Fully Automated Construction of a Deep U-Net Network Model for Medical Image Segmentation

In recent years, the use of deep learning technology for image processing has become mainstream, and the U-Net network has received widespread attention owing to its unique U-shaped structure, which has achieved excellent results in the field of image segmentation, especially in medical image segmentation. To enhance the performance of the U-Net network model and establish better U-Net design variables, in this paper, we propose a fuzzy-controlled multicellular gene expression programming algorithm to automatically build and optimize the U-Net. The algorithm creates an efficient variable-length gene code, generates chromosomes for the optimization of U-Net design variables, decodes the chromosomes to construct the U-Net model, dynamically calculates population fitness and fuzzy affiliation values, and achieves the optimal U-Net network through continuous evolution. The experimental results indicate that the proposed algorithm outperforms U-Net, Fully Convolutional Networks32, and VanillaUnet in image recognition segmentation, especially in the segmentation of COVID-19 CT medical images.