基于HCT和临床数据的水肿分类多模态深度学习框架

AttentionMixer is proposed as a unified deep learning framework for the multimodal detection of brain edema, integrating structural head CT (HCT) with routine clinical metadata. While HCT provides rich spatial information, clinical variables such as age, laboratory values, and scan timing capture complementary context that may be overlooked or naively concatenated. AttentionMixer is designed to fuse these heterogeneous sources in a principled and efficient manner. HCT volumes are first encoded using a self-supervised Vision Transformer Autoencoder (ViT-AE++), eliminating the need for large labeled datasets. Clinical metadata are mapped into the same feature space and utilized as keys and values in a cross-attention module, where the HCT-derived feature vector serves a crucial role.