QA for WSIs

[JHancox]

As mentioned at WG meeting, the need for QA for Whole Slide Imaging is something that seems quite obvious but has not featured in any discussions or feature requests yet (perhaps because it is not that 'exciting').
In fact, many slides in the public domain that are commonly used for model training contain artefacts that have the potential to render nonsense AI results.

The question is therefore:
Do we need one or more QA algorithms?

If so, what artefacts might it detect?

• Tears
• Folds
• Pen marks
• Out-of-focus regions
• stain anomalies

At what granularity? (e.g. per slide, per tile)

Please make any suggestions regarding the scope and priority of this QA theme.