A Deep Learning Method for Autism Spectrum Disorder Classification Based on Multimodal Neuroimaging Data.

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Autism Spectrum Disorder (ASD) is a neurodevelopmental condition impacting social interaction and communication.
• Early and accurate diagnosis of ASD is crucial for effective intervention and treatment.
• Existing diagnostic methods may not fully capture the complex brain features associated with ASD.

Purpose of the Study:

• To develop and evaluate a predictive model for improved ASD classification using multimodal neuroimaging data.
• To investigate the efficacy of fusing functional magnetic resonance imaging (fMRI) and structural magnetic resonance imaging (sMRI) data.
• To enhance the comprehensive feature space for capturing subtle neuropathological signatures of ASD.

Main Methods:

• Proposed a multimodal feature fusion framework integrating fMRI and sMRI data.
• Utilized data from the ABIDE NYU site for model evaluation.
• Employed a five-fold cross-validation scheme to assess performance.

Main Results:

• Achieved an average accuracy of 82.63% in ASD classification.
• Obtained an Area Under the Curve (AUC) of 89.31%.
• Reported a sensitivity of 81.45% and a specificity of 82.86%.

Conclusions:

• Multimodal feature fusion significantly enhances the identification of ASD.
• The proposed approach offers a promising strategy for precise diagnosis of brain disorders.
• This framework supports early clinical decision-making and personalized treatment for ASD.