Genetic Algorithm Enhanced nnU-Net for the MICCAI KiTS23 Challenge

Deep learning-based segmentation techniques have been gaining increasing attention in recent years due to their potential in various medical image segmentation tasks, particularly in the segmentation of kidneys, renal tumors, and renal cysts. One of the major challenges in medical image segmentation is the scarcity of high-quality training data, which often limits the effectiveness and robustness of segmentation algorithms. To address this issue, a novel genetic algorithm (GA) based approach that combines nnU-Net framework is proposed to improve the robustness of medical image segmentation. The proposed approach involves a two-stage process. In the first stage, a set of convolutional neural network (CNN) models are trained with loss function. In the second stage, GA is applied to evolve a population of CNN models with different sets of hyperparameters. This results in a final CNN model with improved robustness and better segmentation performance.