Skip to main content


Fig. 2 | Breast Cancer Research

Fig. 2

From: A whole slide image-based machine learning approach to predict ductal carcinoma in situ (DCIS) recurrence risk

Fig. 2

Full-slide annotation. a List of annotation classes used, and representative examples, alongside the number of ground truth regions available to develop the texture-based annotation classifier. b Multivariate-adjusted p value (Tukey-Kramer) distributions for all 166 features (as points) between all annotated class comparisons. Reference dotted line indicates an adjusted p value of 0.05, with features possessing the significant discriminatory ability (p values < 0.05) situated on the left of it and summarized alongside. c Confusion matrix (which quantifies the performance of the class annotation model) comparing the training ground truth data to the cross-validated annotation classifier test set outputs. The analysis was performed on the original regions before fourfold augmentation

Back to article page