Generative adversarial networks accurately reconstruct pan-cancer histology from pathologic, genomic, and radiographic latent features.

By applying pretrained models to virtual biopsy histology in our cohort, we found that intact p53 module expression ( 24 ) [hazard ratio (HR) for recurrence free interval 0.83 95% CI, 0.70 to 0.98; P = 0.02] and high histologic grade signature ( 27 ) (HR 1.12; 95% CI, 0.95 to 1.13; P = 0.17) identified cases with good/poor prognosis, respectively, although the latter did not reach statistical significance (fig.

also in 111,027 other papers