Abstracts

Rationale:
Divisive normalization is a powerful framework for understanding sensory and cognitive processes in health and disease. Here, a population of neurons provides a ‘normalization signal’ that modulates (via a divisive operation) a neuronal output. Normalization is held to be a canonical computation in the brain. Previously, we proposed that the normalization model encapsulates an excitation-inhibition imbalance in genetic generalized epilepsy (GGE). Specifically, a decreased normalization signal accounted for a difference in the amplitude of visual evoked responses in patients with GGE compared to healthy controls, consistent with abnormal neuronal suppression. However, the role of the normalization framework in understanding epilepsy is not yet firmly established. Recent work (Coen-Cagli and Solomon, J Neurosci: 39,2019, 7344-56) extended the normalization model and showed that visual neurons that are more strongly normalized fire more reliably. We study whether variability in visual evoked responses differs between GGE and controls in accord with the normalization model.
Method:
Twenty-one adult subjects (10 diagnosed with GGE and 11 healthy controls) participated. Subjects also participated in a previous study (Won et al., Clin Neurophysiol: 128, 2017, 340-8). Steady-state visual evoked potentials were recorded using high-density EEG nets. Visual stimuli consisted of a windowed 2 cycles per degree grating pattern contrast reversing at 7.5 Hz. Stimulus contrast was incremented from 1.3% to 94% in 10 equal log steps in each trial. Twenty trials were recorded. Fourier analysis extracted the 2nd harmonic complex response. For each EEG channel and contrast, a mean response vector was calculated. The projected amplitude onto the mean vector was taken as the amplitude response of each trial. Responses were averaged across 7 occipital channels. Trial-to-trial variability was quantified via the Fano factor (FF) and analyzed using a mixed-effect model incorporating stimulus contrast as within-subject factor and diagnosis as between subject factor.
Results:
In both patients and controls, the group mean FF decreased with stimulus contrast (F1,19 = 57, p < 10-6), consistent with previous findings in single neurons and the prediction of the normalization model (Fig. 1). Across contrasts, the FF of patients was significantly greater than that of controls (F1,19 = 7.67, p = 0.012). There was a significant interaction between diagnosis and contrast (F1,19 = 10.33, p = 0.0046).
Conclusion:
Population measures of visually driven responses in humans show a remarkable similarity to single neurons in the decrease in trial-to-trial variability with increasing contrast. This relationship may be understood in terms of the normalization model. Human subjects with GGE manifest a greater variability compared to healthy controls, suggesting a decreased normalization strength in epilepsy. Moreover, normalization acts differently in patients and controls as contrast increases. These results support the normalization as a framework to understand changes in sensory processing in the epileptic brain.
Funding:
:None