Machine learning-enabled detection of attention-deficit/hyperactivity disorder with multimodal physiological data: a
Dimitrios Andrikopoulos1, Georgia Vassiliou2, Panagiotis Fatouros3
1Feel Therapeutics Inc., 479 Jessie St., San Francisco, CA94103, CA, USA. dimitris@feeltherapeutics.com.
BMC Psychiatry
|August 5, 2024
Summary
Physiological data from wearable devices show promise in diagnosing adult Attention-Deficit/Hyperactivity Disorder (ADHD). This study found that multimodal physiological signals, including Electrodermal Activity and Heart Rate Variability, can help distinguish ADHD patients from controls with high accuracy.
Area of Science:
- Neuroscience
- Psychiatry
- Biomedical Engineering
Background:
- Attention-Deficit/Hyperactivity Disorder (ADHD) is a neurodevelopmental condition often persisting into adulthood.
- Current ADHD diagnostic methods are time-consuming, subjective, and rely on recall.
- Objective measures like physiological monitoring are being explored to improve ADHD diagnosis.
Purpose of the Study:
- To investigate the utility of physiological data (Electrodermal Activity, Heart Rate Variability, Skin Temperature) as indicators for adult ADHD.
- To evaluate the effectiveness of these physiological markers in distinguishing ADHD patients from healthy controls.
- To assess the potential of wearable sensor technology in ADHD diagnostics.
Main Methods:
- An observational, case-control study involving 32 adult ADHD patients and 44 healthy controls.
- Passive collection of physiological data during Stroop tests using a multi-sensor wearable device.
- Machine learning algorithms (Logistic Regression, KNN, Random Forests, SVM) applied to classify ADHD patients based on physiological signals.
Main Results:
- The Support Vector Machines (SVM) model achieved 81.6% accuracy in detecting adult ADHD.
- Optimal performance was obtained by integrating data from all physiological signals (Electrodermal Activity, Heart Rate Variability, Skin Temperature).
- The SVM model demonstrated balanced sensitivity (81.4%) and specificity (81.9%) for ADHD detection.
Conclusions:
- Multimodal physiological signals collected via wearable devices show significant potential as objective diagnostic indicators for adult ADHD.
- These findings suggest that physiological markers can complement traditional ADHD diagnostic approaches.
- Further research is needed to explore clinical applications and long-term implications of using physiological data in ADHD diagnosis and management.
