Objective: This study investigates the auxiliary role of resting-state electroencephalography (EEG) in the clinical diagnosis of attention-deficit hyperactivity disorder (ADHD) using machine learning techniques. Metho...Objective: This study investigates the auxiliary role of resting-state electroencephalography (EEG) in the clinical diagnosis of attention-deficit hyperactivity disorder (ADHD) using machine learning techniques. Methods: Resting-state EEG recordings were obtained from 57 children, comprising 28 typically developing children and 29 children diagnosed with ADHD. The EEG signal data from both groups were analyzed. To ensure analytical accuracy, artifacts and noise in the EEG signals were removed using the EEGLAB toolbox within the MATLAB environment. Following preprocessing, a comparative analysis was conducted using various ensemble learning algorithms, including AdaBoost, GBM, LightGBM, RF, XGB, and CatBoost. Model performance was systematically evaluated and optimized, validating the superior efficacy of ensemble learning approaches in identifying ADHD. Conclusion: Applying machine learning techniques to extract features from resting-state EEG signals enabled the development of effective ensemble learning models. Differential entropy and energy features across multiple frequency bands proved particularly valuable for these models. This approach significantly enhances the detection rate of ADHD in children, demonstrating high diagnostic efficacy and sensitivity, and providing a promising tool for clinical application.展开更多
基金This study received financial support from the Jilin Province Health and Technology Capacity Enhancement Project(Project Number:222Lc132).
文摘Objective: This study investigates the auxiliary role of resting-state electroencephalography (EEG) in the clinical diagnosis of attention-deficit hyperactivity disorder (ADHD) using machine learning techniques. Methods: Resting-state EEG recordings were obtained from 57 children, comprising 28 typically developing children and 29 children diagnosed with ADHD. The EEG signal data from both groups were analyzed. To ensure analytical accuracy, artifacts and noise in the EEG signals were removed using the EEGLAB toolbox within the MATLAB environment. Following preprocessing, a comparative analysis was conducted using various ensemble learning algorithms, including AdaBoost, GBM, LightGBM, RF, XGB, and CatBoost. Model performance was systematically evaluated and optimized, validating the superior efficacy of ensemble learning approaches in identifying ADHD. Conclusion: Applying machine learning techniques to extract features from resting-state EEG signals enabled the development of effective ensemble learning models. Differential entropy and energy features across multiple frequency bands proved particularly valuable for these models. This approach significantly enhances the detection rate of ADHD in children, demonstrating high diagnostic efficacy and sensitivity, and providing a promising tool for clinical application.
