Friday, 29 May 2015

Neuroscience is too modular

The advances in neuroimaging during the 1990's provided a strong pull for Psychologists to move into neuroscience and study the brains function in vivo.

This shift has been amazing for developing our understanding of cognitive processes from a neural perspective. However the combination of brain imaging experiments and Psychologists with their baggage of abstract conceptualisations and box and arrow diagrams, has provided fertile ground for hundreds of publications that push a modular, or neo-phrenologic, perspective on brain functioning.

In short Psychology breeding with neuroimaging has spawned many modular neuroscientists. Researchers who label the cortical anatomy by their alleged function: Language, Inhibition, or motor centres. There simply isn't enough space to fit the whole of Psychology onto the cortex!

Neuroscientists are considered modular if they tend to parcellate the cortex into functionally discrete units. i.e. region X is responsible for function Y. Despite acknowledging that processing may rely on spatially disparate regions, fMRI data is too often interpreted though a thin veil of modular rhetoric. The majority of fMRI studies to date have analysed data using a mass univariate analysis and contrasting data from one task or group from another. This approach will reliably reveal regions active during a particular task. The problem is that these contrast maps don't show regions that are solely responsible for the process under investigation because those regions will undoubtedly show up again in separate maps from completely different tasks. Moreover these maps are essentially a snapshot (an average to be more specific) of brain activity. They fail to portray regions interactivity.

In addition to the interpretation of fMRI studies, researchers of particular neuropathology's or psychiatric disorders can also fall into modular interpretations by making bold claims about groups differences being specific, iconic or hallmark representations of their disease of interest. This is likely due to these researchers spending their complete working lives looking at differences between their population and controls and seeing the same differences cropping up in different studies. It is often that you hear of biomarkers or neural correlates of disorder X. However when you take a broad view of neuro-psychiatric complaints you will find that the symptomology is shared by many and not specific pathologies. Impairments in sleep and memory, for example, are known to anyone diagnosed with any neuro-psychiatric complaint. What is more likely is that the same processes are active across many disorders. Perhaps it is differing combinations of a few common pathological processes that give us the multitude of diagnoses available today?

A network perspective takes a slightly different view. Using fMRI data to look at brain connectivity paints a different story. Brain regions do not work in isolation, there are not centres for cognition. Regions are social, they co-activate and influence each other in complicated and highly dynamic ways. Take any region that is reported to be a centre of any cognitive process, and run a basic seed correlation analysis and you will find other, if not many, regions functionally connected to it, co-activating, with it. No region is functioning in isolation.

Even the globalists taking a network perspective are influenced by modular/phrenologic type thinking. Some say Anterior Insula Frontal Opercular network, some say Salience network. The latter label is indicative of modular influences because it implies and endows that network with a specific function, its only role is to detect/be active in response to "salient" stimuli. Admittedly, this is a small step in the right direction but the approach simply moves the problem back from endowing a single region with a function to the network level. When people refer to a "Salience" network, I congratulate the recognition of a spatially distributed set of regions working together, however I disregard the assertion that the network is solely active for the purposes of salience detection.

It is becoming more and more clear that the brain is complicated network of networks. Spatially distributed sets of regions activating together during information processing. These networks often take up numerous roles and so make it difficult for Psychologists using functional imaging to summarise and reify their roles. Psychologists may try to cram all of their concepts onto the cortex, however they will find themselves out of room pretty quickly.

Of course a network perspective is problematic because it paints a more complicated picture. Advanced signal analyses are taking us closer to addressing the temporal dynamics of fMRI data which may provide further insight into frontoparietal networks that seem to share patterns of cortical real estate for example. Perhaps future investigation into frontoparietal domain general cortex will not find more subtle ways of dividing up the anatomy space but will instead look at network dynamics temporal domain. It is likely that differing processes that evoke these networks form differing temporal profiles within the same anatomical space in response to differing task demands.

No comments:

Post a Comment