Fig. 2
From: Distributed radiomics as a signature validation study using the Personal Health Train infrastructure
A schematic diagram explaining the primary methodology for survival analysis used in this study. Details have been provided in the text. Briefly, radiomics features were extracted locally by each institution and then labelled with the radiomics ontology. We then trained a Cox regression model on Lung1 (MAASTRO) and then validated on Lung2 (Radboudumc) by distributing the learning algorithm through the Varian Learning Portal (VLP). Only the event coordinates required to plot a Kaplan-Meier survival curve was returned to MAASTRO, without any identifiable patient-level data.
