Introduction The functional neuroanatomy of depression involves the prefrontal, cingulate and insular cortices, the amygdala, the nucleus accumbens and the hippocampus. The anterior and lateral prefrontal cortices are different in their cognitive and emotional integration roles and their modulation of sub-cortical regions. Both offer a distinct opportunity for targeted antidepressant treatments. Epidural prefrontal cortical stimulation (EpCS) is a unique therapeutic approach. It is more direct than transcranial magnetic stimulation (TMS) or vagus nerve stimulation (VNS), safer than deep brain stimulation (DBS) and, when applied chronically can reduce negative emotional arousal, increases stress tolerance and possibly lead to fewer relapses than observed with antidepressant pharmacotherapy Question Document the direct anatomical connections accessed with bilateral EpCS over fronto-polar and mid-frontal gyrus using diffusion tensor imaging (DTI). Methods Five patients treatment-resistant depression were implanted with bilateral EpCS leads over fronto-polar and mid-frontal gyrus. Patients underwent structural imaging (high resolution T1-weighted (T1w) MRI image and a DTI with 32 gradient directions) before surgery and a CT scan post-implantation. Image analysis was performed using the FMRIB Software Library. BA25, BA10 and BA46 areas, and Amygdala were extracted from the Broadmann area template to serve as guides for tractography. Similarly, individual EpCS paddles identified on CT scan were also transformed. First, an affine (12 parameter model) registration was applied to register Broadmann template onto T1w images and then T1w onto DTI data. Transformations were then applied to the guides. Second, a binary mask of the brain was extracted and applied to all images after correction for eddy currents. BedpostX was then run on the preprocessed data to generate the basis for probabilistic tractography using ProbtrackX. Estimation of tracts was done probabilistically, using a dual-fiber model with 5000 tract-following samples at each voxel. All tractography was done in each subject’s native space. Eight tractography combinations (left & right) were performed: BA10/BA46 as seeds for tractography and Amygdala/BA25 as waypoints masks. Similar approached was applied with EpCS actual lead masks. The resulting fibers were then thresholded at 0.01 (99% likelihood of connectivity) and then binarized, and transferred into MNI152 standard space for cross-subject analysis using a non-linear registration. To generate population-based probability fibers, all subjects’ individual fibers were added up and used as a basis for calculation of a mean image. Finally, the resulting mean fiber was thresholded to show voxels common to 80% or more of the subjects. Results All patients showed robust connections of different pathways in both left and right hemispheres. The resulting pathways connected to each seed combination are shown in Fig. 1 . Electrodes implanted over both BA10 and BA46 areas resulted in direct activation of axons that primarily project from BA10 or BA46 to BA25 and Amygdala, and extend to the cingulum bundle and to the middle occipital gyrus. Conclusion This is the first demonstration of EpCS direct anatomical access to the medial and lateral mood regulation networks. Encouraging clinical data have been reported elsewhere with 3 patients out of 5 showing remission at 7 months (Nahas et al., 2010) and at 5 years (Williams et al., under review). Bilateral anterior and lateral EpCS promises to be an innovative new treatment for a highly morbid disease through adaptive cortical governance of limbic structures but larger studies are needed.