bug: Optimal series selection is handled in a suboptimal way #43
Description
For images with multiple series, for instance containing the section and some QR code or similar, we have two ways of getting the "right" image. Either:
- Select the first series (or another manually selected series)
- We convert all series and then only keep the largest one (for instance based on resulting file size)
The latter seems to work really well for all of our cohorts, but require that all series are converted which may result in redundant compute. What would be better would be to know a priori which of the series had the largest image and then only convert that one. That should be possible if we found where in the metadata this information was stored.
We have gotten some intel on how QuPath does this series selection here, which we could take inspiration from:
https://forum.image.sc/t/selecting-optimal-series-from-olympus-vsi-in-qupath/110152