Abstract

It is necessary to have a precise evaluation of the morphometry of a kidney tumor in make decisions regarding treatment and diagnosis accurately. It is difficult to undertake quantitative research between kidney tumor morphology and clinical outcomes because there is a lack of data and the need to manually evaluate imaging parameters, which is a lengthy process. A conventional generative adversarial network model serves as the foundation for the STGAN approach, which is an autonomous kidney segmentation method that is proposed in this research as a solution to this problem. The primary foundation of this system is comprised of a completely convolutional generating network that is made up of densely linked blocks and a discriminator network that is equipped with multi-scale feature extraction as well. For conducting quantitative and qualitative comparisons with the STGAN methodology, the medical image segmentation networks U-Net, FCN, and SegAN are deployed. STGAN achieves greater performance in comparison to the other neural networks, the model that we have proposed demonstrates potential for improving the accuracy of CT- based kidney segmentation.