Joint one‐sided synthetic unpaired image translation and segmentation for colorectal cancer prevention

Deep learning has shown excellent performance in analysing medical images. However, datasets are difficult to obtain due privacy issues, standardization problems, and lack of annotations. We address these problems by producing realistic synthetic images using a combination 3D technologies generative adversarial networks. propose CUT-seg, joint training where segmentation model jointly trained produce while segment polyps. take advantage recent one-sided translation models because they use significantly less memory, allowing us add the loop. CUT-seg performs better, is computationally expensive, requires real than other memory-intensive image approaches that require two stage training. Promising results achieved on five polyp only one zero As part this study we release Synth-Colon, an entirely dataset includes 20,000 colon additional details about depth geometry: https://enric1994.github.io/synth-colon