Brain Network Analysis and Visualization

Demo Website


The IBM System G Brain Network Analysis tool currently detects possible neuron locations and radius from calcium neuron imaging videos of mouse brains. The tool tries to normalize and register neuron events by compensating the background variants of brain movements and activities. It detects the firing of neurons. By mouse over an individual neuron, neural scientists can exam the luminance changes and detected firing of individual neuron.

The tool analyzes the frames that the neurons are active and construct the dependency graphs of brain neurons based on collective and causal activities of neurons based on the same stimulus.

The current dataset we are working on are from live mice looking at different image patterns:

The following figure shows an example of calcium brain neuron image and the detected neurons. The blue circles are detected neurons (or, more precisely, the "Regions of Interest -- ROIs"). The red circle indicates the signal analysis of the neuron that is being shown on the interface.

In the below figure, you can see the yellow circles are the detected neurons that are in the firing status. The left hand side shows the control of video playing of the original sequence, with the current frame number and the 1x and 5x playing speed. In the right hand side, the analyst can choose to see detected and/or active neurons on the background of the original video or the denoised/normalized video. In this figure, the mouse was seeing a moving diagonal pattern with the stripes angled at 45 degree and the direction of movement was 135 degree.

When an analyst clicks on a neuron, she can see the raw signal, processed signal, and the detection result of that region of interest throughout the entire experiment that consists of alternative periods of mouse seeing different patterns and seeing nothing. This function can help analyst have a clear view on the neuron activity through different stimulus.

At the bottom of the tool includes three canvases to allow analyst to drag and drop any period to compare the detected neuron network of that region. It can help analyst to examine whether the neurons show similar (or related) networks based on similar types of stimilus. The color of the neuron circles indicates a cluster (connected component) in that period. The linkage was created based on the firing at the same time.



This is a collaborative work of the Network Science Department in IBM Research and Prof. Rafael Yuste's Lab in Columbia University.