Skip to main content
. 2015 May 28;68(8):614–621. doi: 10.1136/jclinpath-2015-202888

Figure 4.

Figure 4

Tumour viability assessment in whole-slide images of human lung cancer xenografts stained with H&E. (A) Original sample (left), manually annotated sample (middle) and a result image (right). The tumour regions were identified by manually marking the viable (blue) and non-viable (red) tumour regions. In the result image, a heat map is visualised on top of the automatically identified tumour. The same colour coding is used; blue for predicted viable tumour and red for predicted non-viable tumour tissue. *An example of a region where stromal tissue was not identified by a human expert during the annotation but correctly excluded by the segmentation, and **an example of a region where viable tumour was falsely excluded. (B) An example of manual annotation (white line) in finer detail in the original sample. (C) An example of manual annotation (white line) in finer detail in the result image. (D) The correlation between visual examination and the automated assessment for measuring tumour viability.