Reconstructing high-order sequence features of dynamic functional connectivity networks based on diversified covert attention patterns for Alzheimer's disease classification
This work addresses brain disease classification, specifically for Alzheimer's disease, by introducing a hybrid deep learning approach that integrates attention patterns, representing an incremental improvement over existing methods.
The authors tackled Alzheimer's disease classification by proposing DCA-CRN, a method that combines convolutional recurrent neural networks with self-attention mechanisms to reconstruct high-order sequence features from dynamic functional connectivity networks, achieving improved prediction performance and generalization on ADNI and ADHD-200 datasets.
Recent studies have applied deep learning methods such as convolutional recurrent neural networks (CRNs) and Transformers to brain disease classification based on dynamic functional connectivity networks (dFCNs), such as Alzheimer's disease (AD), achieving better performance than traditional machine learning methods. However, in CRNs, the continuous convolution operations used to obtain high-order aggregation features may overlook the non-linear correlation between different brain regions due to the essence of convolution being the linear weighted sum of local elements. Inspired by modern neuroscience on the research of covert attention in the nervous system, we introduce the self-attention mechanism, a core module of Transformers, to model diversified covert attention patterns and apply these patterns to reconstruct high-order sequence features of dFCNs in order to learn complex dynamic changes in brain information flow. Therefore, we propose a novel CRN method based on diversified covert attention patterns, DCA-CRN, which combines the advantages of CRNs in capturing local spatio-temporal features and sequence change patterns, as well as Transformers in learning global and high-order correlation features. Experimental results on the ADNI and ADHD-200 datasets demonstrate the prediction performance and generalization ability of our proposed method.