The following is a summary of “Machine learning-enabled detection of attention-deficit/hyperactivity disorder with multimodal physiological data: a case-control study,” published in the August 2024 issue of Psychiatry by Andrikopoulos et al.
Attention-Deficit/Hyperactivity Disorder (ADHD) is a neurodevelopmental disorder that persists from childhood into adulthood, with current diagnostic methods being time-consuming and reliant on subjective recall and clinician judgment.
Researchers conducted a retrospective study to determine if physiological measures such as Electrodermal Activity, Heart Rate Variability, Skin Temperature, neuropsychological assessments, and imaging techniques can reliably identify adult ADHD.
They involved 76 participants (32 with ADHD and 44 healthy controls) who completed Stroop tests while physiological data were collected using a multi-sensor wearable device. Univariate analysis identified significant responses, and the Informative k-Nearest Neighbors (KNN) algorithm filtered out less relevant data. A machine-learning decision pipeline incorporating Logistic Regression, KNN, Support Vector Machines (SVM), and Random Forests was used to detect ADHD.
The result showed the highest performance of the SVM-based model, achieving 81.6% accuracy and maintaining a balance between experimental and control groups, with sensitivity and specificity of 81.4% and 81.9%, respectively. Additionally, incorporating data from all physiological signals produced the best outcomes, indicating that each modality captures distinct aspects of ADHD.
Investigators concluded that physiological signals hold promise as diagnostic indicators for adult ADHD. Multimodal data from wearable devices offer the potential to enhance traditional diagnostic methods, warranting further research into their clinical applications and long-term impact.
Source: bmcpsychiatry.biomedcentral.com/articles/10.1186/s12888-024-05987-7