Dear all,

In this post, I give an update about the tiger-algorithm example based on the comments and questions from the webinar.

1. We updated the tiger-example algorithm to check that the writing tile size and the actual size of a tile are the same. Furthermore, we force the tile to be of type uin8, which is required for writing prediction maps with the ASAP software.

2. In some commits (example) of the tiger algorithm example, there is a limit on the number of lymphocytes that could be written (100.000). This was temporarily implemented due to a problem of parsing very large JSON files on grand-challenge. This issue is now fixed and this limitations is not needed anymore (see the latest code here). Please check your code in the rw.py file and check if an if statement limits the number of lymphocytes.

3. In the tiger algorithm example and specific in the function process image tile to detections, we provided illustrative code that you should definitely replace with your own code. However, in the video, I mention that you can detect lymphocytes only in the tumor-stroma segmented region (Label 2), and in the code, I show in an illustrative way how to only create detections in segmented regions with a value of 2 Though this processing step should not be considered as a valid procesing step, but the idea might be useful when you want to detect TILS for cumputing a TILS score. However, this might be confusing as we expect the algorithm to detect lymphocytes everywhere, i.e., not only in the tumor-stroma (Label 2) in leaderboard 1. I hope this clarifies the confusion.

Bet wishes, Mart