Fig. 1From: Assessment of emerging pretraining strategies in interpretable multimodal deep learning for cancer prognosticationStudy workflow overview; neural networks are pretrained using various methods to encode gene expression, DNA methylation, and histopathology whole slide imaging (denoted whole slide graph; WSG) data. Multimodal prognostication models are developed using these encoders pretrained using these methods, to assess the utility of each oneBack to article page