Autism Spectrum Disorder Classification Using a Hybrid Deep Learning Framework

The accurate classification of autism spectrum disorder (ASD) from magnetic resonance images (MRI) is crucial for accommodating early diagnosis and personalised treatment, ultimately leading to improved patient outcomes. This paper proposes a novel hybrid deep learning (DL) framework for ASD classification using functional MRI data. The proposed hybrid deep learning (DL) framework comprises three steps that integrate three advanced techniques to enhance classification accuracy. In the first step, the faster region-based Convolutional Neural Network (R-CNN) architecture is integrated with Soft Non-Maximum Suppression (NMS), which improves classification accuracy. In the second stage, an enhanced ResNet50 architecture is proposed for feature extraction. A Bidirectional-Gated Recurrent Unit (Bi-GRU) is incorporated to process the sequence data in the final step. The proposed hybrid DL framework demonstrated significant performance improvements over individual models, achieving an accuracy of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 5. 1 4 \%}$</tex>. A comparative analysis with other models further reveals the superiority of the proposed method.