Notes from lectures about the brain

From Site
Jump to: navigation, search

Notes I made while watching videos of talks. They are not summaries of content, only points that I found interesting, mixed with my own thoughts at the time.


Functional specialisation in the brain

I watched this talk as part of my quest to collect functional puzzle pieces of the visual cortex from the literature.

Studying functional specificity using fMRI

Resolution of fMRI: 50 000 voxels. Each 3 mm across. Each contains on the order of 2 million neurons.

Fusiform Face Area: sensitive to faces.

Parahippocampal Place Area: responds to pictures of some place scenery (as seen from a 3D viewpoint).

Extrastriate Body Area: bodies and body parts (but not faces).

The above three areas are strongly selective — the response is at least 2:1.

Fusiform Face Area

Discrimination. Humans have to tell one face apart from hundreds of others, all of which are very similar objects.

Experiment: to discriminate similar houses from one another, we have to move the features a lot more, than the corresponding features in the face discrimination task.

In monkeys: Doris Tsao, 2003. Used fMRI to find three areas responding to faces, and one responding to images of bodies. Electrode studies over 320 neurons in those areas show that they respond exclusively to faces (with very weak responses to round objects which look like very blurred faces).

Familiarity with a face does not affect the strong response that the face area gives when it sees a face.

Lateral Occipital Complex

Learning to distinguish quite similar, but distinct objects (made up of similar basic features, but with slight overall positioning differences) with 10 hours of training, affects a small region called the Lateral Occipital Complex. LOC is in the ventral pathway.

This suggests that the machinery for such tasks is localised in the cortex.

Visual word form area

This is for comparison with the face area. We see written words as often as we see faces. Faces were present in our primate ancestors' environments, while written words were not. Natural selection hasn't had the time to put in specialised machinery for the words. A very tiny area for written words was found in 80% of individuals (using a higher resolution fMRI, 1mm voxels). This shows that with a lot of experience, regions can be allocated in the brain for specific tasks. The face area is a lot larger — we infer that it's not all due to experience. (Response to strings of consonants was high, but to strings of digits was low!)


Principles of vision that we can borrow from the brain

  • Brain global control architecture hypothesis: it creates a dynamic clustering across different levels of complexity. (Which area of the brain is responsible for what function changes dynamically, based on context.)
  • Need to work at several levels at once: genetic (molecules), microscopic (neurons and columns), mesoscopic (functional sub-systems), macroscopic (complete behaviour organisation of an intelligent entity). How does each level interact with the neighbouring levels? A mean-field approach may ignore highly important details.
  • Approach a cortical column as a tentative elementary cortical processor. It reduces a noisy and high-dimensional inputs into low-dimensional outputs suitable for decision-making within the current problem's context.

Strategies used throughout thinking processes:

  • Trial and error.
  • Divide and conquer.
  • Search for analogies.
  • Active exploration.

  • View paths: views of an object do not appear randomly, but are changing in a pattern related to the motion, which is often regular and can be predicted. There is a manifold of views and we see a sequence of views which form a curve on the manifold.
  • Complex features learned by sparse coding: commonly-seen corners.
  • Inferotemporal (IT) cortex is like an alphabet to describe more and more complex objects.
  • Combinatorial encoding: an object will have several elementary parts.
  • Tsunoda et al. 2001: macaque monkey brain areas active when shown a full object, a part of the object, and just its outline.


Brain-like Vision video
Edgar Körner, Honda Research Institute
10 October 2008, at Berkeley

Questions raised

  • Suppose that each concept is represented in the brain by a small set of neurons. How is it possible to connect any concept to any other concept? Consider the numbers.
    • It is possible that there are intermediate neurons that can be used for this. These do not have to be dedicated. One intermediate neuron can participate in many association links.

Recovery of consciousness in patients

  • When motor function is fully impaired, it's impossible to tell whether we observe coma (nothing), vegetative state (sometimes eyes are open), minimally conscious state (mind responds to some stimuli) or fully conscious locked-in state: all look similar.
  • ... unless you use fMRI to look at the brain directly. Communicate to patient in normal sensory ways. Patient communicates by thinking (doctor says "imagine that you are swimming when we point to the letter A").
  • Similar areas of the brain activate in all normal humans when they imagine such activities as swimming, playing tennis, walking around the home (distinct patterns of activation for each type of activity).
  • Zolpidem (sleeping pill) administration wakes up patient from minimally conscious state. (Can now walk around.)
  • Zolpidem with/without fMRI study shows twice as much brain activity (metabolism, to be more precise). Most increase in the frontal areas. EEG study shows: 7.7Hz peak disappears, 23.7Hz peak (beta) appears, power increase across the range [15Hz, 100Hz] raised by about 5 dB on average. More than convincing unconscious/conscious distinction.
  • Possible explanation: at the level of a mesocircuit (meso meaning "intermediate scale"?), we could model parts of the brain (frontal, striatum, GP — globus pallidus, central thalamus, parietal/occipital/temporal cortex) and their interactions in terms of excitation and inhibition. One inhibits another, that inhibits a third one.
  • Medicines working on the dopamine system often have effect with brain injured patients.
  • Zolpidem affects GABA A receptor, inhibiting something (GP?) that was no longer being inhibited (as it's normally supposed to have been) following injury.
  • Deep brain stimulation of the central thalamus has a significant positive effect for a patient. It seems that the reasons for this are not known — it is a scientific observation.

Summary: if a brain injury doesn't appear to be major, but consciousness seems to be lost, perhaps some medicine can be used to inhibit or activate some area, which would work like turning on a switch in a system that is mostly structurally sound, but is not functioning because of a broken link in the excitation/inhibition chain. (My guess is: look at things that inhibit the thalamus. If the thalamus is injured, it's not good: the thalamus seems to be the most critical part of the brain for consciousness.) Especially central thalamus, according to Schiff.

Excerpts of explanations in response to a question which seems to be about the nature of consciousness:

"We view consciousness, primarily, as an expression of the cortico-thalamic system. Now, why does central thalamus play such a role? Well, it turns out to have a special anatomy. But it's only special with respect to how it helps to organise the cerebral dynamics of this entire structure. And, to some extent, it becomes replaceable."

"... because of the geometry of its [central thalamus] connections. I would say I focus on the central thalamus not because it's going to explain consciousness to us, but it's practically a very economical choke point for manipulating states of consciousness and understanding some aspects of it."

Another question was about the precuneus (large region in the medial parietal area) and consciousness. There is a link: metabolism in that area is highly correlated with conscious state (as opposed to sleep). Off the top of my head, that area contains association centres, hence this is not surprising.

My question: is the thalamus the clock of the brain? Is it the heart of the brain, that sends timed pulses out to the whole neocortex, synchronising it in some way? What are those beta waves, associated with consciousness, at about 20 Hz? Neurosurgeons have electrodes implanted in patients with various conditions. Are there any close to, or in, the thalamus? What can we say about the neurons of the central thalamus? Do they all fire at the same time? Is there any pattern at all to their firing?


Understanding the Recovery of Consciousness video
Nicholas Schiff, Cornell
15 March 2010, at Stony Brook