Spike analysis lecture

Question 1

spike trains

Answer 1

• time series of action potentials
• can be considered all or nothing events

Question 2

spike waveforms are due to

Answer 2

differences in relative position, impedance of the electrodes and the type of the recorded neuron

Question 3

steps spike sorting

Answer 3

1. spike detection -> high-pass filter and thresholding
2. spike waveforms are summarized in a compact ‘feature vector’ (usually PCA)
3. vectors are devided into groups corresponding to putative neurons using cluster analysis
4. manual curation

Question 4

type 1 errors

Answer 4

• false positives or commission errors -> including a cluster of spikes belonging to a different one -> inter spike intervals

Question 5

type 2 errors

Answer 5

• false negatives or omission errors -> not all spikes fired by a neuron are grouped together

Question 6

kilosort

Answer 6

• generative model of raw electrical voltage
• high pass filtering (>300 Hz) and median subtraction across all electrodes (common average reference)
• remotion of correlated noise across channes (whitening)
• modelling mean spike waveforms with a singular value decomposition (SVD) of its spatiotemporal pattern
• creation of generative model
• once the model is created it uses a template matching to infer the position of the spikes

Question 7

ISI histograms and autocorrelation

Answer 7

• used to further classify spike data

1. regular -> fixed intervals lead to single peak and harmonics
2. bursting -> peaks at +/- Dt of a burst of Aps
3. irregular and Poisson -> pretty random spiking increases with Dt

Question 8

stimulus characterisation

Answer 8

• spike counting functions (PSTH)
• stimulus-response characterisation (spike-triggered average)
• stimulus-response characterisation (Tuning curves)

Question 9

spike counting function

Answer 9

• action potentials can be treated as stereotypical events occuring in time. an idealized version of a spike can be represented as a Dirac function