H&E and CyCIF Slides of Tuberculosis, Lung Adenocarcinoma and Sarcoidosis for Foundation-Model Sparse Autoencoders (Zhao et al. 2026)
- Provider
- Zhao, Maliga et al., Sorger Lab, Harvard Laboratory of Systems Pharmacology
- License
- CC-BY-4.0 (source)
- Format
- TIFF (in sampled files)
- Stored at
- Amazon S3, s3://lsp-public-data/zhao-2026-fmsae (us-east-1)
- Access
- Free and public. No account with the data provider is needed.
- DOI
- 10.5281/zenodo.21299341
- Category
- Health & Biomedical
About this dataset
Whole-slide images for Zhao, Maliga et al. (2026), Compositional and interpretable representation of histology using AI foundation models and sparse autoencoders, from the Sorger lab at Harvard Medical School. The paper trains sparse autoencoders on embeddings from pathology foundation models to obtain interpretable histology features. Per the data index, the study uses five slide sets: DS1, 40 pulmonary TB H&E slides from Brigham and Women's Hospital (stored with the companion TB tertiary lymphoid structure study under ogbonna-2026-tb-tls/); DS2, 14 pulmonary TB slides from Brigham and Women's; DS3, 20 TB slides from the University of KwaZulu-Natal (AHRI); DS4, 12 lung adenocarcinoma slides from Brigham and Women's; and DS5, 5 sarcoidosis slides from Stockholm University and Research Center Borstel. Measured 2026-11-17: this prefix holds DS2/, DS4/, DS5/ and a demo_dataset.zarr. Files are stitched CyCIF pyramids and H&E images in OME-TIFF, one folder per patient sample; open with tifffile, napari or QuPath. Licensed CC BY 4.0; cite Zhao et al. (2026).
How to access this dataset in Python
import boto3
from botocore import UNSIGNED
from botocore.client import Config
s3 = boto3.client('s3', region_name='us-east-1', config=Config(signature_version=UNSIGNED))
s3.list_objects_v2(Bucket='lsp-public-data', Prefix='zhao-2026-fmsae/', Delimiter='/')Lists the files with anonymous access. Runs on any machine, with no cloud account.
Data structure
File types in sampled files: TIFF
Files
Example file: zhao-2026-fmsae/DS2/LSP11103.ome.tif
Analyze this dataset without downloading it
On Scigantic, H&E and CyCIF Slides of Tuberculosis, Lung Adenocarcinoma and Sarcoidosis for Foundation-Model Sparse Autoencoders (Zhao et al. 2026) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Health & Biomedical datasets
- IDC TCGA-LGG: Low Grade GLioma whole-slide microscopy (516 patients)
- IDC Breast-MRI-NACT-Pilot: Breast Cancer MRI (64 patients)
- Nutrition5k: RGB-D Food Imagery with Per-Ingredient Mass, Calories and Macronutrients (Google Research)
- TIGER Training
- WSSS4LUAD: lung adenocarcinoma H&E patches for weakly supervised tissue segmentation (tumor, stroma, normal)
- IDC QIN PET Phantom: Phantom PET (2 patients)
- COph100: infant retinal fundus images for retinopathy of prematurity (ROP) registration, with vessel masks and control points
- IDC VAREPOP-APOLLO: Esophageal Carcinoma, Head and Neck Squamous Cell Carcinoma, Lung Adenocarcinoma and 5 more CT, MRI, PET, radiographs (41 patients)