How Brain Electrical Signals Reveal Autism?

Scientists have found that quantitative EEG (QEEG) can tell apart children with autism from typically developing peers. They examined 19 scalp electrodes and computed three nonlinear measures: Lempel‑Ziv complexity, Tsallis entropy, and Rényi entropy. All three showed significant differences between groups, especially in visual‑processing areas. Machine‑learning models achieved over 90% accuracy in identifying autism. A t‑SNE visualization also revealed two distinct clusters, one for autistic children and another for typical peers. These findings suggest that complexity and entropy metrics, particularly from visual channels, could serve as objective biomarkers for faster diagnosis and for tailoring neurofeedback interventions that help the brain regulate its own electrical activity.

The three nonlinear metrics used in the study stem from information theory and statistical physics. Lempel‑Ziv complexity quantifies the number of "new sequences" in a signal, reflecting the brain's capacity to generate variable patterns. Tsallis entropy, unlike Shannon entropy, introduces a q parameter that emphasizes the tails of the distribution, making it sensitive to rare events typical of disordered brain activity. Rényi entropy, with its order α, allows different weighting of probabilities, highlighting scale differences across brain regions. In children with ASD, occipital regions exhibit higher complexity and entropy values, indicating greater variability and less predictability of electrical activity during visual processing. This may relate to the sensory integration challenges characteristic of autism. Machine‑learning classifiers such as support vector machines or neural networks combined these metrics in a multidimensional space, achieving over 90% accuracy in group separation. t‑SNE visualization reduces dimensionality while preserving neighborhood relationships, revealing non‑linear clusters that confirm the discriminative power of visual channels. These findings pave the way for targeted neurofeedback protocols: by modulating occipital activity, it may be possible to reduce abnormal complexity and improve brain network coherence, potentially alleviating some autism symptoms.