Figure 4.

Cross-correlation of radiomic signature and supervised tumour classification. The Pearson correlation coefficient was calculated for each of the imaging biomarker pairs. If a feature pair had a value of higher than 0.75 it was excluded and replaced by the next highest ranked feature of the same imaging biomarker class. The correlation of the final radiomics signature is depicted in (a). The confusion matrix is based on the training of a linear support vector machine using a 4-fold cross-validation scheme (b). 82.1% of the classified mice were assigned to the correct output class.