Frequently Asked Questions

What is this project about?

This project aims to improve AI-generated tissue masks in histology images containing adipose tissue. By reviewing and correcting these masks, volunteers help create better research data for digital pathology and biomarker discovery.

What am I looking at in each task?

Each task shows a two-panel image:

• the original histology tile
• the predicted tissue mask

You will compare these panels and decide whether the mask is clearly correct, missing a wall, or too hard to judge confidently.

Do I need medical or pathology experience?

No. You do not need previous medical or pathology training to take part. A short tutorial and examples will explain what to look for and how to annotate missing-wall cases.

What does “Clearly correct” mean?

Choose Clearly correct when the predicted mask already follows the visible tissue boundaries well enough that no meaningful correction is needed.

The mask does not need to be perfect. Small imperfections are acceptable.

What does “Missing-wall” mean?

Choose Missing-wall when a real tissue wall or septum is missing, broken, or misplaced in the predicted mask.

In these cases, you will:

• draw where the wall should be
• mark the incorrect predicted mask that should be removed

What does “Hard / ambiguous” mean?

Choose Hard / ambiguous when the tissue boundary is too unclear to judge confidently.

Examples include:

• very faint walls
• torn or folded tissue
• blur or staining problems
• cases where more than one interpretation is possible

If you are unsure, choose Hard / ambiguous instead of guessing.

What counts as tissue in this project?

Thin adipocyte walls and connective septa count as tissue.

Empty adipocyte lumina do not count as tissue.

If a wall is slightly broken because of tissue preparation, it may still be corrected when its intended path is visually clear.

Am I diagnosing cancer?

No. Volunteers are not being asked to diagnose cancer or make clinical decisions.

Your role is to help improve tissue annotations in histology images. These annotations may later support research into biomarkers related to diseases such as cancer.

Why are you studying adipose tissue?

Adipose tissue is easy to see in many biopsy images, but it is often overlooked. Its structure may contain useful biological information related to inflammation, tissue remodeling, and diseases such as cancer.

We are studying whether improved annotations of adipose tissue can help support the discovery of new biomarkers.

Why does AI need help?

AI can analyze large numbers of images quickly, but it still makes mistakes. In adipose tissue, thin walls may appear broken or incomplete during tissue preparation, which can make automatic segmentation difficult.

Human review helps correct these errors and improve the quality of the data.

What should I do if I am not sure?

If you cannot judge the boundary confidently, choose Hard / ambiguous.

It is better to mark a case as uncertain than to guess.

Will my annotations really be useful?

Yes. Volunteer contributions help us improve tissue masks, create better datasets, and refine AI models. These improved annotations may also support research into new tissue-based biomarkers.

Why do some masks look strange?

Some masks contain errors because the AI model can confuse broken or faint tissue walls with open spaces. This is one of the main challenges the project is trying to address.

Do I need to correct the whole image?

No. Only mark the area that needs correction.

If the case is labeled Missing-wall, draw:

• where the wall should be added
• where incorrect predicted mask should be deleted

Do not redraw the full mask.

What happens to difficult cases?

Cases marked as Hard / ambiguous can be separated from clearer examples and reviewed more carefully by the research team or by expert annotators.

How long does one task take?

Most tasks should take only a short time. Some images are easy to review, while others may require more careful attention.

Can students or non-experts participate?

Yes. This project is designed so that non-experts can contribute after a short introduction and tutorial.