Hello, my name is Garrett Banks and I am a 5th year neurosurgery resident interning at Columbia. The work I will be presenting on focuses on creating a diffusion tractographic atlas of the anterior limb of the internal capsule for the purposes of planning approaches for psychiatric neurosurgery interventions. Like most residents, I have nothing to disclose. First, let's discuss the anterior limb of the internal capsule itself, or ALIC for short. The ALIC is almost exclusively comprised of axons that connect the frontal cortex to deeper structures such as the thalamus, the subthalamic nucleus, and other targets such as various areas of the brainstem. Therefore, many of the functions of the frontal lobe, which govern cognitive control, emotional processing and behavior, rely on white matter circuits which traverse the ALIC. It is therefore no surprise that the area has been utilized as a surgical target for treatment of psychiatric disorders. ALIC for OCD, performed either with thermal ablation or gamma knife, involves creation of a lesion in the ALIC. Alternatively, a non-lesional procedure, deep brain stimulation of the ALIC, has been used to treat both OCD and depression. Recently, there has been a move towards improving targeting in deep brain stimulation in psychiatric neurosurgery using tractography. Research into stimulation of the subgenual cingulate for depression has found that stimulation at the intersection between several prominent white matter bundles in the area is correlated with efficacy. In another area, closer to the brainstem, research has shown that stimulation of computationally mapped fiber pathways from the ventral tegmental area through the ALIC leads to better outcomes in deep brain stimulation for depression. However, for the ALIC, consensus has not yet been reached on what exactly should be targeted with a neurosurgical intervention. In this study, we tried to see if an atlas of white matter fibers constructed from normal subjects could aid in targeting in the ALIC. A normalized atlas would potentially allow for more precise targeting of the constituents of the ALIC, while also being a potentially easier method to perform. Of course, a caveat, there will inevitably be patient-to-patient variability in connectivity, which we will be turning a blind eye to by not running individualized tractography. However, we tried to see if the population-normalized location of key fiber bundles is sufficient to guide placement for stimulation. So, to answer this, we first built a population-normalized ALIC fiber bundle atlas, and then compared its behavior results during intraoperative stimulation of the ALIC. First we'll discuss the creation of the fiber bundle atlas. The atlas was created using 100 random subjects from the Human Connectome Project. Tractography for each of the 100 subjects was performed in diffusion space between origin masks of anatomical structures and destinations of interest. Stimulation was limited to voxels with a fractional anastotropy of greater than 0.2 in order to prevent erroneous pathways. Resulting streamlined output distributions were produced and normalized on a per-subject basis by dividing by the total number of successful streamlines. Next, all of these normalized streamlined distributions were warped using linear and nonlinear transformations to MNI 1mm standard space. In 1mm standard space, the data across all subjects was averaged, and the fiber population averages were thresholded at 2% so that only voxels with a significant portion of streamlines across the general population were considered to be part of the fiber bundle. This figure demonstrates the six fibers which were used for the analysis. The anterior nucleus fiber in yellow starts in the anterior nucleus of the thalamus and tracks through the ALIC. The fiber terminates when it tries to exit the ALIC. The medial dorsal nucleus of the thalamus has broad connectivity, so two fibers were generated using it as a starting point. The medial dorsal fiber in light blue starts in the MD nucleus, traverses the ALIC, and terminates at any point in the dorsal frontal cortex. The MD OFC fiber in dark blue starts in the MD nucleus, traverses the ALIC, and terminates at any point in the orbital frontal cortex. The ventral anterior nucleus fiber in green starts in the ventral anterior nucleus of the thalamus and tracks through the ALIC. The fiber terminates when it tries to exit the ALIC, just like the AN fiber. The subthalamic nucleus was also segmented into two fibers due to its broad connectivity. The STN dorsal fiber in orange starts in the STN, traverses the ALIC, and terminates at any point in the dorsal frontal cortex. The STN OFC fiber in red starts in the STN nucleus, traverses the ALIC, and terminates at any point in the orbital frontal cortex. While the full description of the atlas generated is outside the scope of this talk, this slide shows the general organization of the six fibers in a coronal plane near to where the stimulation was performed. The middle slice demonstrates that the VA and AN fibers are the most medial, the STN fibers run the most lateral, and the MD fibers run between the other four. The right slice is the same as the middle slice, except the VA and AN fibers are removed so that one can better visualize the other four fibers. Now we will move to discussing the stimulation data. The stimulation data comes from intraoperative stimulation for patients getting DBS for OCD. Patients had anywhere from two to four separate areas tested intraoperatively, where each of the four contacts was individually tested with monopolar stimulation. Implant locations were either posterior to the anterior commissure near the bed nucleus of the stria terminalis, the BNST, or anterior to the anterior commissure, which will be termed the ventral capsule ventral striatum, or VCVS, for this talk. For each contact's testing, monopolar stimulation was slowly increased by one volt, and patients described changes to their mood and anxiety at each level. A total of 72 different contact locations were tested. Intraoperative CT imaging was obtained for each location, which was used to map electrode locations into standard space. In standard space, 4mm diameter spheres were designed at each area of stimulation to provide a proxy for the volume of activation. To simplify the data for the purposes of analysis, electrodes were classified into four phenotypes. Good electrodes demonstrated only improved mood on stimulation for at least one of the intensities. Mixed electrodes demonstrated improved mood at some intensities, but at other intensities patients described worsening mood or marked anxiety. The most common mixed pattern was improved mood for lower stimulation levels, but marked anxiety at higher stimulation levels. Bad electrodes only demonstrated worsening mood or marked anxiety during stimulation. Null electrodes demonstrated neither positive nor negative effects during stimulation. On the right side is the breakdown of the different phenotypes for all 72 electrodes. If the stimulation data is segregated into VCVS and BNST, the breakdown changes somewhat, but not that much. In other words, stimulation results are not strongly dependent on if you are stimulating at the VCVS or the BNST. A similar effect is seen if the data is segregated into left-sided stimulation and right-sided stimulation. Therefore, stimulation results are not strongly dependent on if you are stimulating the left side or the right side. Now we will look at the stimulation sites as a function of fiber bundles. First, the fiber pathways were compared to the stimulation site volume of activation areas. Only 55 of the electrode sites overlapped with the fibers and were used in the analysis. The excluded sites were inferior near the BNST or in the nucleus accumbens. Only the good and mixed sites were used in the next steps. The VTAs for each of the good and mixed sites were crossed with each of the six normalized fiber distributions to determine which fibers were most stimulated at each site. The mathematical result was termed the activation level. Of note, the units for activation level are meaningless. It is the relative values that are important. The mean activation level was reported for each fiber. When the activation level was averaged across all of the good and mixed sites, two fibers were seen to be much more highly stimulated compared to the rest. The MD-OFC and STN-OFC fibers were more highly stimulated in contacts, whose phenotype was classified as good or mixed. When only the right-sided stimulation areas were examined, the same two fibers were seen to be much more strongly stimulated compared to the others, but there appeared to be a bias towards the STN-OFC fibers. Alternatively, when only left-sided stimulation areas were examined, the same two fibers were seen to be much more strongly stimulated compared to the others like before, but an opposite bias towards the MD-OFC fibers was seen. In conclusion, we found that a standard fiber atlas generated from normal subjects can be used to better understand stimulation data in the ALIC. And intraoperatively, can elicit rapid changes in mood and positive mood changes are associated with fibers traveling to or from the orbital frontal cortex. Future work will investigate into the possibility of the left and right ALIC having similar but slightly different ideal stimulation targets. Finally, I just want to say thank you to all of my mentors and collaborators, as well as to the people of the Human Connectome Project for producing the data used to make the atlas.

neurosurgery

diffusion tractographic atlas

anterior limb of the internal capsule

psychiatric neurosurgery interventions

ALIC