Table 3 Performance results (for OcnVEGFKO vs WT mice image patches) of an SVC (support vector classifier) with an RBF (radial basis function) kernel (\(C=3\)) with stratified 10-fold cross validation, using 10 or 20 persistence features (see main text) to predict whether an image patch in the test set is from a OcnVEGFKO or WT mouse. Classifiers with 10 features use only quadrant 2 of \(H_0\), whereas 20 features adds additional features from quadrant 1 of \(H_1\), with no improvement. All given measures are mean measures over 100 repeats. Numbers in bold show the best performance results (SHG images, 10 features).

From: Persistent homology analysis distinguishes pathological bone microstructure in non-linear microscopy images