interesting project!

[kirbyju]

Hi, I stumbled on to this sort of randomly but it reminded me of a project we did for The Cancer Imaging Archive at RSNA many years back. At the meeting we crowd-sourced 2d annotations from meeting attendees on TCIA datasets. The resulting data were published at https://www.cancerimagingarchive.net/analysis-result/crowds-cure-2018/ and https://www.cancerimagingarchive.net/analysis-result/crowds-cure-2017/. Perhaps you might find it interesting to re-run your pipeline with these data as input?

Best,
Justin

p.s. https://discourse.canceridc.dev/t/text2cohort-a-new-llm-toolkit-to-query-idc-database-using-natural-language-queries/433/4 is what led me to your Github page. If you had any interest in working with me on building an LLM interface to query TCIA data I think that would be really fun and useful! I've thought about it many times, but don't have enough expertise in coding and LLMs to tackle it on my own. I do have a deep knowledge of our APIs and available data though.

[vishwaparekh]

Hi Justin,

Thank you for your comment and feedback. It would definitely be very interesting to try and build an LLM pipeline to query TCIA. Are you available to hop on a quick call early next week to discuss and see how we can best collaborate to build this out? My email is Vishwa.s.parekh@uth.tmc.edu