Fig. 2.

The deep learning model for glioma cancer patient stratification. (A) The architecture of the encoder we established. (B) Training and validation MSE loss iteration during model optimization in cross validation. (C)Glioma patient subgrouping visualization based on survival-associated low-dimensional features from the autoencoder. (D) Survival analysis for high-risk subgroup and low-risk subgroup based on survival-associated low-dimensional features from the autoencoder.