Skip to content

Latest commit

 

History

History
32 lines (18 loc) · 1.89 KB

4_Testing pixel classifier for 900 series.md

File metadata and controls

32 lines (18 loc) · 1.89 KB

Analysis of the 900 series images received in May, 2024

Total files = 39 tiff images Sample information

Screenshot 2024-05-21 at 8 56 08 AM

1. A subset of slides were selected at random while ensuring that all the treatment groups were represented

sample_subset_for_tiling

2. Tiles created from this subset

Screenshot 2024-05-22 at 11 32 25 AM

Total tiles = 902 Average number of tiles / image = ~55

3. A subset of tiles (total=80; 5 tiles/sample) were randomly selected to test the pixel classifier trained on series-700 samples. The subset was selected in a way to include tiles having variable proportions of collagen and tissue.

4. Collagen masks exported from QuPath were overlaid on original MT images using CellProfiler - some example overlays below

900series_classifier_test1

900series_testclassifier_3

900series_testclassifier_2

900series_testclassifier_4

Overall Impression

There is some overcalling in general, but it is avoiding common pitfalls such as incorrectly calling lymphocyte clusters as collagen.