Related Experiment Video
Updated: Jun 18, 2026
Probing the Brain in Autism Using fMRI and Diffusion Tensor Imaging
Published on: September 12, 2011
Dominant cross-frequency analysis-based early diagnosis of Autism Spectrum Disorder in pediatrics using FRHIS and
V Anithalakshmi1, R Thiagarajan2
1Department of Computer Science Engineering, Prathyusha Engineering College, Anna University, India.
Insights
This study introduces a novel method for early Autism Spectrum Disorder (ASD) detection in children under two years old. The approach utilizes Electroencephalogram (EEG) analysis and achieves 99.03% accuracy, significantly improving pediatric ASD diagnosis.
Area of Science:
- Neuroscience
- Pediatric Medicine
- Biomedical Engineering
Background:
- Autism Spectrum Disorder (ASD) is a neurodevelopmental condition impacting cognitive, social, and behavioral skills in children.
- Early diagnosis, especially in children under two years old, is crucial for effective intervention.
- Existing methods often overlook dynamic frequency band variations in Electroencephalogram (EEG) signals, leading to misclassification.
Purpose of the Study:
- To propose a novel, cross-frequency analysis-based method for early ASD detection in pediatric patients.
- To address the limitations of previous studies by incorporating dynamic frequency band variations.
- To achieve high accuracy in identifying ASD in children under two years old.
Main Methods:
- Collected EEG signals from children under two years old.
- Preprocessed EEG data using EWI-BPF for artifact removal and Variation Frechet Distribution Mode Decomposition (VFDMD) for signal decomposition.
- Employed Fuzzy Root Hesitant Inference System (FRHIS) for frequency band determination and feature extraction, including cross-frequency analysis via Phase Amplitude Method (PAM).
- Utilized a Convolutional Restricted Boltzmann Machine (CRBM) for optimal feature selection and a Deep Squared Tau Convolutional Neural Network (DSTCNN) for final ASD classification.
Main Results:
- The proposed method achieved a superior classification accuracy of 99.03% for pediatric ASD detection.
- Demonstrated improved performance compared to traditional ASD detection methods.
- Successfully integrated cross-frequency analysis to enhance diagnostic accuracy.
Conclusions:
- The developed FRHIS and DSTCNN-based approach offers a highly accurate and effective solution for early ASD detection in young children.
- The study highlights the importance of considering dynamic frequency band variations in EEG for improved ASD diagnosis.
- This method holds significant potential for advancing pediatric neurodevelopmental disorder screening.
Abstract:
Autism Spectrum Disorder (ASD) is a prevalent neurodevelopmental condition that affects children's cognitive, social, and behavioral skills. Thus, early diagnosis of ASD for children below 2 years old is very significant. Nevertheless, the prevailing works overlooked the depth of the dynamic frequency band variation during ASD detection, leading to misclassification. Therefore, a dominant cross-frequency analysis-based early ASD detection in pediatrics using a Fuzzy Root Hesitant Inference System (FRHIS) and Deep Squared Tau Convolutional Neural Network (DSTCNN) is proposed in this paper. Primarily, the Electroencephalogram (EEG) signals of children below 2 years old are gathered. Then, the artifacts are removed using the EWI-BPF. Afterward, by using Variation Frechet Distribution Mode Decomposition (VFDMD), the signals are decomposed. Then, by utilizing the FRHIS, the frequency bands are determined, followed by non-linear variables and feature extraction. In the meantime, the cross-frequency analysis is carried out by estimating the modulation depth using the Phase Amplitude Method (PAM) and determining the cross-frequency strength. Then, to select optimal features, the CRBM is applied. Lastly, the ASD classification is carried out using DSTCNN. When analogized with other traditional works, the proposed work attained superior performance in detecting ASD for pediatrics with a higher accuracy of 99.03%.
