Extended Data Fig. 3: Summary of study design and data usage.
From: Deep learning supported discovery of biomarkers for clinical prognosis of liver cancer
a, Information of datasets. b, Training and validation of PaSegNet. c, Acquiring macro mode and micro mode by WSI decoupling and sparsification. d, 10-fold cross-validations of prognosis networks on TCGA dataset. e, Generalization ability test. The prognosis networks were first trained on TCGA dataset and then tested on QHCG dataset. f, Discovery, characterization, and verification of new biomarkers. g, Exploration of macro mode robustness and multiple WSIs selection rule.
