Import FastPathology predictions into QuPath

He we assume that you have some segmentation model that is to be ran on a WSI. The model could be patch-wise or using the low-resolution image.

FastPathology annotation format

In FastPathology, the segmentation predictions are stored in a tiled, pyramidal TIFF format. Each uint of the image correspond to a different class. Hence, we do support multiclass segmentation models, but not multilabel (in the same TIFF file, see below for more information regarding this).

Producing TIFF results

In FastPathology, first create a Project. Then import the images you want to produce predictions on. After import, go to the Process tab, and choose which Pipeline to run. Then click Run pipeline for all images to start generating predictions for all the images in the Project. The result will then be stored within the project. If you only set the name of the project and not the path, the default location is in your user space, e.g., on Windows C:/Users/username/fastpathology/projects/some-project-id/.

Then within the project,

Multilabel data

In order to import multilabel data, that is segmentation results which are produced from different models, for instance. You will need to save the result in different TIFF files in FastPathology. This is only relevant for people that make very custom pipelines. If you are using the built-in methods, multilabel results is handled directly. To import multilabel data, you will need to do two seperate imports and give different names to the annotations as well as change which TIFF image to use.