Scientists use AI to detect ADHD through unique visual rhythms (www.psypost.org)

A PLOS One study used a novel behavioral technique—random temporal sampling—plus machine learning to show adults with ADHD have a distinctive rhythm in visual perception. Participants (49 total: 23 ADHD, 26 controls; 17 of the ADHD group were on stimulants) performed a 200 ms word-recognition task in which signal-to-noise varied as a sum of sine waves; this produced per-subject “classification images” that map perceptual efficiency across time and frequency. Group differences were most pronounced at processing oscillations around 5, 10 and 15 Hz (particularly when the stimulus noise modulated at 30–40 Hz). A classifier trained on these features reached 91.8% accuracy (sensitivity >96%, specificity 87%) using only ~3% of features, and could also distinguish medicated versus unmedicated ADHD with ~91.3% accuracy (100% sensitivity for medicated participants). The result is significant because it points to a potentially objective, brief behavioral marker of ADHD tied to temporal dynamics of perception—suggesting a common atypical perceptual oscillation rather than purely heterogeneous causes. Key caveats: the sample was small and young, medication comparisons underpowered, and the link between classification images and neural oscillations hasn’t been directly validated with EEG/fMRI. If replicated in larger, diverse cohorts and combined with neuroimaging, random temporal sampling could become a scalable adjunct diagnostic tool and a window into the neural timing mechanisms underlying ADHD.