Towards Reliable Stain Transfer: An Iterative Data-Model Co-Optimization Framework Based on Multimodal Expert-Guided Assessment

2026-07-28

AI Summary

This work addresses the challenges of modeling consistency and interpretability in generating immunohistochemistry (IHC) images from hematoxylin and eosin (H&E)-stained counterparts under pixel-unaligned supervision. The authors propose DMCoStain, a novel framework that enhances staining accuracy and pathological structural fidelity through a data–model co-optimization mechanism and a multimodal expert-guided fine selection strategy (MEGFS). Innovatively integrating pathologist reasoning, the method constructs a visual–language model for IHC positive expression and leverages ImmunoInstruction—a large-scale VQA instruction dataset comprising 150K samples—to enable clinically meaningful data optimization and evaluation. The approach achieves state-of-the-art performance across diverse tissue types and biomarkers while delivering interpretable and clinically actionable generation and assessment capabilities.

Abstract

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.

Problem

• stain transfer
• histopathology
• IHC
• heterogeneous biomarkers
• pixel-unaligned supervision

Innovation

• stain transfer
• data-model co-optimization
• vision-language model
• expert-guided selection
• immunohistochemistry synthesis