Neuronal Communication and Electroencephalography (EEG), Lecture notes of Neuroscience

Download Neuronal Communication and Electroencephalography (EEG) and more Lecture notes Neuroscience in PDF only on Docsity! Applied Neuroscience Computational Models of Sleep Fall 2017 Sleep Objective: Computational Models of Sleep Agenda: 1.  Neurobiology of Sleep •  Action Potentials and Conduction •  Origin of Extracellular Currents 2.  Computational Models •  Sleep Time-Frequency Spectra •  Seizure Prediction in Epilepsy Ion Intracellular Extracellular Normal Plasma Value K+ 150 5 3.5-5.0 Na+ 12 140 135-145 Cl- 10 105 100-108 Organic Anions 65 0 •  Difference in ion concentration between compartments gives rise to the resting membrane potential (RMP). Membrane permeability to these ions also influences the RMP. •  Transient changes from the RMP produce electrical signals which transmit information in nerve cells. Changes in the Membrane Potential Produce Electric Signals in Nerve Cells Terminology Associated with Changes in Membrane Potential •  Depolarization- a decrease in the potential difference between the inside and outside of the cell. • Hyperpolarization- an increase in the potential difference between the inside and outside of the cell. •  Repolarization- returning to the RMP from either direction. • Overshoot- when the inside of the cell becomes +ve due to the reversal of the membrane potential polarity. In the nervous system, different channel types are responsible for transmitting electrical signals over long and short distances: A. Graded potentials travel over short distances and are activated by the opening of mechanically or chemically gated channels. B. Action potentials travel over long distances and they are generated by the opening of voltage-gated channels. Gated Channels Are Involved in Neuronal Signaling Gated ion channels in the membrane open to a variety of stimuli: •  Mechanical force, eg. sensory neurons. •  Chemical ligands, eg. neurotransmitters. •  Voltage, eg. changes in the resting membrane potential. §  A graded potential depolarization is called excitatory postsynaptic potential (EPSP). A graded potential hyperpolarization is called an inhibitory postsynaptic potentials (IPSP). §  They occur in the cell body and dendrites of the neuron. §  The wave of depolarization or hyperpolarization which moves through the cell with a graded potential is known as local current flow. Graded Potentials Graded potentials travel through the neuron until they reach the trigger zone. If they depolarize the membrane above threshold voltage (about -55 mV in mammals), an action potential is triggered and it travels down the axon. Graded Potentials Above Threshold Voltage Trigger Action Potentials Frequency of Action Potential Firing is Proportional to the Size of the Graded Potential The amount of neurotransmitter released from the axon terminal is proportional to the frequency of action potentials. Dendritic synapse 20 ms D APs 50 e = T - T 100 200 300 400 Distance from the soma (um) Vsoma A c Somatic synapse _|4mv 10 ms Vdend |i mv Control TTX g e Qo Percentage of control EPSP W 8 = e e e e e® e e e @¢ @ e m uU o o (, 3 a o EPSPs and Action Potentials Source: Hausser et al, Science Vol. 291. 138-141 EPSP shunting depends on synaptic input kinetics. The rise and decay times differ between “fast” and “slow” EPSPs. Electroenchaplography (EEG) Current in the EEG measuring circuit depends on the nature and location of the current sources, on the electrical properties of the brain, skull and scalp and on location of both electrodes. Source: Nunez et al (1891) Test Your Understanding: EEG activity is thought to arise from which of the following? A.  Cortical layers I and VI B.  Axonal action potentials C.  Horizontal dipoles D.  Excitatory and inhibitory post-synaptic potentials Explanation: EEG activity arises from the outermost cortex layer I and does not directly capture axonal action potentials. EEG is most sensitive to post-synaptic potentials generated in the superficial layers of the cortex. Electrodes Electrodes are small metal discs that are places on the scalp in special positions. •  Each electrode site is labeled with a letter and a number •  Letter: F is frontal lobe and T is temporal lobe •  Number: Even number means right side of head and odd number means left side of head •  Can be made of: stainless steel, tin, gold or silver covered with a silver chloride coating Intra(cranial/cerebral) EEG Depth electrodes Grid electrodes Recording Methods of Extracellular Events Electroencephalography (EEG): •  Spatio-temporally smoothed version of the local field potential (LFP), integrated over a larger area •  Used in combination with structural MRI imaging Magnetoencephalography (MEG): •  Measures tiny magnetic fields outside that skull from currents generated by the neurons •  Non-invasive •  High spatio-temporal resolution Electrocorticography (ECoG): •  Uses subdural platinum-iridium or stainless steel electrodes to record electric activity directly from the surface of the cerebral cortex, thereby bypassing signal-distorting skull and intermediate tissue Voltage-sensitive dye imaging: •  Membrane voltage of neurons can be detected optically with voltage-sensitive proteins Simultaneously recorded LFP traces from the superficial (‘surface’) and deep (‘depth’) layers of the motor cortex in an anaesthetized cat and an intracellular trace from a layer 5 pyramidal neuron. Note the alternation of hyperpolarization and depolarization (slow oscillation) of the layer 5 neuron and the corresponding changes in the LFP. The positive waves in the deep layer (close to the recorded neuron) are also known as delta waves. iEEG, intracranial EEG. Recording Methods of Extracellular Events Computational Models of Sleep Prominent Computational Neuroscientists: Gyorgy Buzsaki, NYU Terry Sejnowski, UCSD Roger Traub, IBM What is the biological function of sleep? Why do we dream? What are the underlying brain mechanisms? What is its relation to anesthesia? Sleep and Memory Formation Memory Consolidation At the time of encoding, Pattee elimi litle With time, they become robust fo interference (Mueller and Pilzecker, 1900) PoC TB CTR Telot Brome mols Cee CMCC Crd (Ribot, 1882) Non-REM Stages y 3 rT Spindles Slow wave,delta Sleep and Memory Formation 11 pm lam = 3am 5am 7m awake REM e ci lf ply Stage 2 ks g stage3 |S ¢ stage 4 awake Stage 3 = 70 VIN yay SAM wh IS A, stage 1 _ stage 4 an, rng nt ree pat ee ttn vu Wif CAN ANY = stage 2 REM delta waves Ni, Serene eee PLO pan patting pp appa saree K-complex spindles Brain rhythms Alpha rhythm (8-13 Hz) appears at the occipital cortex when eyes close. [resting condition] {rolandic mu rhythm; temporal tau rhythm} Beta rhythm (13-30 Hz) is associated with alertness. Gamma rhythm (30-80 Hz) is related to sensory integration and feature binding. Theta rhythm (4-8 or 4-10 Hz) Delta rhythm (0.5-4 or 1-4 Hz) Sleep spindle (12-15 Hz or 7-15 Hz) {sigma rhythm} K complex (<0.5 Hz) {(very) slow oscillation} Sleep and Memory Formation Multitaper Spectrogram Sleep Stage Patterns NREM Slow-wave sleep Slow waves (a.k.a up/down Bele UU Intra-cell area 4 -56 mv we) yw" vA we) mn EEG-depth area 4 i i so i. Intra-cell VL 66 mv Neuroscienct37 (2006) 1087-1106 GROUPING OF BRAIN RHYTHMS IN CORTICOTHALAMIC SYSTEMS M. STERIADE* Laboratory of Neurophysiology, Laval University, Faculty of Medicine, aT Te Oe BCEAO af Characteristic patterns of the brain activities in the neocortex and hippocampus Awake Non-REM sleep REM Stage 1 Stage 2 Stage 3-4 sleep Cortex Alpha wave Spindle Delta wave Gamma wave Gamma wave K complex Theta wave Hippocampus Theta wave HVS High-voltage spike (HVS) with high-frequency ripple (~200 Hz) Theta wave Buzsaki, Neuroscience, 31, 551-70, 1989. Gottesmann, Neurosci. Biobehav. Rev. 16, 31-8, 1992. Steriade et al., Science, 262, 679-85, 1993. Steriade, Neuroscience, 101, 243-76. 2000.  ‘ ‘ Wins Neri ts rete ee - - = Sleep, and Memory formation Reactivation In rats, hippocampal activity patterns during behavior are related to patterns during subsequent periods of inactivity (Paviides and Winson, 1989; Wilson and McNaughton, 1994), - specifically, in hippocampal oscillations (kudrimoti et al., 23) - demonstrated to co-occur with slow-waves (Sirota et al., 2003; Battaglia et al., 2004) seen in the hippocampus and neocortex of rats (Qin et al., 1997; Ribeiro et al., 2004; Ji and Wilson, 2007) seen in non-human primate, in multiple ‘disconnected!’ sites, coordinated across hemispheres. (Hoffman and McNaughton, 2002) seen in other structures, maybe under other names (Arieli, Yuste/MacLean, Dan) The upshot of up states in neocortex: From slow oscillations to memory formation Memory Consolidation Trace Reactivation Slow wave Sleep Epilepsy: disorder of brain dynamics •  Characterized by recurrent seizures •  Associated with abnormally excessive or synchronous neuronal activity Current Treatment: •  Anti-Epileptic Drugs (undesirable side effects) •  Surgical Removal of Tissue Motivation for Seizure Prediction: •  Increase quality of life of epilepsy sufferers A robust seizure prediction algorithm requires machine learning. Models of Epilepsy: Animal and Computational Simulated normal Simulated seizure 0 O =, 100 = 20 ®D oD D> Da £ £ 3 200 6 40 > > 300 60 a 5 10 15 0 5 10 16 Time [s] Time [s] Simulated seizure Simulated normal o ° a N Voltage [V] an 5 Voltage [] o & | 3 a N 3 =) om 5 10 15 Time [sl Time [s] Simulated epoch 5 > E 0 5 : \ ‘ 0 10 15 20 seconds Spectrum - normal Spectrum - paroxysmal 0 0 a fan) [50 [50 0 10 20 30 40 50 O 10 20 30 40 50 Hz Hz Animal Models of Epilepsy Species Drosophila melanogaster (fruit fly) Danio rerio (zebra fish) Mus musculus (mouse) Canis familiaris (dog) Papio hamadryas (baboon) First epilepsy studies Dynamin mutant Pentylene- tetrazole Audio-genic Electro convulsive Photosensitive Number of neurons 100,000 100,000 (larvae) 71,000,000 160,000,000 (cortex) 11,000,000,00 0 Percentage of human genes 39% 63% 79% 81% 93% Cost per day <$0.01 ~$0.01 ~$0.20 $27.30 $19.75 Genetic Models of Seizures: •  Knockdown of genes •  SCN1A Mutants Non-Genetic Models of Seizures: •  Kainic acid (activates receptors for glutamate) •  Pilocarpine (compromises the blood-brain barrier) Computational Model of Absence Epilepsy On Left: Frequency Spectrum of Epilepsy Simulation There is intense activity in the 2-4 Hz range from bottom left graph, and the top right graph shows a peak at 3.47 Hz. On Right: Power Spectrum of Epilepsy Simulation It is clear to note the differences between normal data and seizure data visually as evident by spiking patterns. A seizure prediction algorithm can be based on energy analysis in frequency bandwidths of interest. Amplitude v. Frequency with Fourier Transform provides Power v. Frequency The Seizure Prediction Problem Seizure onset Observation window preictal phase intracranial EEG Extraction of features from EEG, pattern recognition classification + interictal phase ictal phase Review of Literature: •  Most methods implement 1D decision boundary •  Machine learning used only for feature selection Trade-off Between: •  Sensitivity (being able to predict seizures) •  Specificity (avoiding false positives) A decision boundary is the region of a problem space in which the output label of a classifier is ambiguous. Interictal phase: period between seizures, or convulsions, that are characteristic of an epilepsy disorder Preictal phase: state immediately before the actual seizure Ictal phase: physiologic state of seizure (Latin: ictus, meaning a blow or a stroke) Slide by Yann LeCunn Hypotheses •  patterns of brainwave synchronization: –  could differentiate preictal from interictal stages –  would be unique for each epileptic patient •  definition of a “pattern” of brainwave synchronization: –  collection of bivariate “features” derived from EEG, –  on all pairs of EEG channels (focal and extrafocal) –  taken at consecutive time-points –  capture transient changes •  a bivariate “feature”: –  captures a relationship: –  over a short time window •  goal: patient-specific automatic learning to differentiate preictal and interictal patterns of brainwave synchronization features [Le Van Quyen et al, 2003; Mirowski et al, 2009] interictal preictal ictal Patterns of bivariate features •  Non-frequential features: –  Max cross-correlation [Mormann et al, 2005] –  Nonlinear interdependence [Arhnold et al, 1999] –  Dynamical entrainment [Iasemidis et al, 2005] •  Frequency-specific features: [Le Van Quyen et al, 2005] –  Phase locking synchrony –  Entropy of phase difference –  Wavelet coherence Varying synchronization of EEG channels [Le Van Quyen et al, 2003; Mirowski et al, 2009] 1min of interictal EEG 1min of preictal EEG 1min interictal pattern 1min preictal pattern Examples of patterns of cross-correlation Slide by Yann LeCunn Example of Seizure Prediction True negatives False negatives False negatives True positives Hypothesis: Patient has seizure. Null Hypothesis: Patient does not have seizure. Reality: Patient does not have seizure. Type I Error (model predicts seizure but patient does NOT have seizure) False Positives Correct Outcome (model predicts no seizure and patient does NOT have seizure) True Negatives Reality: Patient has seizure. Correct Outcome (model predicts seizure and patient does have seizure) True Positives Type II Error (model predicts no seizure but patient does have seizure) False Negatives Next Time: Biophysical Models of Neurons Please bring laptops for programming demo with SpikerBoxes