial dimensions, a temporal dimension and an ‘evidence’ dimension), making it difficult to render on the printed page. This problem is further exacerbated by the volume of data – at present 3,000 pathways have been identified, and the number is rising. It is difficult to present these large quantities of data in a format which is useful to viewers; unlike man-made computing mechanisms, where engineers employ hierarchical and modular design principles to elegantly demarcate processing pathways, biological processing pathways are a result of protracted evolutionary processes and offer no such framework. The second challenge is that of sharing such a large and complex dataset effectively. When the data is complex, the interpretation of shared data can cause large time overheads for the recipients.

 We discuss our approach to tackling these problems: an online interactive visualizer of processing pathways (the Kymata Atlas, https://kymata-atlas.org), that serves as a front-end for a publicly accessible application program interface [API] of pathways data. Emerging web technologies such as HTML5 allow the display of high-dimensionality data, from complex databases, in a format far superior to that achievable on the printed page. Open source graphics libraries allow data to be rendered as interactive, animated graphs, allowing the sorting and partitioning of the data at the users’ discretion. Thus, solving the first problem, of ‘visualising’, is made possible with application of relevant technology.

 The second problem, facilitating the sharing of complex data, is solved twofold. First, users of the atlas can share their chosen view settings online (so that the data viewed by one user can be replicated by another). Second, users can access the raw data through a set of software interfaces, the APIs, which allow structured access to this data for integration into custom applications and novel analyses. The latter ensures open access to all information held in the database, encouraging third-party researchers to explore and use the raw data, thereby best serving continued study of clinical neuroscience.

References

Thwaites, A., Nimmo-Smith, I., Fonteneau, E., Patterson, R. D., Buttery, P. and William D. Marslen-Wilson  (2015) Tracking cortical entrainment in neural activity: auditory processes in human temporal cortex. Front. Comp. Neurosci. 1-14 doi: 10.3389/fncom.2015.00005