Blurry Region Detection Across Varied Slide Samples

[EmanuelSoda]

Hi HistoQC Team,

Firstly, I'd like to express my gratitude for developing such a powerful tool; it's been incredibly useful in my work. I'm reaching out to discuss an issue I've encountered while analyzing a diverse set of slide samples using HistoQC using the pipeline v2.1.

My dataset includes both biopsies (which are quite small) and resections (significantly larger), and I'm curious about the effectiveness of the blurry_removed_percent metric in this context. My questions are twofold:

Given the considerable size difference between biopsies and resections in my dataset, can the blurry_removed_percent column accurately help in identifying slides with blurred regions for both types of samples?

It appears that even though HistoQC performs tissue segmentation, this segmentation is not taken into account when analyzing blurriness; hence, the blurry_removed_percent metric applies to the whole slide rather than just the tissue regions. Is my understanding correct? If so, would it not be more beneficial to focus the blurriness analysis solely on tissue regions? After all, a blurry background doesn't typically impact the following analysis negatively.

I believe that adjusting the analysis to consider only tissue regions could enhance the utility of the blurry_removed_percent metric, especially in datasets like mine that include a wide range of sample sizes.

Thank you for your time and assistance. I look forward to your insights and any suggestions you might have.

Best regards,
Emanuel

[jacksonjacobs1]

Glad to hear you're finding HistoQC useful in your work!

If you're using the v2.1 configuration file, blurry_removed_percent should be calculated relative to the previous mask_use and not the full image.

The previous mask_use in this case already has tissue segmented, so blurry_removed_percent should be relative to the tissue region.

Can you share screenshots of these masks for the same image:

1. The _bright.png mask
2. The _blurry.png mask
3. The _mask_use.png output mask

[EmanuelSoda]

Thanks a lot for the explanation. For the masks, it would be better if I could send them to your email for privacy issues (they come from patient biopsies).
Could you send me your email?
Kind,
Emanuel

[jacksonjacobs1]

Sure.
[email protected]