Multi-Modal Foundation · MahmoodLab
KRONOS
A self-supervised vision model for feature extraction from multiplexed tissue images in spatial proteomics.
What KRONOS is
KRONOS is a foundation vision model developed by MahmoodLab for spatial proteomics and multiplexed tissue imaging. It operates in the domains of biology, medicine, and computational pathology.
The model addresses the challenge of analysing highly multiplexed tissue imagery across varying experimental panels and imaging modalities. By providing standardised feature representations from complex multiplexed microscopy data, it supports researchers studying spatial protein expression and tissue architecture.
How it works
KRONOS is built on a Vision Transformer Small architecture with 16x16 patch tokens (ViT-S/16). It is trained using self-supervised learning techniques directly on highly multiplexed tissue images, enabling it to learn visual representations across diverse imaging modalities and marker panels without requiring manual annotations.
Key facts
- Architecture
- ViT-S/16
- Learning paradigm
- Self-supervised learning
- Primary domain
- Spatial proteomics and pathology
- Pipeline task
- Image feature extraction
- Associated paper
- arXiv:2506.03373
What you can use it for
- Multiplexed Tissue Image Feature Extraction
- Extract numerical feature embeddings from high-dimensional multiplexed tissue imagery for computational analysis.
- Cross-Modality Spatial Proteomics
- Process and represent spatial proteomics data across differing laboratory imaging modalities and antibody panels.
- Computational Pathology Representation Learning
- Generate self-supervised representations of tissue morphology and protein expression to support downstream biological discovery.
Run it here
We list this model but cannot run it yet: gated on the Hub — never run past a gate the user did not accept.
Everything on this page still applies. Running it needs a loader for this repository, not a bigger machine.
Strengths and limits
Good at
- Pretrained across diverse multiplexed tissue imaging modalities and marker panels.
- Employs self-supervised pretraining, avoiding reliance on manually annotated ground truth labels.
- Uses a standard ViT-S/16 architecture tailored for high-dimensional spatial proteomics data.
Watch out for
- Designed specifically for feature extraction rather than end-to-end task prediction.
- Performance and compatibility details for specific custom antibody panels are not specified.
Inputs and outputs
You provide
Image
PNG, JPEG or TIFF. Sent to the model as-is.
You get back
Mean-pooled representation, one vector per input.
Licence
Check the licence before commercial use · cc-by-nc-nd-4.0
This model is gated by its authors — its licence has to be accepted on the source repository before it will run.
Questions
Can I use KRONOS commercially?
Check the licence before commercial use. This model is gated by its authors — its licence has to be accepted on the source repository before it will run.
What architecture is KRONOS based on?
KRONOS is built upon a Vision Transformer Small backbone with 16-by-16 patch tokens (ViT-S/16).
What types of input data does the model accept?
The model is designed for highly multiplexed tissue imagery across different panels and spatial proteomics imaging modalities.
What is the primary function of KRONOS?
Its core function is image feature extraction, producing visual embeddings from multiplexed tissue specimens.
How to cite it
arXiv:2506.03373
Written from KRONOS's own documentation (model card). Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.