Diffusion tensor imaging (DTI) tractography is a novel technique that can delineate the trajectories between cortical region of the human brain non-invasively. In this paper, a novel DTI based white matter fiber tractography using genetic algorithm is presented. Adapting the concepts from evolutionary biology which include selection, recombination and mutation, globally optimized fiber pathways are generated iteratively. Global optimality of the fiber tracts is evaluated using Bayes decision rule, which simultaneously considers both the fiber geometric smoothness and consistency with the tensor field. This global optimality assigns the tracking fibers great immunity to random image noise and other local image artifacts, thus avoiding the detrimental effects of cumulative noise on fiber tracking. Experiments with synthetic and in vivo human DTI data have demonstrated the feasibility and robustness of this new fiber tracking technique, and an improved performance over commonly used probabilistic fiber tracking.