Datasets:

---

MahmoodLab

/

hest

like 91

Follow

AI for Pathology Image Analysis Lab @ HMS / BWH 823

Hi, thanks for the great HEST-1k dataset! We're using the lung Visium samples and noticed something we'd love to confirm with you.
It seems 8 samples may be duplicates of the same 4 organoid sections, listed under two different labels:

• MEND41 = NCBI535
• MEND45 = NCBI536
• MEND47 = NCBI534
• MEND51 = NCBI537
  Our observations:

1. Metadata: NCBI534-537 don't have a Sample ID or source link, and their subseries is "PFA fixed lung_organoid_1-4" with a detailed organoid differentiation protocol. MEND41-51 are labeled "Healthy" but appear to trace to the same Mendeley dataset.
2. Files: For each pair, st/.h5ad and wsis/.tif are byte-for-byte identical (same md5).
3. H&E: The thumbnails are identical, and the tissue morphology looks like cultured cell clusters rather than alveolar lung tissue.
   If this is correct, these 8 samples would be 4 duplicated organoid sections, and the "Healthy" label on the MEND ones may be misleading for users working on normal lung tissue. It would be very helpful if you could take a look — happy to share the file hashes if useful.

(Top: Normal lung samples MEND85, MEND86, MEND89, MEND90 (labeled "Healthy" in metadata). Bottom: Suspected organoid samples MEND41, MEND45, MEND47, MEND51 (also labeled "Healthy" in metadata but showing cultured cell-cluster morphology). H&E thumbnails from HEST-1k.
Thanks for maintaining such a valuable resource!