Figure 7
From: Bladder Cancer Treatment Response Assessment in CT using Radiomics with Deep-Learning
Examples of pre- (on the left side of each image pair) and post- (on the right side of each image pair) treatment bladders that responded fully to treatment, and the differences in the predictions by the computer models and radiologists. (a) All three computer methods and the radiologists correctly predicted the outcome of treatment for this case. (b) The three computer methods correctly identified the case as becoming T0 tumor, while the radiologists did not. There was residual bladder wall thickening, presumably related to the treatment, causing the radiologists to falsely conclude that there was persistent tumor. (c) The radiologist correctly identified that there was no residual tumor on post-treatment images, while the three computer methods failed to classify this case correctly. This was likely due to misidentification of perivesical tissue (arrow) as residual tumor by the computer models. The box on the pre-treatment scan represents the location of the lesion as marked by one of the radiologists.
