This paper presents a new strategy for the segmentation of brain images from the volumetric Magnetic Resonance Imaging (MRI). We propose a new segmentation technique that hybridize an evolutionary algorithm, called the Memetic Programming (MP) algorithm, with the Region Growing (RG) technique. The MP algorithm generates new threshold functions and then the RG uses these thresholds to perform an efficient segmentation of MRI images. The proposed segmentation technique are tested through a set of medical images with different noise and Radio Frequency (RF) levels. The experimental results show that the proposed technique produces more accurate and promising results.