Search papers, labs, and topics across Lattice.
This paper introduces DMCoStain, an innovative Data-Model Co-optimization framework designed to enhance stain transfer by iteratively refining both training data and model capabilities. By employing the Multimodal Expert-Guided Finer Selection (MEGFS) strategy, which leverages a novel IHC-positive-expression vision-language model, the framework significantly improves staining accuracy and interpretability across various biomarkers. Extensive evaluations reveal that DMCoStain achieves state-of-the-art performance in generating immunohistochemistry images from hematoxylin and eosin stains, marking a substantial advancement in computational pathology.
DMCoStain revolutionizes stain transfer accuracy by integrating expert-guided data refinement with cutting-edge vision-language modeling.
Histopathological examination primarily relies on hematoxylin and eosin (H&E) and immunohistochemistry (IHC) staining. Although IHC provides critical molecular information, it is costly and requires specialized expertise. Stain transfer provides an efficient alternative by computationally generating IHC from H&E images, but remains challenged by unified and interpretable modeling for heterogeneous biomarkers under pixel-unaligned supervision. We propose DMCoStain, a novel Data-Model Co-optimization framework for Stain transfer. It iteratively co-refines training data and model capability, improving staining accuracy and interpretability in both pathological and structural consistency. To refine training data in a clinically meaningful manner, it incorporates the Multimodal Expert-Guided Finer Selection (MEGFS) strategy, built upon a pioneering IHC-positive-expression (IPE) vision-language model (VLM) that emulates pathologist reasoning. To support MEGFS, we construct ImmunoInstruction, the first large-scale IPE instruction-following dataset with 150K VQA samples. Extensive experiments on multiple tissues and biomarkers demonstrate that DMCoStain achieves state-of-the-art (SOTA) accuracy. This paradigm offers strong practical value, and MEGFS also functions as a specialized evaluation tool for future model development. Dataset, code, and more details are in https://github.com/SikangSHU/DMCoStain.
