Optimal interval and feature selection in activity data for detecting attention deficit hyperactivity disorder

Area of Science:

• Neuroscience
• Artificial Intelligence
• Clinical Diagnostics

Background:

• Attention deficit hyperactivity disorder (ADHD) is a common neurobehavioral disorder in children and adolescents, characterized by inattention, impulsivity, and hyperactivity.
• Traditional diagnostic methods for ADHD can be time-consuming, leading to delays, increased undiagnosed cases, and higher healthcare costs.
• Objective, data-driven approaches using machine learning (ML) and deep learning (DL) are emerging as efficient tools for early ADHD diagnosis.

Purpose of the Study:

• To evaluate the impact of feature selection techniques on ML model performance for ADHD prediction using activity datasets.
• To compare the effectiveness of specific time-interval activity data versus broader intervals in identifying ADHD.
• To assess the predictive capabilities of five different ML models for ADHD diagnosis.

Main Methods:

• Activity datasets were utilized to train and test five distinct ML models.
• Feature selection techniques were applied to identify significant predictors within the activity data.
• The study analyzed performance metrics using both specific and broader time-interval activity data.

Main Results:

• Implementing precise feature selection improved model accuracy by 0.11.
• Activity data from morning and night intervals proved to be more significant predictors of ADHD.
• Specific time-interval data demonstrated crucial importance for accurate ADHD prediction.
• The Random Forest model achieved the highest performance: 84% accuracy, 79% precision, 85% F1-score, and 92% recall.

Conclusions:

• Feature selection and the use of specific time-interval activity data significantly enhance ML-based ADHD prediction.
• AI combined with activity data offers a promising framework to support clinical decision-making for early ADHD detection.
• Morning and night activity patterns are key indicators for identifying ADHD, suggesting targeted data collection can improve diagnostic efficiency.