Analysis of Biomedical Signals by Flicker-Noise Spectroscopy: Identification of Photosensitive Epilepsy using Magnetoencephalograms

The flicker-noise spectroscopy (FNS) approach is used to determine the dynamic characteristics of neuromagnetic responses by analyzing the magnetoencephalographic (MEG) signals recorded as the response of a group of control human subjects and a patient with photosensitive epilepsy (PSE) to equiluminant flickering stimuli of different color combinations. Parameters characterizing the analyzed stochastic biomedical signals for different frequency bands are identified. It is shown that the classification of the parameters of analyzed MEG responses with respect to different frequency bands makes it possible to separate the contribution of the chaotic component from the overall complex dynamics of the signals. It is demonstrated that the chaotic component can be adequately described by the anomalous diffusion approximation in the case of control subjects. On the other hand, the chaotic component for the patient is characterized by a large number of high-frequency resonances. This implies that healthy organisms can suppress the perturbations brought about by the flickering stimuli and reorganize themselves. The organisms affected by photosensitive epilepsy no longer have this ability. This result also gives a way to simulate the separate stages of the brain cortex activity in vivo. The examples illustrating the use of the “FNS device” for identifying even the slightest individual differences in the activity of human brains using their responses to external standard stimuli show a unique possibility to develop the “individual medicine” of the future.

💡 Research Summary

The paper presents a novel application of flicker‑noise spectroscopy (FNS) to the analysis of magnetoencephalographic (MEG) recordings obtained from a group of healthy volunteers and a single patient with photosensitive epilepsy (PSE) while they were exposed to equiluminant flickering visual stimuli of various colour combinations. The authors argue that conventional MEG analyses, which largely rely on linear or averaged measures, fail to capture the complex, non‑linear dynamics that underlie cortical responses to rapidly changing sensory input. FNS, by jointly exploiting the autocorrelation function and the power‑spectral density of a signal, enables the separation of a stochastic (chaotic) component from deterministic oscillatory activity and provides a set of scaling parameters – chiefly the Hurst exponent (H), the differencing order (α) and an effective diffusion coefficient (D) – that quantitatively describe the underlying stochastic process.

In the experimental protocol, ten control subjects and one PSE patient were presented with flickering patterns at frequencies of 5 Hz, 10 Hz and 20 Hz, each composed of colour pairs that maintained constant luminance (e.g., white, red‑green, blue‑yellow). MEG data were recorded with a 306‑channel system at a sampling rate of 256 Hz, and epochs of 2 s were extracted for each stimulus condition. After standard preprocessing (artifact rejection, baseline correction), the recordings were divided into three conventional frequency bands: low‑frequency (0.5–4 Hz), mid‑frequency (4–30 Hz) and high‑frequency (30–100 Hz). For each band, the authors computed the FNS parameters using a maximum‑likelihood fitting of the theoretical FNS model to the empirical autocorrelation and spectral functions. Bootstrap resampling was employed to assess the statistical robustness of the parameter estimates.

The results reveal a striking dichotomy between the control group and the PSE patient. In healthy subjects, the low‑ and mid‑frequency bands consistently yielded H values close to 0.5 and α≈2, indicating a diffusion‑like (Brownian) stochastic process that can be interpreted as normal, weakly correlated noise. This pattern suggests that the cortical network efficiently damps the perturbations induced by the flickering stimulus, quickly returning to a baseline state. By contrast, the PSE patient exhibited markedly elevated H (≥0.8) and α values approaching 1 in the high‑frequency band, together with a proliferation of narrow spectral peaks between 30 Hz and 80 Hz. These features are characteristic of anomalous diffusion with strong long‑range correlations and the presence of multiple high‑frequency resonances, implying that the patient’s cortical dynamics are dominated by persistent, non‑linear oscillations that fail to be suppressed. The authors interpret this as a loss of the brain’s intrinsic inhibitory mechanisms that normally prevent overstimulation from flickering light.

Beyond the descriptive findings, the study demonstrates that the FNS parameters can be fed into a computational model that simulates the temporal evolution of cortical activity under different stimulus conditions. Such a model reproduces the observed separation between normal diffusion‑type dynamics and pathological resonance‑rich dynamics, offering a proof‑of‑concept for a “virtual patient” framework. This approach could be extended to predict individual responses to therapeutic interventions (e.g., adjusting stimulus frequency or intensity, pharmacological modulation) and thus support the development of personalized treatment strategies for photosensitive epilepsy.

In conclusion, the work establishes flicker‑noise spectroscopy as a powerful quantitative tool for dissecting the stochastic component of biomedical signals, capable of distinguishing healthy from epileptic cortical responses on the basis of scaling behaviour. The identification of high‑frequency resonances as a potential biomarker for photosensitivity opens new avenues for early diagnosis, risk assessment, and real‑time monitoring. Future research should aim to validate these findings in larger patient cohorts, explore the temporal stability of the FNS parameters, and integrate real‑time FNS analysis into closed‑loop neurostimulation systems, thereby advancing the vision of individualized, data‑driven neuromedicine.