This study presents MARQO, an open-source, user-guided automated analysis pipeline that efficiently integrates elastic image registration, iterative nucleus segmentation, and unsupervised clustering to enable full-slide, single-cell-resolution multiparametric spatial tissue analysis, significantly improving the accuracy and efficiency of analyzing complex tissue samples.
Literature Overview
The article, 'Multiparametric cellular and spatial organization in cancer tissue lesions with a streamlined pipeline,' published in Nature Biomedical Engineering, reviews and summarizes the development and validation of MARQO—an open-source, user-guided automated analysis pipeline designed for full-slide, single-cell-resolution analysis of multiplex immunohistochemistry (mIHC) and multiplex immunofluorescence (mIF) images. The method integrates elastic image registration, iterative nucleus segmentation, unsupervised clustering, and graphical interface-guided cell classification, significantly enhancing the ability to resolve cellular spatial organization within the tumor microenvironment. The study validates MARQO's accuracy across multiple tissue types and staining techniques and applies it to clinical trial samples from a neoadjuvant immunotherapy study in hepatocellular carcinoma (HCC), revealing spatial enrichment patterns of CD8+ T cells in treatment responders. The paragraph is coherent and logical, ending with a Chinese period.Background Knowledge
In cancer immunotherapy research, understanding the cellular composition and spatial distribution within the tumor microenvironment is key to unraveling mechanisms of treatment response and resistance. Multiplex immunohistochemistry (mIHC) and multiplex immunofluorescence (mIF) technologies enable the detection of multiple protein markers while preserving tissue spatial architecture, providing a powerful tool for single-cell-level spatial phenotypic analysis. However, existing analysis pipelines are often fragmented, operationally complex, and reliant on manual validation, making high-throughput, standardized quantitative analysis challenging. Although artificial intelligence and deep learning methods have been introduced, their 'black-box' nature limits clinical acceptance, especially in the absence of pathologist verification. Additionally, issues such as image registration errors, ambiguous cell boundaries, and tissue heterogeneity further complicate automated analysis. Therefore, there is an urgent need for an analysis pipeline that is both efficient and interpretable, with user control, to enable standardized, cross-platform, multi-tissue-type analysis. MARQO addresses these challenges by integrating elastic registration, iterative segmentation, and user-guided classification, solving the problems of inaccurate segmentation and unreliable classification in traditional methods for complex tissue samples, providing a generalizable and verifiable analytical framework for cancer spatial biology research.
Research Methods and Experiments
The research team developed the MARQO analysis pipeline, supporting multiple staining techniques, including MICSSS, single-plex IHC, and 20-color mIF. The pipeline employs a tiled parallel computing architecture, enabling execution on local machines or computing clusters, significantly improving processing efficiency. Core modules include: elastic image registration to correct displacements from multi-round staining; iterative nucleus segmentation based on StarDist26, leveraging multi-round nuclear staining to enhance segmentation accuracy; mini-batch k-means for unsupervised cell clustering; and graphical interface-guided user quality control for annotating positive and negative signals in clustering results. The pipeline outputs intermediate files compatible with third-party software such as QuPath, facilitating downstream analysis.Key Conclusions and Perspectives
Research Significance and Prospects
MARQO provides an open-source, scalable, and user-controllable analytical platform for cancer spatial biology research, bridging the gap between automation and interpretability in high-throughput multiplex image analysis. Its modular design allows flexible adaptation to different staining protocols and tissue types, supporting broad applications from basic research to clinical translation. Future enhancements could integrate AI models to improve classification efficiency while retaining user supervision to ensure result credibility.
The study demonstrates MARQO's powerful capability in dissecting immune therapy response mechanisms, revealing spatial dynamics of CD8+ T cells, and offering new perspectives for developing predictive biomarkers. As spatial omics technologies become more widespread, MARQO is poised to become a key component of standardized analysis pipelines, enabling data integration and comparison across multi-center studies.
Conclusion
This study introduces and validates MARQO—an open-source, user-guided automated analysis pipeline for full-slide, single-cell-resolution analysis of multiplex immunohistochemistry and immunofluorescence images. By integrating elastic image registration, iterative nucleus segmentation, unsupervised clustering, and graphical interface-guided cell classification, MARQO significantly enhances the ability to resolve cellular phenotypes and spatial organization in complex tissue samples. In a cohort of hepatocellular carcinoma patients receiving neoadjuvant immunotherapy, MARQO successfully revealed enrichment and spatial clustering patterns of CD8+ T cells in fibrotic and necrotic regions among treatment responders, providing new evidence for understanding immune therapy response mechanisms. The platform supports multiple staining techniques and tissue types, with output highly consistent with pathologist evaluations, and is compatible with mainstream analysis software, demonstrating strong generalizability. MARQO’s open-source nature and user-controllable design make it a bridge connecting high-throughput spatial omics data with clinical pathological interpretation, potentially advancing standardization and reproducibility in cancer immune microenvironment research, and providing a powerful tool for biomarker discovery and personalized therapy development.
