Figure 8
From: Bladder Cancer Treatment Response Assessment in CT using Radiomics with Deep-Learning
Venn diagrams of different methods and their assessments for the test set. The inner three circles compare the methods when at least one method correctly predicted the treatment outcome for a pre- and post-treatment pair. The outer circle contains the pre- and post-treatment pairs for which all three methods incorrectly predicted the treatment outcome. (a) The three computer methods correctly predicted the same outcome of the patients for 39% (21/54) of the pre- and post-treatment pairs. (b,c) The two radiologists correctly agreed on the outcome of 43% of the cases ((18 + 5)/54 for (b) and (17 + 6)/54 for (c)). (d,e) Radiologists 1 correctly agreed with the DL-CNN and the RF-SL methods for 19 and 20 cases, respectively, while Radiologist 2 correctly agreed with the DL-CNN and the RF-SL methods for 24 and 21 cases, respectively.
