Unsupervised spike sorting with wavelets and super-paramagnetic clustering Rodrigo Quian Quiroga Div. of Biology Caltech.

Unsupervised spike sorting with wavelets and super-paramagnetic clustering

Rodrigo Quian Quiroga

Div. of Biology

Caltech

Problem: detect and separate spikes corresponding to different neurons

Goals:• Algorithm for automatic detection and sorting of spikes. • Suitable for on-line analysis.• Improve both detection and sorting in comparison with

previous approaches.

Outline of the method:I - Spike detection: amplitude threshold.

II - Feature extraction: wavelets.

III - Sorting: Super-paramagnetic clustering.

Outline of the method

Simulated dataEx. 2

Simulation results

0/495

3/521

1/507

5/468

Misses

Number of missesExample # Nr. of

[noise level] spikes Wavelets PCA Classic

Ex. 1 [0.05] 474 0 0 16

[0.10] 521 1 6 25

[0.15] 482 1 9 69

[0.20] 490 6 7 280 (2)

Ex. 2 [0.05] 510 0 5 20

[0.10] 468 9 66 247 (2)

[0.15] 462 98 297 (1) 316 (1)

[0.20] 517 193 (2) 329 (1) 366 (1)

Ex. 3 [0.05] 495 0 1 20

[0.10] 484 65 55 223 (2)

[0.15] 479 310 (1) 310 (1) 310 (1)

[0.20] 520 344 (1) 344 (1) 344 (1)

Ex. 4 [0.05] 507 1 32 276 (1)

[0.10] 486 170 (2) 195 (2) 318 (1)

[0.15] 507 251 (2) 313 (1) 313 (1)

[0.20] 490 310 (1) 310 (1) 310 (1)

Conclusions:

• We presented an unsupervised and fast method for spike detection and sorting.

• By using a small set of wavelet coefficients we can focus on localized differences in the spike shapes of the different units.

• Super-paramagnetic clustering does not require a well-defined mean, low variance, Normality or non-overlapping clusters.

Thanks!

Richard AndersenChristof Koch

Zoltan NadasdyYoram Ben-Shaul

Sloan-Swartz Foundation