Automatic Segmentation of Kidney Computed Tomography Images Based on Generative Adversarial Networks

The morphometry of a renal tumor revealed by contrast-enhanced Computed Tomography (CT) imaging is an important factor in clinical decision making surrounding the lesion's diagnosis and treatment. Quantitative study of the relationship between renal tumor morphology and clinical outcomes is difficult due to data scarcity and the laborious nature of manually quantifying imaging predictors. Thus, we proposed an automatic kidney segmentation method, called SegK-GAN. The proposed method comprises a fully convolutional generation network of densely connected blocks and a discrimination network with multi-scale feature extraction. The objective function is optimized using mean absolute error and the dice coefficient. Compared with U-Net, FCN, and SegAN, SegKGAN achieved the highest DSC value of 92.28%, the lowest VOE value of 16.17%, the lowest ASD values of 0.56 mm Our experimental results show that the SegKGAN model have the potential to improve the accuracy of CT-based kidney segmentation.