A Framework for Predicting Neurofeedback Treatment Response in ADHD Using EEG Functional Connectivity and Genetic
Ziyue Li1, Zhinuan Zhu1,2, Dong Li1
1Department of Otolaryngology, The Second Affiliated Hospital & Yuying Children's Hospital of Wenzhou Medical University, Wenzhou, 325027, Zhejiang, China.
Child Psychiatry and Human Development
|July 24, 2026
Summary
This study introduces a computational framework to predict neurofeedback (NF) treatment response in children with Attention-Deficit/Hyperactivity Disorder (ADHD) using EEG data. The model achieved 84.72% accuracy in identifying responders, paving the way for personalized ADHD treatment.
Area of Science:
- Computational Neuroscience
- Neuroimaging Analysis
- Machine Learning in Medicine
Background:
- Attention-Deficit/Hyperactivity Disorder (ADHD) is a prevalent neurodevelopmental disorder.
- Neurofeedback (NF) shows promise for ADHD treatment, but predicting individual response remains challenging.
- Objective prediction of NF treatment success is crucial for optimizing patient care.
Purpose of the Study:
- To develop and validate a computational framework for predicting neurofeedback (NF) treatment response in children with ADHD.
- To utilize electroencephalogram (EEG) functional brain connectivity during early NF treatment for classification.
- To identify key EEG features and channels predictive of treatment outcomes.
Main Methods:
- A six-stage algorithm was developed, including EEG artifact preprocessing, spectral feature extraction (alpha and beta bands), and functional connectivity estimation (Phase Locking Value).
- Dimensionality reduction involved statistical screening (Welch's t-test with FDR correction) and a Genetic Algorithm (GA) for channel selection, identifying a six-channel subset (C3, C4, Cz, Fz, Fp1, T6).
- An ensemble of machine learning classifiers was trained on reduced feature vectors, with performance evaluated using subject-wise cross-validation.
Main Results:
- The computational framework successfully classified eventual responders and non-responders among 60 children with ADHD.
- The optimized model achieved a classification accuracy of 84.72%.
- The study identified a discriminative six-channel EEG subset predictive of NF treatment response.
Conclusions:
- The proposed data-driven framework demonstrates potential for predicting individualized neurofeedback response in ADHD.
- This approach may facilitate personalized treatment strategies for ADHD patients.
- Further validation on independent clinical datasets is warranted to confirm the framework's generalizability.
