Electroencephalogram-Based ConvMixer Architecture for Recognizing Attention Deficit Hyperactivity Disorder in...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Diagnostics

Background:

• Attention Deficit Hyperactivity Disorder (ADHD) is a prevalent neurodevelopmental disorder affecting 5-10% of school-aged children globally.
• Early diagnosis and intervention are critical for improving patient and family quality of life.

Purpose of the Study:

• To introduce ConvMixer-ECA, a novel deep learning architecture for accurate ADHD diagnosis using electroencephalogram (EEG) signals.
• To evaluate the efficacy of ConvMixer-ECA compared to existing deep learning models and attention-based variants.

Main Methods:

• Development and application of the ConvMixer-ECA deep learning model, integrating ConvMixer with Efficient Channel Attention (ECA) blocks.
• Training and evaluation using EEG recordings from 60 healthy children and 61 children diagnosed with ADHD.
• Performance comparison with state-of-the-art models (EEGNet, CNN, RNN, LSTM, GRU) and t-SNE visualization for feature analysis.

Main Results:

• ConvMixer-ECA achieved a high accuracy of 94.52% in recognizing ADHD from EEG data.
• The integration of ECA significantly enhanced ConvMixer's performance, outperforming other attention-based models.
• ConvMixer-ECA demonstrated superior performance compared to established deep learning architectures.

Conclusions:

• ConvMixer-ECA effectively captures distinguishing patterns between ADHD and typically developing individuals through hierarchical feature learning.
• The proposed model shows significant potential as a valuable tool for assisting clinicians in the early diagnosis and intervention of ADHD in children.