A unified connectomic target for deep brain stimulation in obsessive-compulsive disorder

Abstract

Multiple surgical targets for treating obsessive-compulsive disorder with deep brain stimulation (DBS) have been proposed. However, different targets may modulate the same neural network responsible for clinical improvement. We analyzed data from four cohorts of patients (N = 50) that underwent DBS to the anterior limb of the internal capsule (ALIC), the nucleus accumbens or the subthalamic nucleus (STN). The same fiber bundle was associated with optimal clinical response in cohorts targeting either structure. This bundle connected frontal regions to the STN. When informing the tract target based on the first cohort, clinical improvements in the second could be significantly predicted, and vice versa. To further confirm results, clinical improvements in eight patients from a third center and six patients from a fourth center were significantly predicted based on their stimulation overlap with this tract. Our results show that connectivity-derived models may inform clinical improvements across DBS targets, surgeons and centers. The identified tract target is openly available in atlas form.

Fig. 1. Overview of lead electrode placement.

The two training/cross-validation cohorts (left) targeting ALIC (Cologne) and STN (Grenoble), and the two test cohorts (right) targeting NAcc (Madrid) and both ALIC & STN with four electrodes per patient (London) are shown. Subcortical structures defined by CIT-168 Reinforcement Learning Atlas (ALIC/NAcc region) and DISTAL Atlas (STN region), with coronal and axial planes of the T1-weighted ICMB 152 2009b nonlinear template as background.

Fig. 2. Predictive fiber tracts in training cohorts.

Top: all fibers connected to the sum of volumes of tissue activated (VTAs) of each cohort are shown in green. Middle: predictive fibers positively associated with clinical improvement are shown in red. Only positive fibers are shown here for reasons of clarity. See Fig. 3 for negatively associated tracts. The top 20% predictive fibers are displayed. Bottom: correlations between the degree of stimulating positively predictive tracts (sum of aggregated fiber T-scores under each VTA) and clinical improvements. Gray shaded areas represent 95% confidence intervals. This analysis is based on a normative connectome, a replication of it based on anatomically predefined pathways is shown in Supplementary Fig. 1.

Fig. 3. Cross-prediction between ALIC and STN training cohorts.

Top: when the tract was calculated exclusively based on data from the ALIC cohort, it was used to calculate fiber T-scores for all patients in the STN cohort. These were correlated with clinical improvements in the STN cohort. One example patient with strong overlap of the tract (yellow) received a high fiber T-score, whereas one with less overlap received a lower score (blue). The two example patients are marked in the correlation plot on the left. Bottom: here, the tract was calculated exclusively on data from the STN cohort to predict outcome in patients in the ALIC cohort. Again, two example patients are shown. Gray shaded areas in the correlation plots represent 95% confidence intervals. Of note, here, some VTAs barely overlapped with the tract (as the blue example VTA) and consequently received a near-zero score.

Fig. 4. Predictions in test cohorts.

Top: predictive fibers calculated on both training cohorts (Cologne & Grenoble) irrespective of their target. Red fibers are positively associated with clinical improvement, blue fibers negatively. Bottom: the sum of aggregated fiber T-scores under each VTA predicted %-Y-BOCS improvements in eight patients with four settings each (N = 32 stimulations) of the Madrid cohort (left) and six patients of the London cohort with dual stimulation (four electrodes) of STN and ALIC (right). Gray shaded areas represent 95% confidence intervals. Please note that p-values in this manuscript are based on random permutation testing. Based on classical tests, the result shown in the lower right panel would remain significant in a one-sided test, only (p-one-sided = 0.044, p-two-sided = 0.089). A replication of this result based on anatomically predefined pathways is shown in Supplementary Fig. 2.

Fig. 5. Literature defined OCD targets in relationship to the identified tract.

Overview of the positively predictive fiber tracts identified in the present study are shown in synopsis with DBS targets for treatment of OCD from reported studies. Note that most targets were reported for the tip of the electrode, thus, active stimulation may have occurred dorsal to shown targets (Table 2). a, b Reported average targets mapped to standard space. c The degree of weighted overlap between stimulation sites and the identified tract. These were correlated with reported average %-Y-BOCS improvements of published studies (where available, other sites marked in gray; see Supplementary Methods for details). Gray shaded area represents 95% confidence intervals.

Fig. 6. Anatomical course of discriminative fibers shown in MNI space.

The tract is connected to the subthalamic nucleus and mediodorsal nucleus of the thalamus, traverses through the anterior limb of the internal capsule and has a wide array of frontal connections including dorsal anterior cingulate cortex and ventrolateral prefrontal cortex. The tract most negatively associated with clinical improvement was the anterior commissure.

Fig. 7. Summary of methods to define a T-value for each tract.

a For each fiber, VTAs were grouped into either connected (C; yellow) or unconnected (UC; blue) sets across patients. b Two-sample t-tests between clinical improvements in connected and unconnected VTAs were calculated in a mass-univariate fashion for each fiber tract separately. c The resulting T-value of this analysis leads to the “weight” that each fiber received, as well as the color in visualizations throughout the manuscript. Here, red means that the fiber tract is favorably connected to good responders, whereas blue indicates the opposite (and the saturation of tracts denotes how discriminative they are).