Skip to main content
. 2022 Jun 7;20:261. doi: 10.1186/s12967-022-03458-9

Fig. 2.

Fig. 2

Overall workflow of this study. a Two adjacent sections stained with CD3 and CD8 are digitalized by using the whole-slide scanner. b By using a semi-automated image analysis algorithm, a binary mask of segmented tumour region is created for each WSI. c The CD3+ and CD8+ T-cells in the tumour region are segmented and identified by using a fully-automated algorithm. d The CD3-score and CD8-score (normalized CD3+ and CD8+ cell density, ranging from 0 to 100) are classified into low and high based on the cutoffs determined by maximally selected rank statistics, respectively. e, f, g A three-category I-score and a two-category I-score are established by integrating the classifications of CD3-score and CD8-score based on the discovery dataset. h The validation cohort is used to assess the prognostic value of the I-score