基于多原型的自闭症分类OA

Autism spectrum disorder(ASD)is a complex and highly heterogeneous neurodevelopmental disorder whose etiology and neurobiological mechanisms remain largely unknown.Currently,clinical diagnosis relies primarily on behavioral observation for assessment,but this method is subjective and lacks objective biomarkers.Recent advances in machine learning in medical imaging area have provided new impetus for ASD research.Resting-state functional Magnetic Resonance Imaging(rs-fMRI)measures changes in blood oxygen levels in the brain,reflecting brain activity during non-task activity and providing spatiotem-poral information for quantifying functional connectivity between brain regions.Using rs-fMRI data,this paper proposes a convo-lutional neural network model based on multi-prototype learning for the classification of ASD patients and normal controls.This model effectively extracts topological structural features of the functional connectivity network and embeds category information into the learned brain network representation through prototype learning.The experimental results show that the proposed prototype learning-based diagnostic method outperforms state-of-the-art methods in classification accuracy and robustness,with the classi-fication accuracy rate increasing by more than 5 percentage points.This method combines convolutional neural networks with proto-type learning algorithms,providing a more effective new approach for understanding the diagnosis of autism spectrum disorders.