Image Processing Using SVD and Frequency Domain Analysis: Image sequences as multivariate time serie

1
Image Processing Using SVD and Frequency Domain
AnalysisImage sequences as multivariate time
series
Chris Fall
University of Illinois at Chicago
(Andrew Sornborger, University of Georgia)
2
Outline

• A bit of background on the data - CCD based
  cortical imaging
• Movies as multivariate time series
• SVD and spectral analysis for
• de-noising and compressing data movies (MATLAB
  exercise)
• extracting stimulus specific response
  (MATLAB exercise)

3
Problem bridge fMRI and single/multiunit scales
to understand activity patterns in the cortical
microcircuit
?
single/ multiunit, extracellular
fMRI
Golgi stain of human PFC, Mrzljak et al., 1990
4
Yuste et al. bulk loading of fluorescent Ca2
indicators in brain slice
bath-applied membrane permeant form of Ca2
sensitive dye
action potentials...
allow voltage-gated Ca2 influx
fills many to most cells near surface
Mao et al 2002
-single cell resolution, unavailable with current
voltage sensitive dyes
5
optical probing of mouse (pre)frontal cortex
6
De-noising examples
raw
de-noised
2-p imaging of cortex D.Kleinfeld/ CSHL
bio-imaging course
Not 2-P images Not D. Kleinfeld
(movies)
7
(No Transcript)
8
(No Transcript)
9
Image Compression/De-noising
Reformat Data
1
SVD
2
Reconstruct Movie
3
10
We can perform Spectral Analysis on Coefficients
mtspectrumc.m
ftestc.m
1Hz stimulation
11
Spectral Analysis to Choose Components
Reformat Data
1
SVD
2
Spectrum Reconstruct Movie
3
12
(No Transcript)