Expand CDA to NCIT Mappings for 12 tissues & Enhance OpUberonMapper

[rajdeepmondaldotcom]

This introduces a significant expansion to our CDA to NCIT mappings, featuring a comprehensive addition of twelve CSV files for 12 tissues for CDA to NCIT mappings. These files cover a diverse range of tissues: bone, brain, breast, cervix, colon, heart, kidney, liver, lung, pancreas, skin, and thyroid. Please find these detailed mappings organized in src/oncoexporter/ncit_mapping_files/cda_to_ncit_tissue_wise_mappings.

Furthermore, OpUberonMapper was updated within src/oncoexporter/cda/mapper/op_uberon_mapper.py, adding new terms and mappings that translate the string representations of anatomical locations into their corrosponding UBERON terms.

Kindly please take a look,

Thanks a lot,
Rajdeep

[justaddcoffee]

great @rajdeepmondal-el ! Maybe we can review tomorrow?

[justaddcoffee]

cc: @pnrobinson

[justaddcoffee]

@rajdeepmondal-el these mappings seem very good

Manually checking these, I did spot 1-2 that probably need updating, e.g. this one should probably be NCIT:C3749 Alveolar Rhabdomyosarcoma and not NCIT:C194245 Malignant Neoplasm of Connective and Soft Tissue of Head, Face and Neck

Maybe @pnrobinson, you and I could manually review some more of these today on the call?

[rajdeepmondaldotcom]

Thanks a lot for pointing it out, Agree with you @justaddcoffee I can explain why the result is that so, and some probable ways to make the predictions even more accurate. For this specific example, the primary_diagnosis_site consists of quite a lot of spurious information that might make the model a bit confused, it is trying to be more context-aware than necessary, which I can solve by assigning weights to the primary diagnosis part.

Also, there are some edge cases which i will also share.

[justaddcoffee]

handing this off to @ielis

[ielis]

Hi @rajdeepmondal-el

thanks a lot for your excellent work!

I plugged the mapping files into the library and tested to run one of the driver script and it looks OK.

It's looking good, please feel free to merge the PR!

[rajdeepmondaldotcom]

Thank you very much @ielis, I have merged it.