Module A - Brain machine interface Flashcards
Describe the observations from the 1870 study by Fritsch and Hitzig on stimulation of the cortex:
Electric excitability of the cerebellum in dogs and observations from battlefield open head wounds
1. Motor cortex was responsible for motor control
2. Areas of cortex connected to specific muscles
Limitation: Not very well localised
Describe the observations from the 1954 study by Wilder Penfield on mapping by stimulation:
Mechanisms of voluntary movement
Epilepsy study of proving different parts of the brain
Overlapping areas not well defined but recorded motor homunculus
Describe the observations from the 1968 study by Evarts on neural activity and force:
Relation of pyramidal tract activity to force exerted during voluntary movement
Single cell firing encodes force of contraction (neurons behave like muscles)
Describe Evarts observations on muscles with no load:
Flexors don’t fire during extension and fire during flexion (tonic EMG)
Extensors fire during extension and don’t fire during flexion (tonic discharge)
CTN active with agonist muscle
Describe Evarts observations on muscles with flexor load:
Flexors fire during flexion with increased firing rate and tonic discharge
Extensors don’t fire
CTN activity increases with increased load
Describe Evarts observations on muscles with extensor load:
Flexors don’t fire
Extensors fire during extension and fire at the end of flexion to hold against the load
No CTN activity because flexion movement results from relaxation of antagonist
Describe the observations from the 1983 study by Georgopoulos on neural activity and direction:
The relations between 2D arm movements and cell discharge in the primate motor cortex
Neurons represent direction (neurons fire when the arm moves in a particular direction, not in response to force)
Describe the directionality of firing and preferred directions:
Maximal firing rate occurs in a neurons preferred direction
What does the following equation calculate:
Firing rate=baseline + modulation depth (movement direction-preferred direction)
Predicts the max firing rate of neurons
Describe the observations from the 1988 study by Swchartz on preferred directions in 3D:
Primate motor cortex and free arm movements
Populations of neurons in preferred directions are uniformly distributed in space (all directions equally represented)
Describe the observations from the 1988 study by Georgopoulos on the population vector algorithm:
Coding of the direction of movement by a neuronal population
Take an average vector to give population vector which always points in the direction but have different firing rates
This depends on knowing scaling factor beforehand to normalise firing rate
Describe the certainty of the PVA:
Certainty detected by bootstrapping to form a confidence cone around the preferred direction vector
Calculate the preferred direction from random combinations of neurons activity from the multiple trials’ recording
Do this 100 times and report angle encompassing 95% of the resulting
A smaller angle and tighter confidence cone is desirable
Describe the 4 parameters of the PVA:
Normalised (firing rate, baseline, modulation depth)
Smoothened
Pooled (preferred direction)
Integrated
Describe the observations of the 1999 study by Kaheri, Hoffman and Strick on muscles and direction:
Monkey can move hand up/down in 3 wrist conformations
Substantial group of M1 neurons displayed changes in actvity that were muscle-like, and an even larger group displayed change sin activity that were related to change sin actual wrist movements in space (independent of muscle activity generated)
Both muscles and direction strongly represented in M1
Describe the other considerations (regions) needed to be taken into account during the muscles vs direction argument:
Primary motor cortex Supplementary motor area Premotor cortex Dorsolateral prefrontal association cortex Somatosensory cortex Posterior parietal cortex (area 5 + 7)
Describe the observations from the muscle vs direction study:
PMv (ventral premotor) definitely extrinsic
Muscles are definitely intrinsic
Most M1 cells behave like extrinsic cells
Most M1 modulation scales according to rotation
Describe the observations from the 2005 study by Graziano on more complex movements:
Arm movements evoked by electrical stimulation in the motor cortex of monkeys
Single electrode in M1
Measured movements from different start points
Stimulating a particualr location ends up in the same place regard less of starting position
Describe the observations from the 2007 sutdy by Churchland and Shenoy on complex signals:
Temporal complexity and heterogeneity of singlel-neuron activity in premotor and motor cortex
Describe the dynamically causative theory of neural network dynamics:
Individual neurons may not represent anything
Instead their activity is whatever needed to generate to correct movement
Neural networks show complexity
This makes neural prosthetics (limit in interpretation, not perfect representations)
Describe the observations from the 2012 study by Sussillo, Churchland, Shenoy on dynamic neural networks:
Neural state is represented as a location in space (1 axis for each positron of that neuron at that time)
What is the Fano factor?
Variability of a quantity between trials
Describe the observations from the 1997 study by Nicolelis on ensemble recording:
Reconstructing an engram-simultaneous, multisite, many single neuron recordings
Need calibration to come up with parameters (before PVA)
Real time interfacing requires recording from lots of neurons simultaneously
Describe the electrical recordings of ensemble recording:
Records EC voltages related by APs in a few neurons
Different neurons have different shaped waveforms
Describe the calibration (multiple linear regression) loop:
Start –> (better) parameters –> present target –> allow some movement, assume intention is towards target –> regress firing rates against target
This provides visual feedback of performance to subject
Describe the observations of the 2002 study by Taylor and Schwartz on 3D cursor control:
Neurons aren’t exactly cosine tuned
Cosine-ness of units improves overtime and performance improves
The brain is adapting and learning the new tool its been given, and gets better at adapting behaviour (become more cosine-like)
Describe the observations of the 1992 study on Utah electrode arrays in creating electrodes:
- Cut keffs
- Fill with glass
- Cut out electrodes
- Coal tips with metal
- Insulate with rest of silicon with polyimide insulation
Cut skull open, stick into hand/arm area of motor cortex, screw down connectors, close skull up
Describe the first human implant in 2006 by Hochberg and Donoghue:
Tetraplegia Record single neurons Modulated by movement intent Consistent in response Neurons show directional tuning Neurons used in cursor control
Allows open/closing hand, move arm and shoulder
Describe the observations from the 2008 study by Velliste and Schwartz on self-feeding:
Cortical control of a prosthetic arm for self-feeding
PVA to achieve robotic arm and marshmallow
Neural control in gripper limited
Describe the decoding technology of OLE and PVA:
PVA
With small numbers, preferred directions may not be uniformly distributed
Biased sampling fives biased encoding
Optimal linear regression:
OLA fixes preferred directions to give unbiased decoding
Describe the offline and online performance in terms of OLE and PVA:
Offline performance reconstructing trajectories OLE»PVA
Online performance in brain control OLE~PVA
Describe the observations of the 2013 study by Collinger on 7D control:
High performance neuroprosthetic control by individual with tetraplegia Spinocerebellar degeneration (lost all movement below neck)
Describe ‘looking forward’ in brain machine interfacing:
3-axis force sensor
Receptive field of each electrode with somatotopical arrangement
