Purpose/Significance To develop a more effective magnetic resonance imaging (MRI) classification model, so as to improve the early diagnosis accuracy of Alzheimer’s disease (AD) and enhance the quality of life of patients. Method/Process Based on the Swin Transformer model, integrated with a multi-resolution feature fusion (MRFF) module, an improved network structure, Swin Transformer + MRFF, is proposed. Data augmentation and expansion strategies are adopted, and the problems of insufficient samples and category imbalance are solved. Through comparative experiments with other deep learning models, the advantages of the improved model in the recognition of AD are evaluated. Result/Conclusion The improved Swin Transformer + MRFF model significantly outperforms the base model on the OASIS-1 dataset, notably enhancing the recognition ability for mild dementia and non-dementia categories.