In the quest for more effective diagnostic methodologies for Alzheimer’s disease (AD), the integration of multimodal imaging techniques with advanced machine learning models holds significant promise. This study introduces a novel diagnostic framework that combines Discrete Wavelet Transform (DWT)-based fusion of MRI and PET images with a deep learning architecture to enhance the accuracy of AD classification. Our model employs a 10-layer convolutional neural network (CNN) enhanced with channel-spatial attention mechanisms to extract and prioritize salient features from the fused images. For classification, an Ensemble Deep RVFL (edRVFL) is utilized, which leverages the strength of multiple RVFL networks to improve robustness and accuracy. We compare our model’s performance against traditional classifiers and other single-layer feedforward networks, demonstrating superior sensitivity, specificity, precision, and F1 scores. The results substantiate the efficacy of combining attention mechanisms with ensemble learning in a deep learning context, significantly outperforming existing state-of-the-art approaches in AD classification. The source code of the proposed model is available at https://github.com/rsharma2612/Attentive-CNN.