Abstract:

Alzheimer's disease is a prevalent dementia with a slow progression and complex imaging features. Traditional diagnostics is inefficient and accuracy varies. To address this problem, this paper proposes a classification method based on hybrid attention and multi-scale information fusion (3D HAMSNet). The method leverages image data and a convolutional neural network to enhances the model's attention to the hippocampus, amygdala, and temporal lobe by introducing the hybrid attention mechanism, and integrates multiscale spatial scale features of Alzheimer's disease by using a multiscale information fusion module based on dilated convolution and soft attention, enhancing early diagnosis and prediction. Finally, tested on 198 Alzheimer's patients, 200 with mild cognitive impairment, and 139 controls, it achieved 94.14% accuracy, 97.07% specificity, and 94.17% F1 score—improvements of 9.88%, 4.94%, and 10.17% over the baseline. It excels over existing methods, offering a new path for early Alzheimer's diagnosis.

Key words: Alzheimer's disease, convolutional neural network, hybrid attention, multiscale information fusion