Skip to main content
Fig. 2 | BMC Cancer

Fig. 2

From: Deep learning-based tumor microenvironment segmentation is predictive of tumor mutations and patient survival in non-small-cell lung cancer

Fig. 2

Calculation and utilization of TIP and TMEC features. a H&E slides from TCGA were downloaded and split into tissue patches. Each patch was classified with ARA-CNN, producing tissue segmentations. These segmentations were next used to calculate the TIP and TMEC features. b Distribution of individual component features in TIP and TMEC. The most often occurring features for TIP were tTUMOR and tLUNG. For TMEC, these were mLUNG, mIMMUNE and mMIXED. c Tasks performed with the help of the TIP and TMEC features. In addition to the TIP and TMEC features, clinical and mutation data was also sourced from TCGA. These datasets were combined and served as input in two tasks: survival prediction and gene mutation classification. The results were compared to those obtained using previous spatial metrics instead of TIP and TMEC

Back to article page