Error running benchmark code:

Hi there,

I am using the dataset v.1.2.1 from physio.net and nov2025 ver of SNOMED CT International version as required, however, when I run the benchmark code, I got the KeyError: 310054005 when I run the following cell - which suggest the SNOMED CT concept id in the annotation is NOT in the nov2025 ver of SNOMED CT.

concepts_in_scope = [
SG.get_concept_details(a) for a in annotations_df.concept_id.unique()
]

print(f"{len(concepts_in_scope)} concepts have been selected.")

The error is reproducible if you run the bench mark code with the datasets in the current challenge.

Please help to ensure the correctness of the annotation file and/or SNOMED CT version. Thank you!

Cheers,

TT

FYI I checked that there are a total of 101 concept_ids that is in annotation training dataframe but NOT in the SNOMED CT nov 2025 version.

I checked, and the code you mentioned was present in the 2023 Edition used for the Challenge, but no more in the current version; so likely the training set maintains the same annotations. However, since the training set does not cover all possible codes anyway, likely it is a safe approach to just remove the obsolete codes, being not so many.

Hi @tt2026 - @vdellamea is correct, some codes in the dataset may not be current with the version of SNOMED used for the challenge, and you should handle this however you choose.