Transcript No Slide Title

Brain
Project
Neurobench: a prototype for a
neurophysiological recording database
and analyses tool
Michael Arbib and Erhan Oztop: Database Issues (Kyoto) 10-00
1
Brain
Project
What is actually recorded ?
Experiment
Conditions
Trials
Spikes
Michael Arbib and Erhan Oztop: Database Issues (Kyoto) 10-00
2
Brain
Project
How much is available to the scientific community ?
• Usually in a publication a small fraction of the recorded
neurons are reported.
• Because of space limitations,
• Ambiguity of the neurons,
•Lack of quantitative analysis tools and/or
•Interest of the experimenter
• Experimental errors (incomplete trials etc.)
Michael Arbib and Erhan Oztop: Database Issues (Kyoto) 10-00
3
Brain
Project
What do the computational modelers need?
• Lot of data!
• Quantitative reports, answers to questions: when, how
much etc. in numeric terms.
• Ease of accessing data
• Using only publications for the basis of modeling is not
efficient/not feasible
• Very little reported
• Not enough data points to perform analysis
• Not easy!
Michael Arbib and Erhan Oztop: Database Issues (Kyoto) 10-00
4
Brain
Project
What kind of data do modelers need?
• Neurophysiological
Neurobench
• Anatomical
• Bibliographical
Neurohomolgy
database
• Behavioral
Michael Arbib and Erhan Oztop: Database Issues (Kyoto) 10-00
5
WWW
Access for Neurobench2.0
NeuronBench
Brain
Project
http://java.usc.edu/~erhan/nbench1.0
Functionality:
A tool for examining
cell records,
adjusting
time alignments, and
producing
histograms.
In terms of
implementation:
• Java applet
• Plain file structure
Data:
(not accessible to
public)
14 mirror neuron
recordings from area
F5. Supplied by the
Rizzolatti group in
Michael Arbib and Erhan Oztop: Database Issues (Kyoto) 10-00
6
Brain
Project
Is Hand-object interaction required for Mirror Neurons to fire?
Grasping observation
Observation of mimic of
grasping
Monkey grasps himself
Michael Arbib and Erhan Oztop: Database Issues (Kyoto) 10-00
7
Brain
Project
10 minutes Neurobench
DEMONSTRATION
Michael Arbib and Erhan Oztop: Database Issues (Kyoto) 10-00
8
Brain
Project
Neurobench improvements and new features
• Use of a commercial database
• New analysis tools
•
Outlier detection
•
Correlation analysis, especially for multi-electrode
recording studies
•
And any custom analysis (need help from
neuroscientists)
• Unification of different labs data sets
• Transparent data storing in the process of data acquisition.
• Security: different levels of access
Michael Arbib and Erhan Oztop: Database Issues (Kyoto) 10-00
9