[Back to home]

Visualization techniques for multisite recordings of seizure propagation

Marcel Favini and David Stanley's class project for Intelligent Image Processing (2008)

Taught by Dr. Steve Mann. Advised by Dr. Berj Bardakjian.

Based on work by M. Derchansky, D. Rokni, J.T. Rick, R. Wennberg, B.L. Bardakjian, L. Zhang, Y. Yarom, and P.L. Carlen

Approximately one third of patients with epilepsy are not responsive to anti-epileptic drugs. In these cases, a possible treatment is surgical ablation of the brain region responsible for initiating epileptiform activity. This region is called the focus and must be properly localized prior to surgery. However, such localization is difficult and often multiple foci are found.

To improve the localization of seizure foci, it is important to consider the rich and complex dynamics of seizure propagation. When analyzing data, not only are spatial and temporal dimensions important, but frequency information is vital as well. For example, it has recently been shown that within an epileptiform burst, seizure propagation can actually be bidirectional, with different frequencies propagating in opposite directions. (Derchansky et al., Neurobiology of Disease 2006).

A total of four dimensions are relevant in analysing brain EEG: time, space, frequency, and signal power (voltage). We compared two frameworks for such 4D visualization. Samples of the raw data obtained from four different electrodes spanning the septotemporal axis of the CA1 subfield are shown below.

A first approach was to to represent this data using animated wavelets. Here, time is plotted along the abscissa. Wavelet scale from the continuous wavelet transform is plotted along the ordinate (scalebars are 50ms each). Colour shows wavelet coefficeint amplitude. Progression of the animation in time represents spatial progression along the hippocampal septotemporal axis.

Our second approach was to design a fully interactive Matlab utility to represent the data in a cube with time along the x-axis, wavelet scale along the y-axis, and septotemporal displacement along the z-axis. The utility enabled user manipulation of the cube, allowing it to be rotated and "sliced." This approach has advantages over animation, in that two dimensions can be compared while the user controls navigation through the third. This project was continued by Diane Kostka and Rohan Karkhanis in the Bardakjian lab, with the ultimate goal of developing a practical visualization tool for seizure focus identification in a clinical setting.

Last updated June 2011