Machine learning approaches to uncover the neural mechanisms of motivated behaviour: from ADHD to individual differences in effort and reward sensitivity
Machine learning models decoded ADHD from EEG signals and linked motivation to specific white matter tracts.
A new PhD thesis from Dublin City University researcher Nam Trinh leverages machine learning to map the neural circuitry of motivated behavior, with direct implications for understanding conditions like ADHD. The 194-page work, published on arXiv, presents three core studies that combine electroencephalography (EEG), diffusion MRI, and structural MRI with advanced classifiers to move beyond subjective diagnoses. In the first study, machine learning models trained on task-based EEG data—specifically gamma-band power from fronto-central and parietal regions during a stop-signal task—successfully classified adults with ADHD, outperforming models using resting-state data.
The subsequent studies linked individual differences in motivation to specific brain structures. Using diffusion MRI, the research identified that white matter integrity, particularly in tracts connected to the Supplementary Motor Area (SMA), correlated with computational parameters of effort and reward sensitivity. A third study used structural MRI to show that grey matter volumes could robustly decode an individual's reward sensitivity and subclinical apathy levels through machine learning. Across all analyses, fronto-parietal brain circuits emerged as a central network for evaluating cost and benefit, governing goal-directed behavior.
This research provides a quantitative, data-driven framework that could translate into clinical tools. The identified neural features—from specific EEG signatures to structural brain correlates—serve as candidate biomarkers. This could lead to more objective diagnostic aids for ADHD and motivational disorders, reducing reliance on behavioral questionnaires alone. Furthermore, by pinpointing the specific circuits involved, the work lays a foundation for developing targeted neurotechnological interventions, such as neuromodulation therapies, tailored to an individual's unique neural profile.
- ML models classified ADHD using task-based EEG gamma power, outperforming resting-state data.
- Diffusion MRI linked white matter integrity in SMA-connected tracts to computational parameters of effort/reward sensitivity.
- Structural MRI and ML robustly decoded individual reward sensitivity and apathy levels from grey matter volumes.
Why It Matters
Provides objective neural biomarkers for ADHD and motivation disorders, paving the way for precise diagnostics and personalized brain-targeted therapies.