Published June 16, 2019
| Version v0.1
Dataset
Open
Tissue blocks for EBV detection in stomach cancer
Description
Tissue blocks from hematoxylin and eosin (H&E) images of stomach adenocarcinoma samples (STAD) from our validation cohort
They are grouped into four classes:
- false negative: true EBV+ patient, block predicted to be EBV-
- false positive: true EBV- patient, block predicted to be EBV+
- true negative: true EBV- patient, block predicted to be EBV-
- true positive: true EBV+ patient, block predicted to be EBV+
all blocks are 224 x 224 px, corresponding to 256x256 µm
The blocks represent only a subset of the full validation cohort, selected as follows
- from each patient, only a maximum of 50 randomly drawn blocks was used
- we saved the upper and lower quartile (sorted by EBV score)
- we then grouped into true and false positive (high EBV score) and negative (low EBV score), respectively