Datasets:

ArXiv:
License:

Mismatch case numbers in hugging face repo and metadata.json

#5
by lely475 - opened

Dear Histai Team,

Thanks a lot for providing this large and insightful dataset.
When looking into the metadata.json for the colorectal cancer sets I stumbled upon an inconsistency in case numbers based on the metadata.json compared to the case numbers provided in your publication and huggingface repo:

Dataset Total Cases metadata.json Total Cases Paper / Huggingface Repo
histai/HISTAI-colorectal-b1 878 998
histai/HISTAI-colorectal-b2 57 62

Heres the code snipped I used for getting the case numbers for HISTAI-colorectal-b1 and HISTAI-colorectal-b2 form the metadata.json:

metadata = pd.DataFrame(json.load(open("path-to-json/metadata.json")))
crc_b1_df = metadata[metadata["case_mapping"].str.contains("HISTAI-colorectal-b1")]
crc_b2_df = metadata[metadata["case_mapping"].str.contains("HISTAI-colorectal-b2")]
print("HISTAI-colorectal-b1", len(crc_b1_df), "HISTAI-colorectal-b2", len(crc_b2_df))

Am I missing something? (e.g. can some cases have a different case_mapping, not including the set identifier)
Or are there cases with WSIs but no entry in the metadata.json?

Thank you already!
Best,
Lydia

HistAI org

Some cases are missing metadata, we are looking for it.

alpchel changed discussion status to closed

Sign up or log in to comment