BMT 823 Neural & Cognitive Systems Slides Series …...Prof. Dr. rer. nat. Dr. rer. med. Daniel J. Strauss BMT 823 Neural & Cognitive Systems Slides Series 1 Introduction and Basics

Prof. Dr. rer. nat. Dr. rer. med. Daniel J. Strauss

BMT 823 Neural & Cognitive Systems Slides Series 1

Introduction and Basics of Neurobiology

Systems Neuroscience &Neurotechnology Unit

(Computational) Neuroscience

Artificial Intelligence Machine Learning

Neural Systems

Regularization Networks

Cognitive Systems

Approximation Networks

(Computational) Neurobiology

Neuropsychology

Expert Systems

Artificial Neural Networks

Computational Cognition

Neuroinformatics

Biocybernetics


Medical Research

Rembrand. Anatomy of Dr. Tulp, Oil on Canvas, 1632

Dali. Three Sphinxes of Bikini, Oil on Canvas, 1947

Brain Research

http://www.fotos.org/galeria/showphoto.php/photo/1565/size/big/sort/1


Philosophical Issues (in which we are not interested)

• Can a system be smart enough to understand how it works by analyzing itself ?

• Is the brain the hardware and the mind the software ?

• Can a computer based on a Von-Neumann architecture ever be self-aware ?

Systems Neuroscience &Neurotechnology Unit Cognition

mental process of knowing, including • awareness • perception • reasoning • judgment • attention • recognition


(Non-Engineering) Basics of Neural Engineering: The History of Theoretical Neuroscience

Systems Neuroscience &Neurotechnology Unit The History of Theoretical Neuroscience

• Hebb (1949) – Neural Populations – Plasticity

• Hubel & Wiesel (1959)

– Receptive Field – Early visual processing


• Von Neumann (1947) – Artificial Automata-> Cellular Automata – Artificial Life – Computer Architectures


• Rosenblatt (1958) – Simple Perceptron

• Minskey & Pappert (1968) – Linear separability


• Hodgkin & Huxley (1952) – Theory of action potential generation

• Roll (1960-1970) – Cable theory

• Fitzhugh & Nagumo (1969) – Analysis of excitability


• Hopfield (1982) – feedback networks

• Rumelhardt, Mc (1986)

- Multilayered perceptrons * * * * *


• Connectionism

• Large-scale computer simulations

• Brain as a complex dynamical system


The History of Theoretical Neuroscience (13)

• David Marr (1945-1980) – A theory of cerebellar cortex (1969) – A theory for cerebral neocortex (1970) – Vision (1982)


Albert Bregman 1990

Systems Neuroscience &Neurotechnology Unit Levels of Neural Modeling

• Brain as a whole • Specific brain systems (visual system,…) • Large scale neural networks • Small neural networks • Neurons • Ion channels and synapses • Molecular processes

Systems Neuroscience &Neurotechnology Unit Methods in Neural Modeling

• Analytical Solutions

• Numerical/Computer Simulation


Approaches in Neural Modelling

Level

Method Analytical Simulation

Single cell

Population

Rinzel, Termann, Izhikevich

Ermentrout Koppel

Bressloff

Amitt, Sompolinsky van Wreesvijk, Treves, Hertz, Hopfield

Poggio, PDP, Edelman, Grossberg

De Schutter

Abbott, Sejnowsky, Aertsen, Gerstner, Bienenstock, Koch,…

Destexhe


(Non-Engineering) Basics of Neural Engineering The Human Brain




(axial plane)

R.B. Graham (1990) Physiological Psychology


dorsal

rostral caudal

R.B. Graham (1990) Physiological Psychology


Brain

Cerebral cortex Forebrain

Midbrain

Hindbrain

Anatomy of the Brain


The central nervous system is divided into seven parts:

1. The spinal cord. Receives and processes sensory information from skin and muscles; sends commands to muscles; houses simple reflexes and controls locomotion.

2. Medulla. Autonomic function: digestion, breathing, heart rate. Sleep function: maintaining quiet.

3. Pons. Control of posture and balance.

4. Cerebellum. Learning, memory, and control of movement.

Kandel et al. (2000) Principles of Neural Science


5. Midbrain. Eye movements.

6. Diencephalon:

• Thalamus. Nearly all sensory information arrives here first.

• Hypothalamus. Regulates autonomic and endocrine function.

7. Cerebral hemispheres.

• Hippocampus. Memory of facts, events, places, faces, etc.

• Basal ganglia. Control of movement.

• Amygdala. Autonomic and endocrine response in emotional states.

• Cerebral cortex. Kandel et al. (2000) Principles of Neural Science

Systems Neuroscience &Neurotechnology Unit Triune Brain

Logical Brain (Neocortex)

Emotional Brain (Limbic System)

Survival Brain (Brain Stem)


Frontal lobe control skilled muscle movements, mood, planning for the future, setting goals and judging priorities

Parietal lobe receives and processes information about temperature, taste, touch, and movement coming from the rest of the body. Reading and arithmetic are also processed in this region

Occipital lobe process visual information

Cerebellum coined as the “little brain”, it governs movement, postural adjustments and stores memories for simple learned motor responses

Temporal lobe hearing, memory and language functions

Pons contains centres for the control of respiration and cardiovascular functions. It is also involved in the coordination of eye movements and balance

Medulla oblongata contains centres for the control of heart rate, respiration, blood pressure and swallowing

Brain Lobes

CEREBRUM

Systems Neuroscience &Neurotechnology Unit Limbic system

>> a group of interconnected structures that mediate emotions, learning and memory

Hippocampus plays a significant role in the formation of long-term memories

Amygdala involves in many brain functions, including emotion, learning and memory. It is part of a system that processes "reflexive" emotions like fear and anxiety

Thalamus a major relay station between the senses and the cortex (the outer layer of the brain consisting of the parietal, occipital, frontal and temporal lobes)

Parahippocampal gyrus an important connecting pathway of the limbic system

Fornix an arch-like structure that connects the hippocampus to other parts of the limbic system

Cingulate gyrus plays a role in processing conscious emotional experience and pain

Systems Neuroscience &Neurotechnology Unit The Papez Circuit

James Papez, 1937

Systems Neuroscience &Neurotechnology Unit The Papez Circuit…cont

Functions

Emotional Expression

Memory acquisition

Fear conditioning

Violence

Aggression


• In 1949, Paul McLean expanded the Papez circuit by adding the amygdala, and other areas that feed into or receive outputs from the Papez regions

• And so, we get the upgraded version of the Papez circuit, which is the Limbic System.

The Papez Circuit…cont


(Non-Engineering) Basics of Neural Engineering Neurons and Neural Communication


• The central nervous system (CNS) is entirely composed of two kinds of specialized cells : neurons and glia

• Neurons are the basic information processing structures in the CNS.

• Function of a neuron is to receive INPUT "information" from other neurons, to process that information, then to send "information" as OUTPUT to other neurons.

• Glia (or glial cells) are the cells that provide support to the neurons (old view - nowadays considered to be important in neural processing) -> blackboard


the major components of a neuron


• Neurons like any other cells, are packed with a huge number and variety of ions and molecules.

• Neurons are enclosed by a membrane separating interior from extra cellular space. Cell membrane is a lipid bilayer and is impermeable to most charged molecules and so acts as a capacitor by separating the charges lying on either side of the membrane.

• Ion conducting channels are embedded in the cell membrane which lowers the effective membrane resistance.

• -ve ions build up on the inside surface of the membrane and an equal amount of +ve ions build up on the outside

• The difference in concentration generates an electrical potential (membrane potential) which plays an important role in neuronal dynamics.


thickness of the membrane: approx. 7.5 - 10nm (in comparison, thickness of the cell body : approx. 50µm and more)

macromolecular pores form the ionic channels

Cell membrane


K : Potassium (Kalium)

Na : Sodium (Natrium)

Cl : Chloride (Chlorid)

A : Anion

Systems Neuroscience &Neurotechnology Unit Ion channels and pumps

Adenosine 5'-triphosphate (ATP)


dynamic equilibrium diffusional forces

balance electric-field forces

resting state:

stable active state that needs metabolic energy

(Na K - pump which keeps the ionic concentration stable)

- IT IS NOT A PASSIVE STATE -


Dendrites ⇒ „small branches“, often receive (or sometimes,send) electric signals from other cells Cell body ⇒ a.k.a soma, sums (“integrates”) electrical potentials from many dendrites. Contains necessary structures for keeping the neuron functional; nucleus, mitochondria and other organelles. Axon ⇒ conducts electrical signals and transmits them to other cells. Axon Hillock ⇒ The junction between the soma and the axon. When the received electrical potentials result in a sufficient change in the membrane potential, a rapid depolarization is initiated here. Myelin sheath ⇒ an insulating layer which prevents gates on that part of the axon from opening and exchanging their ions with the outside environment Nodes of Ranvier ⇒ are holes between the myelin sheath where ions exchange for the production of an action potential

Schematic of a generic neuron

Systems Neuroscience &Neurotechnology Unit Schematic of a generic neuron…cont

Systems Neuroscience &Neurotechnology Unit Neuron morphology

• Unipolar cells : have one primary process that give rise to several branches:

- One of these is the axon and the rest serve as dendritic receiving structures.

- Unipolar cells have no dendrites arising directly from the cell's soma. These cells occur in certain ganglia of the autonomic nervous system of vertebrates.

• Morphologically, on the basis of the number of processes arising from the cell body, neurons are classified into three groups.


• Bipolar cells : have two processes emerging from the cell soma: – a peripheral process or dendrite

which conveys information from the periphery and a central process, the axon, which carries information toward the brain.

– These cells have mainly sensory functions: retina, olfactory epithelium and sensory cells of the spinal ganglia

Neuron morphology…cont


• Multipolar cells : have a single axon and one or more dendritic branches emerging from all parts of the cell body. – Multipolar cells vary in the number and length of their dendrites and the

length of their axons. The number and extent of dendritic processes depend on the number of synaptic contacts that other neurons make onto it.

– For instance a spinal motor cell, with a moderate dendritic tree, receives about 10,000 contacts. The much larger dendritic tree of a purkinje cell in the cerebellum, receives up to 150,000 contacts from other neurons.

Neuron morphology…cont

Systems Neuroscience &Neurotechnology Unit Shapes and sizes of neurons

Purkinje cells (or Purkinje neurons) are a class of

GABAergic neuron located in the cerebellar cortex.

http://en.wikipedia.org/wiki/GABA

http://en.wikipedia.org/wiki/Cerebellum

Systems Neuroscience &Neurotechnology Unit Action potential

• Electrical potential across the membrane of the cell´s axon is determined by the intracellular and extracellular concentrations of three single element ions potassium (K+), sodium(Na+) and chloride(Cl-)

• Two distinct phases of transmembrance electrical activity, the resting membrane potential (cell not conducting an electrical impulse) and the action potential (actual nerve impulse“)

• The action potential propagating down the axon is about 60-70mV negative inside w.r.t.t outside membrane during the resting potential and about 40mV positive during the action potential.

• The action potential is also referred to as spike because of its shape.

• The sodium pump keeps sodium(Na+) more concentrated on the outside during the resting state


Hodgkin-Huxley Model 1952

Nobel Prize, 1963

Hodgkin Huxley

Systems Neuroscience &Neurotechnology Unit Action potential…cont

• Conduction of action potential involves an active biochemical process (as

discoverd by Hodgkin and Huxley, 1952. Which will be discussed in detail later)

• Action potential is a membrane phenomenon.

• This was proven when action potential was observed even after squeezing

out all the axoplasma of the squid‘s axon.

• It depends strongly on the presence of Na+ ions in the extracellular medium

• Potential change results in the inward movement of Na+ ions - The resting membrane is much less permeable to Na+ ions than K+ and

Cl- ions - As the action potential generation increases, the membrane

permeability to the Na+ ions increases as well. - Hence allowing the inward movement of the Na+ ions ⇒ DEPOLARIZATION


• If the cell membrane depolarized from its resting state of -70mV, it will revert back to its resting potential

• UNLESS the depolarization reaches the threshold transmembrane voltage of about -40mV (in the case of the squid axon), an action potential is triggered

• The amplitude of the action potential does not diminish as it travels along the length of the axon

• Increasing the strength of the stimulus, does not increase the size of the propagated action potential

• This behaviour is called the all-or-none

Action potential…cont





• The frequency of impulse generation is limited. A few miliseconds after an impulse, no additional impulses can be generated ⇐ absolute refractory period

• A few miliseconds after that it is the relative refractory period. This is when an action potential can be initiated, but the strength of current required is larger than normal

• Refractory periods results from the same mechanism that terminates the impulse.

• Following the increasing of Na+ conductance, is the increasing conductance of K+, which in return decreases the conductance of Na+ ⇒ impulse terminated

• The cell`s ability to generate impulses recovers when both ionic conductances are back to their resting levels

Action potential…cont




Conduction of the nerve impulses would be far too slow if it was merely electrical

transmission through a wire

So how does the nerve impulses propagate along the axon?

Systems Neuroscience &Neurotechnology Unit The propagating generic axon

Systems Neuroscience &Neurotechnology Unit Unmyelinated axon

Continous conduction

Systems Neuroscience &Neurotechnology Unit Unmyelinated axon…cont

[F/m]length unit per ecapacitanc membranec/m][length unit per resistance axial

[V] voltagethreshold

[A/m]length unit per current sodium maximum [m/s] impulse nerve theofvelocity

m

max

=Ω=

=

==

i

th

Na

rV

iv

where

thmi

Na

Vcri

v 2max=

Systems Neuroscience &Neurotechnology Unit Myelinated axon

The action potential at one node is sufficient to excite a response at the next node, so the nerve signal can propagate faster by these discrete jumps rather than by the continuous propagation of depolarization/repolarization along the membrane.

This enhanced signal transmission is called saltatory (sprunghafte) conduction.


dv 6≈

m][diameter axon [m/s] impulse nerve theofvelocity

µ==

dv

where

Myelinated axon…cont






Systems Neuroscience &Neurotechnology Unit Synapse

• Synapse : Site of transmission of electric nerve impulses

between two nerve cells or between a nerve cell and a gland or muscle cell.

Electrical Synapse Chemical Synapse


Electrical synapses

Impulse travels across the synaptic gap (cleft/junction) by direct electrical conduction.

The gap (or cleft) between the neurons are approx. 3,5nm. Gap junctions consist of two half-channels - one in each cell.

The transmission is bidirectional, but not in all cases. Some kinds of electrical synapses can block one direction (rectifying synapses).

About 10 times faster than chemical synapses.

Synapse…cont

Electrical synapses (also called gap junctions) are quite rare compared to the number of chemical synapses. They are frequently found in the vertebral retina and cerebral cortex.


Chemical synapses At chemical synapses, impulses are transmitted across synaptic gap via chemical substances called neurotransmitters.

The gap (or cleft) between the neurons are approx. 20-50nm.

The transmission in unidirectional.

Chemical transmission seems to have evolved in large, complex vertebrate nervous systems, in which multiple messages must be transmitted over long distances.

Synapse…cont


• Synapses can be :

a) presynaptic axon to the postsynaptic dendrite

(axodendritic)

a) Axon to axon (axoaxonic)

a) Axon to cell body (axosomatic)

a) Dendrite to dendrite (dendrodendritic) Most common is the axodendritic.

Synapse…cont


Neurotransmitters are released from the presynaptic cell, diffuse across the synaptic clefts, and bind to receptors on the postsynaptic cell. Chemical synapses are found only between nerve cells or between nerve cells and the gland cells and muscle cells that they innervate.

Synapse…cont


• Typical chemical synapses, regardless of what kind of transmitters substance they release, will cause a passive (without action potential) increase or decrease in the postsynaptic membrane (commonly dendrite) potential.

• In excitatory synapses, passive depolarization is called excitatory postsynaptic potential (EPSP)

• In inhibitory synapses, passive hyperpolarization is called inhibitory postsynaptic potential (IPSP)

• EPSPs and IPSPs from numerous synaptic inputs on the dendrites or cell body will accumulate at the axon hillock.

• At any given moment, the collective influence from these inputs will determine whether or not an action potential will be initiated at the axon hillock and propagated along the axon.

Synapse…cont


• EPSP occurs when transmitter causes an inward movement of +ve charge, by increasing the membrane conductance to Na+, K+ and Ca++

might or might not be large enough to depolarize the

postsynaptic neuron to its firing threshold a postsynaptic neuron has thousands of dendrites receiving

neurotransmitters from different presynaptic cells hence, a neuron‘s firing depends on the sum of EPSP

minus the IPSP from other dendrites • IPSP can occur as a result of increased conductance either for

outward movement of +ve charge (K+) or for inward movement of –ve charge (Cl-)

Synapse…cont


A postsynaptic neuron has thousands of dendrites receiving neurotransmitters from different presynaptic cells

Synapse…cont


Spatial integration -ripples can collide and combine


Spatial integration -ripples can go up (excitatory) or down (inhibitory)


Spatial integration -ripples can combine


Another look at spatial summation


Neurotransmitters

⇒ chemicals that allow the movement of information from one neuron across the gap between it and the adjacent neuron.

Neurotransmitter substances :

Substances are amino acids or derivatives of amino acids

• Acetylcholine (Ach) : very common in the CNS and PNS • Norepinephrine (NE, also known as noradrenaline NA): released by

neurons that control muscles in places such as the intestine, blood vessels, etc.

• Dopamine (DA) : important in parts of the brain • Serotonin (5HT) • Glutamate (GLU) : important in retina • Glycine (GLY) • Gamma-amino butyric acid (GABA) • etc


• Neurotransmitters can either be excitatory or inhibitory

• With the exception of GLY and GABA, which are almost always inhibitory

• And GLU, which is always excitatory

• In the cortex, GLU is the main excitatory and GABA the main inhibitory neurotransmitter

• The monoamine neurotransmitters, DA, NE and 5HT, are broadcast by certain midbrain regions out to vast areas of the cortex and the limbic system. Sometimes, they have found to modulate GLU or GABA

Neurotransmitters…cont


• Cholinergic synapses uses ACh as their neurotransmitter

• In this type of synapse, when depolarization occurs in the presynaptic membrane, it increases the movement of the calcium ion (Ca++)

• This Ca++ stimulates the release of neurotransmitter from the vesicles.

• Calcium also plays a variety of other roles in neurochemistry. Together with the cyclic nucleotides (cAMP and cGMP), they are the “second messenger” in plastic changes at the synapses.

Neurotransmitters…cont

Systems Neuroscience &Neurotechnology Unit Neurotransmitters…cont

• Not all chemical substances present in the brain are actual neurotransmitters.

• But, they still are important in modulating cellular reactions. For example, cyclic nucleotides and neuroactive peptides.

• The peptides include endorphins. A morphin-like substance, which are associated with positive reinforcement (relieves pain or having the sense of well-being/happy)

Systems Neuroscience &Neurotechnology Unit Hodgkin-Huxley Model (1952)

• Hodgkin-Huxley model is a model that tries to simulate the biological nerve cell.

• What is a model? – Mathematical representation. – Logical description of a system.

• Hodgkin and Huxley experimented on the giant axon fibers of squids and discovered how the signal is produced within the neuron


The space clamp Marmont (1949) and Cole (1949) developed the space clamp technique to maintain a uniform spatial distribution of Vm over a region of the cell where one tried to record currents

This was accomplished by threading the squid axon with silver wires to provide a very low axial resistance and hence eliminating longitudinal voltage gradients

The voltage clamp Cole and colleagues developed a method for maintaining Vm at any desired voltage level

Required monitoring voltage changes, feeding it through an amplifier to then drive current into or out of the cell to dynamically maintain the voltage while recording the current required to do so


saline solution : Kochsalzlösung


Simplified principle and electric model of the space clamp measurement procedure. (A) The physical structure of the device that ensures axial uniformity, hence current flow that is in the radial direction only. The problem is thus reduced to one dimension. (B) The total current (im), through the membrane (per unit length), consisting of the components of ionic current imI and capacitive current imC.

space clamp measurement


voltage clamp measurement

Realistic voltage clamp measurement circuit. Current is applied through electrodes (a) and (e), while the transmembrane voltage, Vm, is measured with electrodes (b) and (c). The current source is controlled to maintain the membrane voltage at some preselected value Vc


equivalent circuit of the HHM

http://butler.cc.tut.fi/~malmivuo/bem/bembook/04/04x/0410x.htm


Hodgkin-Huxley-Equations

probabilities that specific ion channels are open or closed

In these equations V' = Vm - Vr, where Vr is the resting voltage.

Systems Neuroscience &Neurotechnology Unit Hodgkin-Huxley-Equations


potassium conductance as function of time

sodium conductance as function of time

potassium & sodium conductance, their sum, and the membrane voltage

Systems Neuroscience &Neurotechnology Unit Propagating nerve impulse

Under steady state conditions Θ = propagation velocity [m/s]

K = constant [1/s] a = axon radius [cm]

ρi = axoplasm resistivity [Ωcm]

Hodgkin-Huxley-Equation for the propagating nerve impulse

A detailed derivation can be found in

Malmivuo & Plonsey. Bioelectromagnetism. Oxford University Press. 1995

Systems Neuroscience &Neurotechnology Unit Anode break

! hyperpolarization may cause an action potential ! reason: a current impulse which duration exceeds a particular time constant deinactivates the sodium (Natrium) channels and deactivates the potassium (Kalium) channels


HHsim: Graphical Hodgkin-Huxley Simulator (exe & Matlab) http://www-2.cs.cmu.edu/~dst/HHsim/

See also Christof's book http://www.klab.caltech.edu/~koch/biophysics-book/

Systems Neuroscience &Neurotechnology Unit Hodgkin, 1977

• Finally there was the difficulty of computing the action potentials from the equations which we had developed. We had settled all the equations and constants by March 1951 and hoped to get these solved on the Cambridge University computer. However, before anything could be done we learnt that the computer would be off the air for 6 months or so while it underwent a major modification. Andrew Huxley got us out of that difficulty by solving the differential equations numerically using a hand-operated Brunsviga. The propagated action potential took about three weeks to complete and must have been an enormous labor for Andrew. But it was exciting to see it come out with the right shape and velocity and we began to feel that we had not wasted the many months that we had spent in analysing the records.

Systems Neuroscience &Neurotechnology Unit Huxley, 1964

• The computations … were done by hand. This was a laborious business: a membrane action potential took a matter of days to compute and a propagated action potential took a matter of weeks. But it was often quite exciting. For example, when calculating the effect of a stimulus close to the threshold value, one would see the forces of accommodation – inactivation of the sodium channel and the delayed rise of the potassium permeability. Would the membrane potential get away into a spike, or die in a subthreshold oscillation? Very often my expectations turned out to be wrong, and an important lesson I learned from these manual computations was the complete inadequacy of one’s intuition in trying to deal with a system of this complexity.


Neurons connected to a computer chip

Systems Neuroscience &Neurotechnology Unit Cool links

• http://highered.mcgraw-hill.com/sites/0072437316/student_view0/chapter45/animations.html#

• http://www.klab.caltech.edu/cgi-bin/publication/reference.pl • http://www.getbodysmart.com/ap/nervoussystem/neurophysiology/acti

onpotentials/conductionrates/tutorial.html • http://butler.cc.tut.fi/~malmivuo/bem/bembook/v • http://www.klab.caltech.edu/~koch/biophysics-book/

http://highered.mcgraw-hill.com/sites/0072437316/student_view0/chapter45/animations.html

http://highered.mcgraw-hill.com/sites/0072437316/student_view0/chapter45/animations.html

http://www.klab.caltech.edu/cgi-bin/publication/reference.pl

http://www.getbodysmart.com/ap/nervoussystem/neurophysiology/actionpotentials/conductionrates/tutorial.html

http://www.getbodysmart.com/ap/nervoussystem/neurophysiology/actionpotentials/conductionrates/tutorial.html

http://butler.cc.tut.fi/~malmivuo/bem/bembook/v

Documents

BMT 823 Neural & Cognitive Systems Slides Series …...Prof. Dr. rer. nat. Dr. rer. med. Daniel J. Strauss BMT 823 Neural & Cognitive Systems Slides Series 1 Introduction and Basics