Speaker: Semir Zeki
Department: Dept. of Neurobiology, Univ. College London, UK
Subject: A re-appraisal of brain strategies for constructing the visual image
Location: Erasmus MC
Author: Nemo Andrea
Professor Zeki has a proven track record in neurobiology and his lecture was fairly accessible for someone with limited knowledge of the neurology/neurobiology field. Semir Zeki is professor of Neuroesthetics at University College London. He has received various prestigious awards such as the Golden Brain Award (1985), King Faisal International Prize in Biology (2004), Erasmus Medal (Academia Europeae, 2008), and Aristotle Gold Medal (2011). He has given over 60 lectures in his career and published three books. His experience was reflected in the seminar, as it was clear that he was comfortable in the position of lecturer, which made his message that much more coherent. In this seminar he challenged current ideas – or what he deemed misconceptions – regarding image processing in the brain and touched upon the topic of parallelism.
As mentioned before, Professor Zeki believes that the current believes regarding how image processing in the brain is handled is inaccurate. He stressed how in light of contrary evidence, many people in the neurobiology field still held on to outdated belief regarding image processing in the brain. He outlined how he saw the evidence as convincing that his main point was indeed the correct interpretation considering the current understanding of the brain
The main point of contention was the question of whether all visual information is passed through the visual cortex (V1) before being passed through to other visual cortexes or processing regions in the brain (the hierarchical model). This would mean that the processing done in the V1 section would be critical for the other visual processing sections in the brain. Professor Zeki’s was driving the point home that this is simply not the case and that visual signals are fed into V1 and other visual processing regions (V2, V3, V4, V5) in parallel fashion, allowing for asynchronous and parallel processing of visual data in the brain. He stressed how this asynchronous nature is often overlooked and not properly reflected in computer based models of human vision.
One of the main arguments and points of evidence against the hierarchical model discussed by the speaker was the fact that there are known cases in which a patient has tragically damaged the V1 area of the visual cortex, yet retains the ability to see fast motion; something attributed to the V5 area. If the hierarchial model would hold, the signals from the eye should pass through V1 first, but, seeing how this would be damaged in the patients in question, these signals would then never make it to V5 and should subsequently not be observable by the patient. The fact that the patient is able to see the motion attributed to this region of the brain suggests the model is not accurate.
This rejection of the model proposes a more complex model of interactions between the visual cortex regions. This model heavily relies on the idea that image processing is a highly parallelised process. While the concept of parallelisation of processes in the brain is not a controversial or even new concept, the speaker stressed how image processing must be a parallel and asynchronous process. His main arguments were that 1) with the new model for image processing, the same signals would arrive at the different visual areas at different times, requiring some form of asynchrony in order for the visual cortext to produce a meaningful image out of all the areas 2) Given the short response time and general efficiency of image analysis in the human brain, the visual cortex must employ the vast ‘’performance’’ benefits provided by parallel processing.
This new insight in the functioning of human image processing is interesting, as this new view of the system would allow for more dynamic behavior depending on the type of image. This would be a very complex and useful thing to quantify further and to develop accurate simulations for this system. I believe nanobiologists could play a relevant role in this process, as they should have a good mastery of the mathematics required to model such a system and the physics to consider the physical limitations of the system. While the neuron itself is already a highly dynamic system, the brain is a beautiful example of how (roughly speaking) local interactions can lead to the highly dynamic and complex behaviour known as consciousness. These new insights into the brain can help us better understand it and if we pair this with a more nanoscale understanding of neurons, it may be possible to either improve the current computational image analysis algorithms or give us fundamental insight into the inner functioning of the brain.
If you wish to read his full article on this topic, visit: http://journal.frontiersin.org/article/10.3389/fnint.2015.00021/full