Multi-kernel Learning Fusion Algorithm Based on RNN and GRU for ASD Diagnosis and Pathogenic Brain Region Extraction

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder impacting social and communication skills.
• Current ASD research often relies on single neuroimaging data types, potentially missing crucial complementary information.
• Effective classification and understanding of ASD pathogenesis require integrating multi-modal data.

Purpose of the Study:

• To develop a novel framework for enhanced autism spectrum disorder (ASD) classification using multi-modal neuroimaging data.
• To leverage Recurrent Neural Network (RNN) and Gated Recurrent Unit (GRU) for feature selection across different data modalities.
• To fuse selected features using a Multi-Kernel Learning Fusion (MKLF) algorithm for improved ASD detection and identification of relevant brain regions.

Main Methods:

• Implementation of a novel Multi-Kernel Learning Fusion algorithm based on RNN and GRU (MKLF-RAG).
• Utilizing RNN and GRU for feature selection from diverse neuroimaging data modalities.
• Applying the MKLF algorithm to fuse features for pathological mechanism detection and Region of Interest (ROI) extraction in ASD.
• Validation using the Autism Brain Imaging Data Exchange (ABIDE) database.

Main Results:

• The MKLF-RAG framework significantly enhances classification accuracy for ASD.
• Experimental results demonstrate superior performance of MKLF-RAG compared to existing methods across various evaluation metrics.
• The framework effectively identifies key Regions of Interest (ROIs) associated with ASD.

Conclusions:

• The proposed MKLF-RAG framework offers a powerful approach for ASD classification by integrating multi-modal neuroimaging data.
• This method provides valuable insights for the early diagnosis of ASD.
• The study highlights the importance of multi-modal data analysis in understanding ASD pathogenesis.