Fig. 2: Classification performances of imCMS models trained and evaluated using data from different combinations of cohorts. | npj Precision Oncology

Fig. 2: Classification performances of imCMS models trained and evaluated using data from different combinations of cohorts.

From: Image-based consensus molecular subtyping in rectal cancer biopsies and response to neoadjuvant chemoradiotherapy

Fig. 2

Numbers are macro-averages of the AUROC for ensembles of five models trained and validated with five different folds of the same development set. Each model of an ensemble model was evaluated separately: the mean and standard deviation of the corresponding macro-average AUROCs is reported in parenthesis. The ensemble model resulting of the experiment [E3a] was used for subsequent analysis of association between imCMS and pCR to neoadjuvant LCRT.

Back to article page