Skip to main content
. 2020 Jul 23;10:12366. doi: 10.1038/s41598-020-69106-8

Figure 1.

Figure 1

Proposed DRTOP model, where 3D CT and PET images go through separate networks, which are unsupervisedly pre-trained on an independent dataset. The outputs of the two networks (referred to as the CT risk and the PET risk) are combined with other clinical factors and fed to a Cox PHM.