Show/calculate 75th percentile for DRLs

[Tim de Wit]

Currently histogram-data have to be exported to excel in order to calculate e.g. the 75th percentile, required for determining the diagnostic reference levels (DRLs). It would be great if these could be calculated automatically and/or be visualized in the histogram plots.

[Ed McDonagh]

Hi @tcdewit. Personally, I wouldn't like that to be a default feature, though if someone wants to add it in and make it an option I'd welcome the pull request.

The reason for this is that calculating the 75th percentile is commonly done due to a misunderstanding of the technique. I am speaking from a position of the UK protocols, but as I understand it the technique is the same across Europe. I could be wrong though

The UK strategy for local and national DRLs is as follows:

• At a local level DRLs are calculated for common procedures, either for all imaging equipment across an institution, or for individual rooms where there are justifiable differences
• The DRL in this context is the mean or median dose indicator for 'average' sized patients
• The DRLs for each common procedure (or the room means) are then periodically submitted to a national dose survey
• At a national level, the distribution of the DRLs submitted is then considered, and the national DRL is set to be the 75th percentile of this distribution.

Therefore, the national DRL is the 75th percentile of the mean dose indicators from each of the contributing centres.

Back at a local level, we can then compare our DRL (our mean values) to the NDRL, knowing that most institutions should have a local DRL that is less than the National DRL, and if we don't, we may need to do something about it!

Is this your understanding of how DRLs should be calculated and used?

[David Platten]

@edmcdonagh, I agree with your post.

I'd like to add that it's likely that the ICRP will be recommending median rather than mean at some point in the future. They had a draft report open for consultation called "Diagnostic Reference Levels in Medical Imaging". It's closed for comment now, but you can still see the draft: http://www.icrp.org/page.asp?id=256. It advocates the use of median rather than mean when analysing local data.

[Tim de Wit]

You're completely right... my mistake! Issue can be closed.

[Ed McDonagh]

Marking it as wontfix rather than closed. Thanks @tcdewit

[David Platten]

@Ed McDonagh and @tcdewit,

ICRP report 135 is now published (https://www.icrp.org/publication.asp?id=icrp publication 135).

On page 22 this defines a Local DRL as:

“A DRL for an x-ray procedure set in healthcare facilities within part of a country for a defined clinical imaging task, based on the 75th percentile value of the distribution of the appropriate DRL quantity in a reasonable number (e.g. 10–20) of x-ray rooms. Local DRLs may be set for procedures for which no national DRL is available, or where there is a national value but local equipment or techniques have enabled a greater degree of optimisation to be achieved so that a value less than the corresponding national DRL can be implemented.”

This reads to me as though we should be calculating the 75th percentile of the DRL quantities (DAP, DLP etc.) from a group of rooms, and using that as our Local DRL.

Or is it suggesting that we take the 75th percentile of the room medians and use that as our Local DRL?

Thoughts?

[David Platten]

See issue #918 where I’ve created charts of user-defined percentile.