Fig. 5: Model performance for prediction of BCE and non-BCE.
From: Multi-protein spatial signatures in ductal carcinoma in situ (DCIS) of breast
a Classification model was developed using logistic regression to predict BCE and non-BCE using two input variables, one comprising of cluster 2 and 4, and the other comprising of cluster 5 and 6. The model had a sensitivity of 77%, specificity of 79% with an error rate of 21.6% and AUC of 0.785. The AUC in the leave one out cross-validation was 0.739. b Escore was developed from the classification model and binary Escore was used for log-rank analysis and data represented using Kaplan–Meier plots.
