Roman

Project Title

Simulation of Cone Responses for Photosensitive Epilepsy

Introduction

Photosensitive epilepsy occurs in patients who experience seizures when presented with flashing lights. In 1997, thousands of children reported being affected by watching a pokemon episode in Japan. While many these symptoms did not turn out to be seizures, 700 of these children did experience epileptic seizures caused by flashing lights in this pokemon episode [1].

Background

Epidemiology

While only 4% of adults affected by epilepsy suffer seizures triggered by visual stimuli, photosensitive epilepsy affects 10% of children who have been diagnosed with epilepsy [2]. Not all visual stimuli are equally linked to photosensitive epilepsy. Rapidly flashing red lights are more likely to trigger photosensitive epileptic seizures, perhaps due to how they stimulate long cones [3]. Additionally, multicolor modulations are also highly epileptic, specially red and blue modulations at 15 hz [2].

Electrophysiology

Using EEG, doctors can non-invasively record brain activity in epileptic patients. The photoparoxysmal response (PPR) is an abnormal and amplified pattern of brain activity observed only in epileptic patients while observing flashing stimuli [4]. Recent unpublished data shows that PPRs are observed in more patients when observing a red light flashing stimulus compared to when observing other light colors [5].

Research Goals

While researchers have identified the stimuli and brain responses associated with epilepsy, elucidating the mechanisms that transform epileptic stimuli into epileptic cortical responses is a challenging task. Studying how the human eye processes stimuli and transforms them into electrical signals is critical to gain a better understanding of photosensitive epilepsy. However, this is difficult without using invasive methods. Nonetheless, recent tools designed to carry out computer simulations of the human eye may help.

Questions

Using the Image System Engineering Toolbox for Biology (ISETBIO), a toolbox to simulate processes at the front-end of the visual system [6], we ask:

1. How does the human eye process stimuli that may trigger epileptic seizures?

2. Could cone activity intensify when processing epileptic stimuli (i.e. red lights)?

Hypotheses

More specifically, we test the following hypotheses:

1. Compared to M-cones and S-cones, L-Cones absorb more photons in red stimuli.

2. PPRs caused by colored stimuli are correlated with cone absorption levels.

3. Red flashing stimuli will trigger Photocurrents with a larger dynamic range in L-Cones compared to M-cones and S-cones.

Methods

The Wikipede edits Myriapoda.

To test our hypotheses, we used ISETBIO to simulate how the human eye process stimuli delivered by a xenon light source.

Stimuli

Six different colored filters were applied to the xenon light source. The figure on the right shows the spectral power distributions (SPDs) for the filtered xenon light (solid lines; the black SPD corresponds to the white light), and the cone sensitivity curves (dashed lines; red: L-cones, green: M-cones, blue: S-cones) as reference. Stimuli had an average luminance of 100 $cd/m^2$.

Eye simulations

We used the ISETBIO toolbox to process the xenon light stimuli with the default human optics. To create flashing stimuli we generated sequences where the xenon light filters were sinusoidally modulated by a black screen.

Measurements

To characterize cone absorptions, we calculated the cone absorptions for each type of cone and each stimulus color with duration of 5ms.

To positively correlate PPR incidence in epileptic patients and levels of cone absorption for different stimuli colors, we found the least squares linear regression coefficients that explain such positive correlation.

To characterize the transduction of colored stimuli over time, we calculated photocurrents while cones processed 10Hz flashing stimuli of different colors for 1.2 seconds.

Results

L-cones absorb more photons in red light compared to M-cones and S-cones

When assessing how different cone types absorbed light colors, we found that red light was absorbed the most by L cones. The figure bellow shows how other light colors (black bars corresponds to white light) were absorbed by L, M, and S cones. Note that L-cones also absorbed orange, yellow, green and white colors more than M-cones and S-cones.

PPRs do not linearly correlate with absorption levels

We plotted cone absorptions vs PPR percentages in order to assess whether a linear relationship between cone absorption and PPRs exists. The plots bellow show L, M, and S cones absorptions vs PPR percentage for each color (black line corresponds to white light). The rightmost plot shows the absorptions for all three types of cones summed vs PPR percentages.

The Wikipede edits Myriapoda.

Next, we tried finding a weighted sum of L, M, and S cones that could explain a linear relation between cone absorptions and PPR percentages. The goal was then to find a vector $x\in {𝐑}^{\mathbf{(}\mathrm{𝟒}𝐱\mathrm{𝟏}\mathbf{)}}$ that would minimize the squared error between ${\hat{y}}_{ppr}=Ax$ and $y_{ppr}$, where $A\in {𝐑}^{(6,4)}$ is a matrix whose columns are: (1) a vector of ones (to obtain a global bias term that explains the linear relationship), (2) the absorptions for L-cones, (3) the absorptions for M-cones, (4) the absorptions for S-cones. $y_{ppr}$ is a vector containing the true PPR percents known from empirical data [5]. After running the linear regression, the vector $x$ we found was:

$\left(\begin{array}{c}585e^{-3}\\ 0.0786e^{-3}\\ -0.0973e^{-3}\\ 0.0743e^{-3}\end{array}\right)$

In this $x$ vector the first element is the global bias for the linear regression, and terms 2, 3, and 4 are weights multiplying the absorptions in L-cones, M-cones, and S-cones, respectively. The figure on the right shows the resulting linear relationship between PPRs and weighted cone absorptions, as well as the line of best fit.

Red flashing stimuli trigger Photocurrents with a larger dynamic range in L-Cones compared to M-cones and S-cones.

As the figure bellow shows, the photocurrents for L-cones was larger when processing red stimuli flashing at 10Hz compared to the M-cones and S-cones.

The photocurrents triggered by the rest of the colored flashing lights (orange, yellow, green, blue, and white) can be seen in the figure below:

Conclusions

Red light is absorbed more by L-cones, compared to M-cones and S-cones

A weighted sum of L-cone, M-cone, and S-cone absorptions can explain a linear relation relationship between cone absorptions for different colors and the percentage of epileptic patents showing PPRs when stimulated with colored flashing lights.

Red light drives a photocurrent with more dynamic range in L-cones compared to M-cones and S-cones.

Appendix

Presentation Slides

The matlab code to run the simulations and generate the figures can be found here, and the relevant data needed to run the code can be downloaded from File:RomanPsych221data.zip

Acknowledgements

Thanks to Dora Hermes for her mentorship while carrying out this project.

Thanks to Trisha Lian for her help with project and coursework as TA in the class.

References

[1] Radford, B., & Bartholomew, R. (2001). Pokémon contagion: photosensitive epilepsy or mass psychogenic illness?. Southern medical journal, 94(2), 197-204.

[2] Parra, J. (2017). Epileptic Photosensitivity: Towards Implementation of Preventative Measures. In Converging Clinical and Engineering Research on Neurorehabilitation II(pp. 103-106). Springer International Publishing.

[3] Harding, G. R. (1998). TV can be bad for your health. Nature medicine, 4(3), 265-267.

[4] Fisher, R. S., Harding, G., Erba, G., Barkley, G. L., & Wilkins, A. (2005). Photic‐and pattern‐induced seizures: a review for the Epilepsy Foundation of America Working Group. Epilepsia, 46(9), 1426-1441.

[5] Kasteleijn-Nolst-Trenité, unpublished data.

[6] Brainard, D. H., Jiang, H., Cottaris, N. P., Rieke, F., Chichilnisky, E. J., Farrell, J. E., & Wandell, B. A. (2015, June). Isetbio: Computational tools for modeling early human vision. In Imaging Systems and Applications (pp. IT4A-4). Optical Society of America.