Question: Methodology for Inflammatory Cell Dot Annotations in PAS/IHC WSIsLabel question

I'm seeking clarification about the exact methodology used to create the inflammatory cell dot annotations in the dataset.

Current observation: When overlaying the dot annotations onto the IHC WSIs, I notice variable quality (presumably due to registration issues), yet they align perfectly with the PAS WSIs. Specific questions:

What was the precise workflow for creating these dot annotations?

Were annotations created directly on PAS slides with IHC as visual reference only? Were annotations created on IHC slides and computationally mapped to PAS coordinates? Was some combination of both approaches used?

Understanding the ground truth generation process will help ensure appropriate algorithm development and validation approaches. Thank you for your assistance.

Thank you for this question.

You noticed correctly that the allignment of PAS WSI and IHC WSI are not always perfect due to suboptimal registration. So you can indeed encounter a minor shift of dot annotations for IHC WSI.

The dot annotations were initially made on IHC WSI and afterwards transfered/mapped onto PAS WSI. We made sure to align the dot annotations to correspond with the morphology in PAS WSI as this will be the most important ground truth from our perspective (Models will be trained using PAS WSI and corresponding dot annotations).

I hope this answers your question. Let us know if you need more clarification.