This paper presents a new segmentation approach based on hybridization of the genetic algorithms (GAs) and seed region growing to produce accurate medical image segmentation, and to overcome the oversegmentation problem. A new fitness function is presented for generating global minima of the objective function, and a chromosome representation suitable for the process of segmentation is proposed. The proposed approach starts by selecting a set of data randomly distributed all over the image as initial population. Each chromosome contains three parts: control genes, gray-levels genes, and position genes. Each gene associates the intensity values by their positions. The region growing algorithm uses these values as an initial seeds to find accurate regions for each control gene. The proposed fitness function is used to evolve the population to find the best region for each control gene. Chromosomes are updated by applying the operators of GAs to evolve segmentation results. Applying the proposed approach to real MRI datasets, better results were achieved compared with the clustering-based fuzzy method.