Notes: Tanglethon Validation

These are just some notes on how PHAS was used to validate deep learning-based tangle counting in IHC slides stained for tau protein.

Create a new task for validation

First, create a new labelset that will be used for this::

flask labelset-add -d “Tanglethon validation” tangleval samples/labelsets/tangleval.json

Then add the actual task::

flask tasks-add –json samples/tasks/tangleval.json

Generate actual samples at random from curves

This is the command::

flask samples-random-from-annot -r 100 -n 40 -u pyushkevich -c 1 3 candidate_roi 2048

Notes: * 100 is the standard deviation of the random offset of the box center from curves * 40 is the number of boxes per slide to generate * 1 is the ID of the annotation task used to draw the curves * 3 is the ID of the newly added task * candidate_roi is the label assigned to new boxes * -c clobbers existing boxes with label candidate_roi * 2048 is the size of the box