MMIR-TCM Framework for Multimodal TCM Clinical Decision Support
July 3, 2026
MMIR-TCM integrates a Memory-SAM module for tongue extraction with a fine-tuned Qwen3-VL and a Qwen3-based RAG component. The architecture aims to bridge the semantic gap between visual tongue features and textual Traditional Chinese Medicine reasoning.
HOW THIS AFFECTS YOU
●
builderYou can study this three-stage architecture for combining specialized segmentation with RAG and MLLMs.
●
healthThis offers a pathway toward more reproducible and objective digital tongue inspection in TCM.