AIMAFE: Autism spectrum disorder identification with multi-atlas deep feature representation and ensemble learning

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Autism spectrum disorder (ASD) diagnosis relies on subjective behavioral criteria, often leading to delays.
• Functional magnetic resonance imaging (fMRI) shows potential for objective ASD identification but faces accuracy limitations.

Purpose of the Study:

• To enhance the accuracy of autism spectrum disorder (ASD) identification.
• To develop an objective diagnostic tool for ASD using neuroimaging data.

Main Methods:

• Utilized multi-atlas functional connectivity from fMRI data.
• Applied a stacked denoising autoencoder (SDA) for deep feature representation.
• Employed multilayer perceptron (MLP) and ensemble learning for classification.

Main Results:

• Achieved 74.52% accuracy in identifying ASD from fMRI data.
• Demonstrated high sensitivity (80.69%) and an AUC of 0.8026.
• Outperformed previous ASD identification methods.

Conclusions:

• The proposed method significantly improves ASD identification accuracy.
• This approach shows promise as an objective tool for clinical ASD diagnosis.