Presentation @ ICCR

[SimonBiggs]

Abstracts are due before the end of January. Does anyone from the PyMedPhys community want to undergo a presentation on behalf of PyMedPhys?

https://www.iccr2024.org/abstract.html

I'd personally be interested in a PyMedPhys streamlit app that utilises the Claude API in combination with Mosaiq... that'd be pretty awesome...

[SimonBiggs]

@sjswerdloff, @pchlap, @nlambriICH or @Matthew-Jennings do you have any thoughts here?

[SimonBiggs]

...if anyone wants Claude API access to be able to utilise it within PyMedPhys I should be able to make that happen.

[Matthew-Jennings]

I'm keen in the general sense! No access to Mosaiq though...

[SimonBiggs]

Hmm, maybe something else instead? What's something you might like to do?

[pchlap]

I'd be keen to be involved in some way, I do hope to attend ICCR next year.

However, I'll be working on an abstract related to my PhD work as well. So not sure how much time I will have to commit to this one (I'm also moving to the US on Jan 3).

One idea would be to report on the overall development of pymedphys, reflecting on the progress since the last ICCR which I think was when it started to gain momentum (Simon definitely sold it to me there :D). It could even be a survey of the current open-source landscape in Medical Physics.

Perhaps not as exciting as your idea Simon, but I think this would be manageable without too big of a time commitment from any of us by the end of Jan. And I think it would certainly be of interest to the ICCR attendees.

[SimonBiggs]

@pchlap, I think that's a great idea 🙂

Would still be great to have a "demoable feature" though... Does someone have a favourite they'd like to demo?

[nlambriICH]

Unfortunately, I won't be able to attend ICCR. I'd like to contribute what I can.

[SimonBiggs]

Here is an interesting tool that integrates with MSSQL (and therefore the PyMedPhys mosaiq implementation) that will help support anonymisation of certain information:

https://pypi.org/project/pynonymizer/

This might be quite amazing to include within PyMedPhys... being able to undergo a process of anonymising Mosaiq data might be an amazing feature within PyMedPhys?

[SimonBiggs]

I could imagine a default could be to delete data, and then step through an allow list as strategy files are created:

https://github.com/rwnx/pynonymizer/blob/master/doc/strategyfiles.md

Essentially, if a strategy file for that datatable column entry doesn't exist then delete the data. And only export data that is both in the allow list and has an appropriate strategy file...

[SimonBiggs]

Anyway, let's say we're running with the simple version that @pchlap mentioned, who would be interested in presenting + putting in an abstract?

(they're due in Jan)

[pchlap]

Hi @SimonBiggs, I am definitely interested, but as mentioned above can't fully commit to driving this one forward as I'm moving to the US in a couple of weeks. I'd love to be involved though if we can get a few more hands on deck :)

I do hope to attend ICCR so I'd be happy to present the work if no one else wants/is able to.

[Matthew-Jennings]

Why don't we arrange a meeting to brainstorm ideas and form a plan? Perhaps first week of Jan (mindful of tight deadline)?

[SimonBiggs]

Sure, are you happy to spearhead making a meeting like that happen @Matthew-Jennings?

[sjswerdloff]

I could imagine a default could be to delete data, and then step through an allow list as strategy files are created:

https://github.com/rwnx/pynonymizer/blob/master/doc/strategyfiles.md

Essentially, if a strategy file for that datatable column entry doesn't exist then delete the data. And only export data that is both in the allow list and has an appropriate strategy file...

I'm game for starting on a strategy file that specifies "update_columns" for PHI (using Faker or whatever other approach is specified).
My constraints are that I will only list columns that contain PHI (no allow list with "everything else" that would essentially enumerate the whole schema), and for testing purposes we would have to mock a database that has those columns but otherwise doesn't use schema copied from MOSAIQ (to minimise exposure of the actual schema to those who haven't agreed to the license conditions for the schema). I can get started on this next week (week of January 8th, 2024).

Where would you want the strategy file to go within PyMedPhys...
lib/pymedphys/_mosaiq ?

[SimonBiggs]

That sounds amazing Stuart! :)

And yes, location seems right.

[sjswerdloff]

That sounds amazing Stuart! :)

And yes, location seems right.

I see that there is a place in tests/mosaiq/data for CSV files whose basename is the Table and whose contents includes the column names in the first row; and a toml file for mapping the column data types. So the infrastructure is present for providing additional mock data for testing purposes. I would prefer to only add just enough schema in new CSV files for testing purposes (and make sure it correlates properly with the existing test data).

[SimonBiggs]

Yup, that sounds perfect