-
Notifications
You must be signed in to change notification settings - Fork 1.1k
April 26, 2022 (Dev Meeting)
David Manthey edited this page Apr 27, 2022
·
12 revisions
# | Owner | Title | Time (min) |
---|---|---|---|
1 | Sachi, David | MONAI Label update | 20 |
2 | All | Pathology Annotations and Inference Outputs | 20 |
3 | Behrooz, Jonny | WSI Datasets | 20 |
- MONAI Pathology: Jonny, David, Shan, Lee, Fernando, Kenza, Behrooz, Andres, Gigon, Ziyue, Jeff
- David has proposed some changes in MONAI Label to Sachi
- The current implementation support training and annotation
- No need to standardize the annotation since it is fairly easy to convert between existing ones.
- Usually the available tool define what research use. Supporting GeoJson would be a good idea.
- We can use post-transforms to generate heatmaps in MONAI label, tools needed to put together inference outputs.
- Augmentation using slide specific information (random, derived, or fixed)
PIC | Item |
---|---|
Sachi | Give a demo in the next meeting |
David | Share the tools to create a single tiff file from numpy arrays of patches |
All | Add feedbacks and feature request to https://github.com/Project-MONAI/MONAI/discussions/3608 |
There is some documentation on using the large-image-source-vips
package to create a single tiff from numpy arrays: https://girder.github.io/large_image/example_usage.html#writing-an-image . This is used, for example, here: https://github.com/DigitalSlideArchive/HistomicsTK/blob/master/histomicstk/cli/PositivePixelCount/PositivePixelCount.py