Lakshmi19921992/Multimodal-Alzheimer-s-Disease-Classification-Using-CA-CNN-BiLSTM-with-BAM-: Multimodal Alzheimer's Disease Classification Using Cross-Attention CNN–BiLSTM with Bottleneck Attention from Heterogeneous Unpaired Datasets

This release contains the implementation of the multimodal deep learning framework for Alzheimer's Disease classification. The framework integrates heterogeneous modalities including MRI imaging, PET imaging, speech-derived audio features, and structured clinical variables. Feature extraction is performed using ResNet-50 for imaging modalities and MFCC representations for speech signals. A CNN–BiLSTM architecture with Bottleneck Attention Module (BAM) and cross-attention based fusion is employed for multimodal representation learning and classification. This release corresponds to the experimental implementation used in the research manuscript titled: "Multimodal Alzheimer's Disease Classification Using Cross-Attention CNN–BiLSTM with Bottleneck Attention from Heterogeneous Unpaired Datasets" Version: v1.0