Researchers have developed a framework called MKG-CARE for medical image diagnosis that uses multimodal knowledge graphs to incorporate similar cases and external knowledge. The system constructs a structured diagnostic memory where diseases, images, and symptoms are organized hierarchically. It then retrieves similar cases from this memory and extracts their corresponding subgraphs. A Graph Attention Network aggregates the semantics within these subgraphs and aligns them with visual representations using bidirectional cross-modal attention. To mitigate retrieval noise, the system estimates each case's reliability based on prediction confidence and sample similarity, and reweights its contribution to the final prediction. --- Why it matters: This matters because it addresses a limitation of existing medical image diagnosis methods that rely on isolated visual evidence without considering similar cases or external knowledge. By incorporating these factors, MKG-CARE has shown consistent improvements over strong baselines in extensive experiments. Source: https://arxiv.org/abs/2605.22547