Neuroscience 4 QuantumBrain

This is what I was researching for my 2nd lab rotation for the MSc under the supervision of Dr. Simon Stringer. It was a computational project with an emphasis on a proof of principle that the independent motion of objects in natural scenes allows us to learn them as individual objects (that is, form separate representations of them in our brain), even if they are always seen together.

I’ll briefly describe how the computational model works. We have 2 sets of arrays, an input and an output, where each element represents a neuron, the element’s value representing it’s firing rate (between 0 and 1). The input array is formulated as a ring of 75 neurons and there are two of these, one for each object. A block of 20 contiguous neurons in each ring are switched on which represent the object. This packet of activity can be moved around the ring to simulate different views of the same object.

These input arrays are connected to the output array, a 50×50 sheet of neurons, such that each input neuron is connected to every output neuron. The strength of these connections are described by the synaptic weight (a synapse is a connection between two neurons). The synaptic weight between one input and one output neuron evolves according to the activity of these neurons. Thus, if both neurons have a high firing rate, the weight will increase a lot. If one or both have a low firing rate, the weight will decrease. So the weight only increases if the input neuron actively drives the output neuron.

Furthermore, there are local excitatory connections in the output layer such that neurons near each other help to drive each other. This means that if an input neuron actively drives an output neuron then it will likely also drive other nearby output neurons. Consequently, its synaptic weight to these other neurons will also likely increase.

How do we train the output layer to recognise the objects? If we present the objects, via the input neurons, to the output layer, in many different positions on the ring many times (this allows the synaptic weights to evolve lots) then different parts of the output layer will eventually respond only to “preferred” positions on the ring. This is because the output layer also has global inhibition - that is, the more one neuron fires, the harder it is for another to fire. This essentially acts to drive competition between neurons so that they don’t all respond to the same input as this would be redundant. This type of competition is present in real brains. If it wasn’t, we would only be able to form a single memory which isn’t very useful!

These local excitatory connections provide a basis for self organisation - that is, neurons near each other respond similarly to each other (to similar inputs) and those further apart respond to different inputs (in this case, different “views” of the object).

Now, what is interesting is that if we move both objects around their respective rings in exactly the same way, allowing the synaptic weights to evolve, the output layer develops exactly the same response for one object as for the other. What does this mean? Let’s say we test the output layer by presenting it with just one of the two objects (remember, both are presented when we train the output layer). How can we tell which object is being presented if the output neurons respond in the same way to this object as to the other? Very simply, we can’t. The output layer thinks that both objects are the same object.

Now, if we train the output layer on objects that do not move around their rings in the same way, then we reduce the symmetry between the objects. They become “statistically decoupled” from each other from the point of view of the output layer. As such, the output layer is able to form one set of responses for one object and another set of responses for the other object. Now, when we come to test the output layer with just one object, we can tell which object is being presented based on the pattern of activity in the output layer. And so the output layer has learnt to recognise each individual object.

This phenomena can help us to understand many other behavioural attributes that rely on vision - in fact, it applies to all sensory modalities (touch, sound, taste, smell) as the important aspect of the inputs is that they’re statistically decoupled.

My supervisor and I are hoping to publish these results as a proof of principle for other, more applied work going on in the lab.

So the first term has finally come to an end with the handing in of my (rushed) dissertation. The project was pretty open ended and, as with any science, changed significantly since my post describing what I “would” be doing.
So here’s a wee description of what I was looking at and how I did it.

The population code is basically the ultimate goal for neuroscientists and as such is probably the furthest from our reach. By population code, I mean how the entire population of neurons, or a subset of the population, encodes sensory stimuli. Several techniques are available to do this, on of which is using an array of electrodes, for example a 8×8 grid of them. These have excellent temporal resolution but lack spatial information. As such, complex algorithms lacking in accuracy must be used to determine from where the voltage signals came from. Another method (and the method I employed) is 2-photon calcium imaging. This requires a dye which is injected into the brain area of interest. The dye is taken into the cell where it is cleaved such that it cannot escape. It has a high affinity for calcium with which it binds. Synchronous binding of calcium and excitation by laser light causes the dye to fluoresce which can be detected using photomultiplier tubes. Since large intracellular increases in calcium concentration are indicative of neuronal action potentials, this provides a good measure of neuronal activity and, as such, a movie of the dye-labelled brain can be taken to capture the neural activity. What’s more, the technique is much less intrusive than multi-electrode recording, has much better spatial accuracy and can have up to equal temporal accuracy whilst allowing measurement of up to hundreds of cells at any one time. This is important as there are large correlations between neurons in terms of their activity (intuitively so as this is how the brain encodes sensory stimuli).

So the questions I ultimately wanted to ask is whether the optic tectum, the main visual processing area of Xenopus tadpoles, exhibits any functional architecture, by which I mean do different neurons encode different features of visual stimuli and how are these neurons organised in the tectum? Why the Xenopus tadpole? Well, it’s a model organism which has been investigated in many areas of biology to a large degree. Also, it develops ex utero (outside of the womb) and thus provides us a rare opportunity to explore how its neural development from the outset.

A great deal of literature is out there showing that functional architectures exist in many animals, from primates to cats to ferrets to zebrafish (which are remarkably similar to tadpoles in their larval stage), at different degrees of complexity.

However, calcium imaging of Xenopus tadpoles has never been used to explore the population code. My project then needed to perfect the art of imaging in this species which, as I’ve experienced, is not an easy task.

Thus my project ended up crunching through methodological procedures including: 1) bulk loading of the fluorescent dye into the whole brain; 2) movie processing; 3) composition of analytical tools (using Matlab) which can extract functional information; 4) invesitgation of the effects of cell excitability on neural activity discrimination and 5) investigation of response characteristics to different visual stimuli. These tasks out of the way, the lab is now equipped to explore the neural ode in populations of 100s of neurons. Very exciting!

Unfortunately, my time in the lab has ended and I’m now starting my next lab rotation with Dr. Simon Stringer, composing computational models which explore the properties of self organising maps in neural networks and how the structure which creates them allows for transform invariant recognition of complex objects in cluttered visual environments. I’ll report on this in a few months time!

I’ve been pretty poor up to date with blogging on random things I’ve been up to. This is an overview of my first Michaelmas Term but to save me from writing a humungous blog and you from reading it, I’ll keep it brief! Any questions then just comment

Lectures

We had 5 modules running this term: Introduction to the brain - an overview of historical achievements in neuroscience, methodological approaches and a word of caution when someone declares that they have discovered something (general rule of thumb: neuroscience is so multi faceted and so complex that you can usually either refute something from one aspect or another or the discovery simply begs another question, as is as expected with a relatively young field of science); Neuroanatomy - does what it says on the tin. Thankfully this module was not assessed as there are approximately one kabillion new words to learn. I probably learnt 5. Still, it was quite interesting learning that specific functions of the brain are tailored for by the specific and differentiated structures of the brain; Neuronal cell and molecular biology - this should have been 2 modules! Generally, it included the cell biology of neurons and glia and the development of the brain as a consequence of the development of individual cell types. Molecular biology just covered transcription and translation. “Just covered” is a bit misleading as there are firstly many processes involved with this and many variations of each which makes sense seeing as genetics are the programming language of life!; Synapses and transduction - this was my favourite course as it was, in parts, very mechanistic (especially sensory transduction from physical stimuli to electrical nervous signals). It looked at the scale of ion channels and the different mechanism by which they operate and govern the signalling properties of neurons. Within this topic comes pharmacology - the study of what chemicals are involved in transduction and what drugs can alter this; Overview of systems neuroscience - this focussed on the most prominant (i.e have been studied the most/take up the most processing power of the brain) of sensory systems such as the visual, olfactory, auditory systems etc. with a larger focus on the motor system and motor diseases. This largely comes down to the physical structure of the brain but an interesting question remains (and is being studied intensely): how do sensory signals control the growth of the structure of the brain? Genomic information sets the rules for how a particular cell will grow, where these rules give rise to functional responses to the neural signals that the cell receives.

Animal handling training

These lectures and practicals covered a wide range of ethical, legal and practical aspects to breeding, housing, experiments and the mimisation of suffering of animals used in neuroscience.

Questions we must think about when taking an animal outside of their natural environment are centred around, for example: its emotional response or ability to have an emotional response; how to care for it in a way that minimises its suffering/stress/pain. This is of vital importance not only because these are living, feeling animals who are, to a greater or lesser extent, “conscious” of their being, but also because animal stressors produce physiological changes in the animal that can render experimental data invalid. This can be from data about behavioural responses to simple blood samples. How to check for signs of suffering are taught and these are dependent on the physiology/anatomy of the animal as well as its evolutionary behavioural traits. What is emphasised is the recognition of that animal’s normal behaviour, when healthy! If an animal has always shown low inquisitiveness and little activity, one may think that this is normal. In some cases it will be but in others, one’s entire relationship with an animal may be during an extremely stressful time for that animal, e.g. after an operation, and thus abnormal behaviour can seem normal to that person.

There is a multi hierarchical organisation of vets, carers, liscence holders etc. right up to the Home Office who must be involved in the training of any scientist who wishes to use an animal in their research. These laws are very unique to the UK and often are very lax, if they exist at all, in other countries. Oxford University receives applications from all over the world to attend this course - it is very thorough and the course leader is very knowledgeable and experienced in the field.

One interesting fact is that only particular animals are protected, these being all vertebrates except humans, who can enter an experiment in the full knowledge of the procedure and implications (usually, though not always e.g. in psychological experiments!) plus octopus vulgaris which is apparently very bright! Further protection exists for animals who seem to have a special connection with humans, even though they may not show any more cognitive abilities than a “mere” rat, which are surprisingly intelligent! One such animal is the horse. It seems that, to a certain extent, our treatment of animals is based on our placing anthropomorphic attributes to these animals.

One last thing I will mention on the theory behind handling animals is what is hammered home again and again when designing an experiment - the 3 Rs: replacement - using non-animal alternatives or less sentient animals, reduction - use of good statistical analysis and planning to optimally (this is important as, if one animal returns invalid data, the whole experiment may be rendered invalid. Thus, a small excess in the number reduces the chance that the experiment is made void and therefore the unnecessary use of animals) minimise the number of animals needed, refinement - ensuring appropriate housing and care for the animals pre and post experiment.

The practical side of the course was very timid, us not having licenses. I have never had a pet and so was quite uncomfortable around the animals. Strangely (now obviously), the more assertion you apply when handling animals, the less discomfort they seem to show. For example, when learning how to hold a rat or pick it out of the cage, I gingerly cupped it in my hand and it wouldn’t stop wriggling! I thought it was going to jump off and fall to the ground. Rats on the loose are not good (even more so with mice as they can fit into smaller holes). I was afraid of crushing them in my hand. Actually, all I needed to do was to keep holding them tighter until they stopped wriggling. At this point, they felt secure. The only reason they wriggled beforehand was because they didn’t feel secure and thought that they might fall! You can tell that you’re not holding them too tight becasue, when you do, they let you know, claws, teeth and all.

Anatomy practicals

I was at first apprehensive on two counts: one that I was going to see real human heads (donated by that person to science, before you think that we are a murdering cult!) sitting on a table with parts of the skull removed and the other that I had been warned of it! Turns out that, apart from the smell of preservative, it’s not all that bad. The faces were respectufully covered in cloth and, even though that proved useless as it began to fall away whenever someone handled it to get a better view, all that was underneath was, surprise surprise, a pair of eyes under some eyebrows, a nose, the odd whole (or otherwise) ear etc. The skin had yellowed with the preservative and become very loose and rubbery, giving it a zombified appearance but, again, as is to be expected.

The prupose of the practical was to see the meninges, tough layers of skin that envelop the brain like a protective sheath. In the skull, the dura mater (the thickest, outer layer of the meninges) remains attached to the bone and does not some out with the brain, hence one of the reasons for using whole heads.

Different cross sections allowed us to view the capsule for the brain and cerebellum as well as the cranial nerves and arterial blood supply to the brain.

The practicals continued with elucidating the various distinctive parts of the meninges on the spinal cord, cross sectional changes along the extension of the spinal cord from rostral (head) to caudal (tail) ends, the cerebellar anatomy etc. We also had an intact human brain to handle and see how the various compartments are structurally connected, along with fixed brain slices so that we could see the internal structure.It was very difficult to hold a representation of the many areas of the brain in my own brain because, as you might expect, the different parts are not tinted with bright and contrasting colours as you find in text books! It had a pinky grey colour…throughout. Without good staining techniques and a microscope, the novice has only the ventricles (chambers of fluid), small swellings where neuronal tracts converge and slight changes in pinky grey to elucidate the different areas. Indeed, to think of isolated compartmentalised areas is destructive when trying to inderstand the brain’s functioning. Its connections are massively complicated, both locally and in relation to the other areas. Only in terms of these directional connections can one think of compartments of neuronal circuits and, even then, one cannot completely ignore what is going on in the processing of signals before it reaches these circuits, both in terms of signalling but also pharmacological and molecular states of the cell. Compare this with the first studies of comparmentalised functioning of the brain one or two hundred years ago, where personality traits were attributed to areas of the brain based on bumps in the skull!

Finally, we go to disect an eyeball. The eye is, in one word, amazing! I cannot begin to describe how overwhelming it makes me feel to think that such a refined piece of equipment has evolved from nature. The eye is not a separate organ but in fact an extension of the brain. To find out more about the eye, I will write an introduction to it’s structure and function on a dedicated page on the main website.

Psychophysics Practicals

We attended three practicals, exploring three visual psychophysical experiments including EEG (electroencephalogrophy) responses to attention, the ability to perceive contrasts in light intensity and voltage recordings of saccadic eye movements (moving the focus of your eye from one part of your visual field to another).

The EEG experiments take a Fourier transform of oscillations in the brain’s surface activity. These are measured by electrodes placed over the scalp. Oscillations are observed at specific frequencies which relate to the processing of certain tasks in the brain and are termed alpha, beta, gamma waves etc. For example, alpha waves correspond to oscillations in the brain of awake, resting subjects.

Contrast perception was measured by observing a sinusoidal grating of monochromatic stripes on a computer screen. The contrast between dark and light stripes was reduced and the subject made to state whether or not he/she could distinguish the individual stripes. The “yes/no” response to this followed a binomial distribution with contrast magnitude and the threshold is defined to be the contrast magnitude (in decibels) at which the subjuect could no longer distinguish individual stripes.

Eye movements can be quantified quite accurately by measuring the tiny voltage differences at a subject’s temples due to the slight polarity in voltage of the eyeball. By asking a subject to follow a dot on a computer screen as it either jumps from one spot to the another or tracks smoothly across the screen, one can measure these voltage changes and callibrate them to the angular position of the eyes in that subject’s visual field. Interestingly, corrective responses can be seen with increasing magnitude for larger saccades. The saccade is made, the eye focusses and if the focus isn’t at the desired point, small corrective saccades are made via the occular reflex system until focus is at the desired point. Furthermore, one builds up a detailed picture of their surroundings by making many small saccades over their entire visual field at an average of five per second! This happens even when you focus on one particular thing so as to avoid saturation of photoreceptors! However, when the dot tracks smoothly across the screen, the eye traces it smoothly as well, without saccadic movements. So it is clear that focus and occular reflexes strongly influence the control of occular muscles.

Communications Course

This was a two day workshop run by Peter Evans of BBC Radio 4 and Bernard Dixon, editor for New Scientist. Going into the seminar room on the first day, I thought that I was going to have to endure hours of obvious trawl with no real specific help. Wrong. This was a really enjoyable course and some really good points were made which, so long as I remember to think about them, will really boost how well I present my research to journals, peers and the public (something I am quite bad at doing judging by the number of blank faces I see when talking about what I’m doing!).

Peter and Bernard are really nice, down to earth guys. Being full of interesting experiences to tell of their lives in the world of science and media has given them a realistic outlook on how we communicate our ideas to others and how this final hurdle of communication can really alter the facts of one’s research simply because of the audience’s perception of what you’re trying to put across. I’d like to thank Peter and Bernard for their invaluable input.

Extra Curricular

Aside from the neuroscience course, I’ve attended various seminars plus a conference on Consciousness and Experiential Psychology, organised by the group of that name within the British Psychological Society in London. The focus of the conference was the classic study carried out by Bidet (a quick google search will give info on the experiment itself) which gave evidence towards preconscious activity in the brain, i.e. the brain begins to change its state of activity associated with making a choice before we are conscious of making that choice! Quite profound to the realm of philosophy, I think you’ll agree. I had a major headache for half the day so decided to sit out a couple of the talks in a nearby park where I watched a group of pigeons searching for food. This got me thinking about free will, as I will describe in a blog I’m going to write in the Philosophical Musings section.

Well, that was my time here in Oxford so far. It’s been a great first term and I haven’t even mentioned how nice the people in my group are, how it’s great to have people who want to talk about interesting and relevant things all around me and how addictive the enthusiasm for research is here. Needless to say, I’m very excited about what lays ahead of me. It’s going to be a very demanding time from now till September as lab rotations start alongside advanced modules and essay writing. It’ll be worth it, though, and then I can finally get on with the DPhil. This is what I look forward to most of all - being truly immersed in the world of research, developing ideas and expanding my mind to the truth of the life we live.

I am very happy today as I have been given my first choice lab projects for both Hilary and Trinity terms (Oxford jargon for 2nd and 3rd terms).

In Hilary term, I will be working with Colin Akerman in the pharmacology department. The project will be looking at neural coding of the developing visual system in Xenopus Laevis tadpoles. This includes tracing individual neurons at birth from a progenitor cell in the tadpole’s retinotectal system and recording the development of their spiking activity using calcium imaging with two photon fluorescence microscopy. The aim is to relate the developing activity with a mathematical model using receptive field mapping over the entire tectum to elucidate the process of development of an individual neuron’s functional identity within the network.

In Trinity term I will be working with Simon Stringer who leads the Oxford Foundation for Theoretical Neuroscience and Artifical Intelligence. The project will involve developing computational models of object invariant recognition in vision by projecting 1D objects onto a 1D input neural network and mapping this to a 2D neural network using feedforward and self-organising-map (SOM) connections. The network will be trained to recognise the objects such that the output receptive field will comprise object invariant domains surrounded by orientaion (or position, in the case of 1D objects) selective maps (one for each object). This results from sparse activity in the output layer so as to allow for orthogonal (i.e. non-overlapping) hyperdimensional (I didn’t really need to include that word but it’s cool, isn’t it?) vector representations of individual objects. This isn’t the case in the real brain where activity in cells further along in the feed forward chain (e.g. inferotemporal cortex) can involve ~50% of the cells, leading to highly convoluted representations. Time permitting, the project will aim to extract individual object recognition in the output layer from highly overlapping vector representations.

I’m dead chuffed with this as it will allow me to use lots of maths which, I must admit, I have been missing from my physics days. When I get the time (hopefully this Christmas holiday) I will expand on these projects on the dedicated neuroscience pages of the main website.

Recommended for all neuroscience graduates, I have attended the Autumn School in Cognitive Neuroscience at the University of Oxford this week from Monday to Thursday (although not really on Thursday due to a cold induced sneezing blitz!).

The school composed of 8 45 minute lectures a day on selected topics such as MEG approaches, the cerebellum, dyslexia, network modelling, goal directed behaviour etc. within the cognitive neurosciences. I am going to focus, here, on only a couple of points made in the talks which most caught my attention:

• “Distributed representations for response selection“, Ivan Toni, Donders Institute

This talk largely focused on a subject’s ability to judge whether a hand shape, presented on a screen in front of a subject, was a left or right hand. Once the image had been presented, the subject was timed for their response (usually a couple of seconds), the pressing of a button corresponding to either “left hand” or “right hand” and the number of successful/unsuccessful judgements was tallied.

It was found that they were more successful and had shorter response times when a hand was held in the main vertical line of the body, eg. at the belly button. This was an interesting result as one of the hypotheses stated that the subject orientated a mental representation of their own hand to match that on the screen, this proces being enabled further with the hand in-line with the body.

Another much more interesting result, I find, came from EEG (electroencephalography). EEG places a mesh of electrodes onto the subject’s scalp such that each electrode measures the local electrical activity of the brain at the near-surface. The area of interest in the brain was the fronto-striatal system in the frontal cortex which is believed to be involved with incorporating contextual information in action selection. Traces of the EEG signal (voltage-at-electrode versus time) were averaged over many trials such that EEG signals were equally supported for both left and right hands being displayed on the screen. The trace covered the time from the cue onset (hand shows up on screen) till after the subject confirmed their judgement with the pressing of the button.

A noticeable difference in voltage traces was observed between left hand or right hand images being displayed. The interesting fact is that this difference arose very shortly after the image appeared on screen and long before the subject consciously gave their response. This suggested that the brain had determined the result before the subject was even conscious of the result.

The point I would like to make is that this is, as is easily evident, susceptible to much speculation. There is, of course, a response time involved with the motor coordination for the finger to press the right button. There is also the possible fact that a subject will think of either their right or left hand and then proceed to fit the features of the hand displayed with their own such that, throughout the time course of the trial, they hold a mental representation of either a right or left hand in their mind. I’m sure there are many other arguments against the given conclusion as well but merely use this as a pointer to one of the central problems in cognitive studies.

• “What does the cerebellum really do?“, Mitch Glickstein, University College London

The cerebellum, located at the rear and bottom of the brain, has long been thought to contributed to motor coordination. The talk looks at a variety of experiments that correlate the complexity of motor actions with the size of the cerebellum in several species. There is strong evidence that motor coordination is the sole function of the cerebellum. Recently, there has been growing speculation that it may also contribute to cognitive processes. Glickstein presented evidence in criticism of such speculation but I just wonder whether time, yet again, will reveal the confounded nature of operations performed in the cerebellum as has been fund in other areas of the brain. Don’t get me wrong, there is strong evidence in neurology, this being successfully applied to the treatment of many ailments in people’s cognitive processes, that particular areas of the brain contribute to specific cognitive processes and motor actions. However, these are inevitably correlated with higher or lower order processes in the brain. For example, the recognition of a particular face is a high order process thought to be carried out in the inferotemporal cortex (TE) but is only possible due to lower order recognition of visual features such as contrasting shades which form straight, radial or concentric lines etc. which are categorised in the V3, V2, V1 (primary visual cortex) etc. areas of the brain and ultimately the retina which all form neural pathways that lead to the TE. No doubt areas of the brain which have stored memories of particular faces are called into action as well. More over, the cerebellum must be called into action to so that the eyes’ focus can be moved over the face in question.

It seems to me that no one area of the brain can be isolated to perform a specific function. If one was to create a lesion in the brain such that a particular area was isolated, the overall performance of the subject may change completely or may only just suffer a little, depending on it’s contribution to the task involved.

I would suggest that looking for connections with a directional flow from the cerebellum to the cerebrum (thought to deal with more with the many cognitive processes required for motor function and, really, most other thought processes) would help elucidate the cerebellum’s role in task performance. I would imagine that as more data becomes available, neuroscientists will find that, although they may manage to further isolate areas of the brain for specific functions, they will have to consider more areas of the brain for each task an experiment involves such that the confounding nature of data available to neuroscientists will stay roughly the same!