ADHD EEG live demo
Select a real 19-channel EEG recording below — each is a 2-second window from a child the model never saw during training. It is classified in your browser; nothing is uploaded.
Select a recording
Each is labeled with the group the child belonged to in the study: Control or ADHD. Select one to view its 19-channel EEG and the model's classification.
This model separates the study's two groups with about 92% accuracy on unseen subjects — a genuinely strong result, but on a small cohort and for group membership rather than diagnosis. See the write-up for the full figures and caveats.
What am I looking at?
Each row is one scalp electrode (Fp1, Fp2, F3, … following the international 10–20 layout), and each trace is 2 seconds of electrical brain activity sampled 128 times per second. The model reads all 19 channels jointly. It was trained solely on recordings from other children, so a correct classification here means the learned pattern transfers to individuals it has never encountered.