Fig. 7:

Classification performance for prostate cancer classification repeatability for 1694 features derived from K (kurtosity) of kurtosis function, using the area under receiver operator characteristic curve (AUC). The scatterplots show repeatability of radiomic features (x-axis) with respect to AUC (y-axis), for all features (top), and features having high repeatability only (ICC(3,1)>0.8, bottom). Measurements are from single DWI scan (Scan1 and Scan2) and when radiomic feature values are averaged (Avg Scan 1&2).