MIFSpace: A Cross-Panel Framework for Expanding Marker Space in Multiplex Immunofluorescence

Background: Clinical-grade multiplex immunofluorescence (mIF) imaging enables spatially resolved profiling of tissue microenvironments but is constrained by limited panel sizes. To interrogate complex cellular states, clinical studies increasingly rely on sequential staining across adjacent tissue sections; however, existing computational tools are largely designed for single-panel analysis and support for cross-panel integration remains limited.

Methods: We developed MIFSpace, an open-source scalable computational framework designed with a cross-panel-first architecture for antibody-based mIF images from human FFPE tissues. It integrates (1) deep learning-driven nuclear segmentation using custom StarDist models, (2) per-cell marker quantification, (3) supervised cell classification, and (4) spatial analysis and visualization. These modules are unified through (5) an interactive co-registration interface for aligning adjacent tissue sections, and (6) concordance quantification assessing marker consistency across panels, enabling robust analysis of shared and panel-specific features.

Results: StarDist models achieved accurate and robust nuclear segmentation across multiplex datasets. Cross-panel co-registration enabled spatial alignment of adjacent sections. Concordance analysis revealed marker-dependent variability across panels, providing a quantitative framework for evaluating cross-section consistency. Integrated cross-panel analysis enabled identification of cellular phenotypes and multicellular spatial organization not resolvable within individual panels with high translational impact.

Conclusion: MIFSpace overcomes inherent marker limitations of mIF platforms by enabling integration, validation, and joint interpretation of sequential staining panels. By effectively expanding the measurable marker space, this approach enables reconstruction of higher-dimensional tissue states and advances the analytical capabilities of spatial profiling in clinical oncology.

Learning Objectives:

1. Describe the modular components of MIFSpace for mIF spatial analysis.
2. Explain how deep learning-based segmentation improves nucleus detection in multiplex tissue images.
3. Discuss how across-panel integration expands marker space and reconstructs tissue biology.