Abstract
The imminent release of tissue atlases combining multi-channel microscopy with single cell sequencing and other omics data from normal and diseased specimens creates an urgent need for data and metadata standards that guide data deposition, curation and release. We describe a Minimum Information about highly multiplexed Tissue Imaging (MITI) standard that applies best practices developed for genomics and other microscopy data to highly multiplexed tissue images and traditional histology.
Introduction
Highly multiplexed tissue imaging is revolutionizing our understanding of the tumor microenvironment and infectious diseases by capturing 20 to 60+ molecular markers simultaneously within a single tissue section. However, these methods generate massive, heterogeneous datasets—often exceeding 1TB per sample—that are notoriously difficult to share, reproduce, or integrate.
Unlike the field of genomics, which has long-established standards like FASTQ and BAM, multiplexed imaging has historically lacked a unified "language" for metadata. Without standardized information regarding biospecimen handling, antibody reagents, and image processing steps, it is nearly impossible for outside researchers to validate findings or integrate data from disparate labs into larger "tissue atlases." The MITI (Minimum Information about highly multiplexed Tissue Imaging) standard addresses this gap by establishing a rigorous framework for metadata and data-level definitions.
The MITI Framework: A Five-Level Data Hierarchy
The core of the MITI proposal is a five-level data hierarchy that mirrors the processing stages used in genomics. This structure provides a mental model that bioinformaticians can immediately grasp, treating imaging data with the same rigor as sequencing data.
- Level 1 (Raw): The original numerical output from the instrument. These are often stored in proprietary vendor formats and serve as the "FASTQ-like" foundation of the dataset.
- Level 2 (Assembled): Full-resolution mosaics that have undergone automated stitching, registration, and illumination correction. These are stored in the universal OME-TIFF format and are analogous to BAM files in genomics.
- Level 3 (Processed): Images that incorporate "interpretive intent." This includes quality-controlled mosaics (where failed antibody channels have been removed) and segmentation masks that identify the boundaries of individual cells.
- Level 4 (Quantified): "Spatial feature tables" containing single-cell measurements such as signal intensity, spatial coordinates, and morphology. These are analogous to count matrices in single-cell RNA-seq.
- Level 5 (Interpreted): High-level results derived from the data, such as cell-type annotations, dimensionality reduction plots (UMAP/t-SNE), and interactive "data stories" (e.g., via the MINERVA platform).
Metadata and FAIR Principles
To ensure that data is Findable, Accessible, Interoperable, and Reusable (FAIR), MITI specifies mandatory and recommended metadata fields across clinical context, biospecimen handling, and experimental reagents. These standards are implemented in machine-readable YAML and JSON formats.
One critical distinction in the MITI standard is the transition from Level 2 to Level 3. While Level 2 is purely automated, Level 3 involves human intervention and quality control decisions. By documenting these interventions, MITI ensures that downstream biological interpretations are built on a foundation of trust and transparency.
Implementation and Adoption
While MITI is a guidelines-focused framework, its utility has been demonstrated through detailed implementations in major research initiatives:
- CRC1: A colorectal cancer specimen from the Human Tumor Atlas Network (HTAN).
- COVID-19: A patient lung specimen used to study SARS-CoV-2 pathology.
The standard is already being used to structure data within major international projects, including HTAN and the Human BioMolecular Atlas Program (HuBMAP). To facilitate adoption, the standard is supported by the SCHEMATIC tool, which allows researchers to use Google Sheets with controlled vocabularies to generate valid metadata. Furthermore, MITI is compatible with major open-source tools such as QuPath, napari, MCMICRO, and OMERO.
Discussion and Future Directions
MITI addresses the "tragedy of the commons" in data hosting by suggesting practical models for large-scale data sharing, such as "requester pays" cloud egress, ensuring long-term availability without overtaxing hosting institutions.
As the field evolves, the MITI standard will need to adapt. While currently optimized for 2D sections, future iterations are being developed to better support 3D datasets and subcellular resolution imaging (e.g., MERFISH). Additionally, as the community moves toward next-generation cloud-optimized formats like OME-NGFF (Zarr), MITI will evolve to remain format-agnostic. By remaining independent of any specific commercial entity and governed by an open GitHub repository, MITI provides a living framework for the future of spatial biology.
Summary updated