CNN-FEBAC: A framework for attention measurement of autistic individuals

Electroencephalogram (EEG) signals are a cost-effective and efficient method to measure and analyse neurological data and brain-related ailments. Autism Spectrum Disorder (ASD) is a globally prevalent neurological disorder that is of significant concern to the medical research community regarding its diagnosis and treatment. Artificial Intelligence (AI) algorithms utilized to study EEG signals of autistic patients have shown promising results to make progress in this domain. In this study, the authors have used the BCIAUT-P300 dataset for attention measurement and analysis of EEG signals of autistic patients. The dataset comprises the EEG signal data of ASD patients when they are exposed to external stimuli in a controlled environment. The authors propose a Convolutional Neural Network based Feature Extractor for BCI Attention Classification (CNN-FEBAC) framework to achieve the research objective of predicting the response of ASD patients by studying their EEG signal recordings. The CNN-FEBAC framework consists of a feature extractor architecture followed by a shallow classifier to predict the patient's response to the stimuli. The proposed model was evaluated using performance metrics such as — confusion matrix, accuracy and F1 scores. The best accuracy achieved by the proposed model was 91%. The authors have explored and described the limitations of previously established methods and highlighted the performance improvements achieved with the proposed CNN-FEBAC framework. The authors further highlight the challenges encountered in the study and suggest the scope for improvement.