Request access to Eva-data

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

This dataset is released under CC BY-NC-ND 4.0 and may only be used for non-commercial research with proper attribution. Any commercial use, sale or other monetisation of the data or of models trained on it is prohibited and requires prior approval. The primary email of your Hugging Face account must match your institutional email: requests from free providers (@gmail, @outlook, @hotmail, @qq, @163, @foxmail and the like) are denied. By downloading the data you attest that all information given below is correct and up-to-date, and you agree not to distribute, publish, redistribute or reproduce a copy of this dataset in whole or in part. If another user in your organisation wishes to use it, they must register as an individual user and agree to these terms. You may not attempt to re-identify the patients from whom these data were derived. Commercial entities should contact the corresponding author.

Log in or Sign Up to review the conditions and access this dataset content.

Eva-data

Multiplexed imaging patches for the Eva study. One .npz per region, named by its acquisition id, under a per-cohort directory. K varies per region and the channel order follows that file's own biomarkers, so index by marker name, never by position.

key
mif (n, 224, 224, K) float16 multiplexed patches
he (n, 224, 224, 3) uint8 paired H&E, same patches; absent where not acquired
coords (n, 2) int64 top-left pixel of each patch within the region
biomarkers (K,) str that region's panel, in channel order

Patches tile the region on a regular 224-pixel grid. Where a cohort has paired H&E, only the H&E-paired regions are included.

cohort regions patches K H&E size
UPMC-HNC 327 33,123 41 yes 42.6 GB
Stanford-CRC 406 13,461 40 no 14.6 GB
Stanford-HCC 35 9,017 41 yes 9.3 GB
DFCI-HNC 58 2,266 40 no 2.9 GB
MDACC-HCC 30 16,116 48 yes 17.3 GB
import numpy as np

d = np.load("UPMC-HNC/UPMC_c001_v001_r001_reg001.npz")
mif, coords, markers = d["mif"], d["coords"], d["biomarkers"]

h, w = coords.max(0) + 224
region = np.zeros((h, w, mif.shape[-1]), dtype=np.float16)
for patch, (y, x) in zip(mif, coords):
    region[y:y + 224, x:x + 224] = patch

cd8 = region[..., list(markers).index("CD8")]
from huggingface_hub import snapshot_download
snapshot_download("yandrewl/Eva-data", repo_type="dataset",
                  allow_patterns="UPMC-HNC/*", local_dir="data")
Downloads last month
29