OPTIMAL: An OPTimized Imaging Mass cytometry AnaLysis framework for benchmarking segmentation and data exploration.

Area of Science:

• Biomedical Imaging
• Computational Biology
• Immunology

Background:

• Multiplexed tissue imaging, including imaging mass cytometry (IMC), faces challenges in single-cell segmentation and data analysis.
• Suboptimal analysis can lead to inaccurate cell phenotype, state, and spatial relationship identification compared to single-cell suspension data.

Purpose of the Study:

• To develop and validate the "OPTimized Imaging Mass cytometry AnaLysis (OPTIMAL)" framework.
• To benchmark various approaches for IMC data analysis, including segmentation, parameter transformation, batch correction, visualization, clustering, and spatial neighborhood analysis.

Main Methods:

• Tested multiple cell segmentation models, arcsinh cofactor values, dimensionality reduction algorithms (e.g., PacMap), and clustering methods (e.g., FLOWSOM).
• Utilized FFPE human tonsil tissue microarrays stained with 27 metal-tagged antibodies across 12 batches.
• Assessed neighborhood analysis techniques, including "disc" pixel expansion versus "bounding box" methods.

Main Results:

• Ilastik-derived probability maps improved single-cell segmentation, though issues were more apparent post-clustering.
• An arcsinh cofactor of 1 and Z-score normalization effectively transformed parameters and corrected batch effects.
• PacMap excelled in dimensionality reduction, and FLOWSOM outperformed Phenograph for cell type identification.
• "Disc" pixel expansion was superior for neighborhood analysis, with object filtering crucial.

Conclusions:

• The OPTIMAL framework provides a standardized approach to evaluate and integrate IMC data analysis pipelines.
• It facilitates accurate cell identification, state determination, and spatial analysis in multiplexed tissue imaging.
• Generated .FCS files enable single-cell exploration using familiar flow cytometry software and algorithms.