Related Experiment Video
Updated: Aug 5, 2026
05:58
Application of an Amplitude-integrated EEG Monitor (Cerebral Function Monitor) to Neonates
Published on: September 6, 2017
Classification of neurodevelopmental disorders and typical development using deep learning and a portable patch-type
Byambadorj Nyamradnaa1, Maya Izumoto2,3, Yoshiko Iwatani2,3
1PGV Inc, PMO Nihonbashi 2 Building, 7th Floor, 2-15-5 Nihonbashi, Chuo-Ku, Tokyo, 103-0027, Japan.
Neuroimage. Reports
|July 28, 2026
Summary
A portable EEG device and deep learning model show promise for identifying neurodevelopmental disorders (NDDs) like autism spectrum disorder (ASD) and attention-deficit/hyperactivity disorder (ADHD) in children. This approach offers a more accessible diagnostic support tool for NDD screening.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Artificial Intelligence
Background:
- Autism spectrum disorder (ASD) and attention-deficit/hyperactivity disorder (ADHD) are common neurodevelopmental disorders (NDDs) that frequently co-occur, complicating diagnosis.
- Current diagnostic methods are subjective, lengthy, and rely on expert knowledge, limiting accessibility.
- Electroencephalography (EEG) shows potential as a biomarker but requires skilled technicians and can be stressful for children.
Purpose of the Study:
- To develop a more accessible diagnostic support tool for neurodevelopmental disorders (NDDs) in children.
- To utilize a portable EEG device with low participant burden and a deep learning model for NDD classification.
- To distinguish between typical development (TD) and NDD groups (ASD, ADHD, ASD + ADHD).
Main Methods:
- Resting-state EEG data were collected for 5 minutes using a portable, three-channel HARU-2 EEG device from 163 participants (87 TD, 76 NDD).
- A deep learning model, combining a 1D convolutional neural network and a transformer encoder, was developed to analyze EEG data.
- A 5-fold cross-validation strategy was employed for model evaluation.
Main Results:
- The deep learning model achieved an AUC of 0.713 and bACC of 67.5% for classifying NDD vs. TD.
- Exploratory analysis showed higher performance for specific NDDs: AUC 0.793 (ASD vs. TD), AUC 0.817 (ADHD vs. TD), and AUC 0.667 (ASD + ADHD vs. TD).
- The model demonstrated varying accuracy across different NDD phenotypes.
Conclusions:
- Portable EEG devices combined with deep learning models show potential as accessible adjunctive tools for NDD screening in children.
- This technology may help overcome limitations of current subjective and lengthy diagnostic processes.
- Further research can refine this approach for broader clinical application in neurodevelopmental disorder assessment.