Multimodal Kernel-based Discriminant Correlation Analysis Data-Fusion Approach: An Automated Autism Spectrum Disorder Diagnostic System

Abstract

Abstract Autism Spectrum Disorder (ASD) diagnostic systems based on association of multimodal tools such as combination of Electroencephalogram (EEG) and eye-tracking have emerged as an analytical to provide objective biomarkers. However, the existing systems lack in providing robust feature set and knowledge of fusion approaches employed for feature redundancy. The present paper aims to reduce disorder homogeneity by proposing a multimodal diagnostic system that can incorporate cognitive data of ASD individuals. The paper fused computational, neural and visual cognitive data collected from three modalities-laptop-performance tool, EEG machine, and Eye-tracker, respectively. The multimodal features are analyzed via proposed Multimodal Kernel-based Discriminant Correlation Analysis (MKDCA) fusion approach. The proposed framework considered the distinct cardinality of the feature vectors and fused the group structure among multiple samples after ranking them in increasing order. The acquired fused feature set is evaluated using state-of-the-art machine-learning based classifiers to provide influential features and reduce adversarial features in ASD. The results reflected that proposed multimodal system with SVM classifier provided reduced fused feature set of 11 features from total 39 features that diagnosed ASD with 96% accuracy and 0.988 AUC(ROC). The correlation of reduced feature set identified by SVM with ASD clinical symptoms accounted by ADOS and MoCA reflected clinical relevance of proposed multimodal features. The proposed automated fusion-based system has the potential to classify disorder by reducing the disorder heterogeneity and stratifying ASD individuals into homogeneous sub-groups.