Published June 26, 2026 | Version v3

Decoding Fibrosis: Histology Test Data, Collagen Segmentation Annotations, and CDP Clustering Tile Pool

Authors/Creators

  • 1. ROR icon University of Oxford

Description

This dataset contains the artefacts accompanying our paper:

Decoding Fibrosis: Transcriptomic and Clinical Insights via AI-Derived Collagen Deposition Phenotypes in MASLD

The associated analysis code, trained models, and retraining workflows are available at: https://github.com/mkatw/decoding-fibrosis

Contents:

This record includes supporting data for testing, reproducing, and extending the Decoding Fibrosis histology analysis workflows. 

  1. Example histology whole slide image with expected results.
    An example PSR-stained liver WSI with expected pipeline outputs, intended for testing the public analysis code.
  2. Collagen segmentation training tiles and annotations.
    Paired PSR liver tiles and collagen annotations for training or fine-tuning of collagen segmentation models.
    • The "collagen segmentation tiles cocomasld" folder contains tiles from the discovery cohort introduced in the paper.
    • The "collagen segmentation tiles cocomasld extended retraining" contains additional tiles from the validation cohort described in the supplementary material.
  3. CoCoMASLD CDP clustering tile pool.
    Sampled collagen probability-map tiles used for collagen deposition phenotype (CDP) clustering model development. Tile order has been randomised and the filenames retain no connection to the original slides. These tiles support comparison and extension of CDP-style clustering methods.

This dataset is inteded as a modular research source. Users may reproduce the workflows, retrain individual components, compare alternative models, or adapt selected parts of the pipeline for related cohorts and modelling questions.

Preparation of this dataset was supported by Novo Nordisk A/S as part of the CoCoMASLD study.

Files

cdp_clustering_tiles_cocomasld_v1.zip

Files (3.4 GB)

Name Size
md5:db5f6a78c9bf7ecbe258643e4ab070cd
2.0 GB Preview Download
md5:228326b13c5a80f2f81c18ec9ef75790
471.0 MB Preview Download
md5:b94aba2c742cb722f71399e76aee7587
310.7 MB Preview Download
md5:6f87e07cf454449bb3f1cba567e59ea6
575.8 MB Preview Download

Additional details

Related works

Is supplement to
Publication: 10.1097/HEP.0000000000001811 (DOI)

Software

Repository URL
https://github.com/mkatw/decoding-fibrosis
Programming language
Python