Targeting based on patient-specific 7 Tesla MRI connectivity analysis improves deep brain stimulation for Parkinson’s disease Authors: Yarit Wiggerts, Rick Schuurman, Rob M. A. de Bie, Martijn Beudel, Wietske van der Zwaag, Pepijn van den Munckhof, Maarten Bot Published: 2026 ABSTRACT Deep brain stimulation (DBS) of the subthalamic nucleus (STN) is an effective treatment for Parkinson’s disease (PD), yet individual motor outcome remains insufficiently predictable. The effect of DBS depends on modulating brain networks, and this study evaluates whether targeting using patient-specific connectivity with 7-Tesla (7T) MRI probabilistic tractography enhances motor outcome. For connectivity analysis, selective visualization of STN connectivity to the primary cortical motor areas was performed before surgery, and electrode lead placement was directed at the connectivity-derived motor subdivision of the STN. Motor outcome, measured using the Movement Disorders Society Unified Parkinson Disease Rating Scale motor part (MDS-UPDRS part III), of 102 PD-DBS patients for whom 7T MRI connectivity analysis was used for DBS targeting, was compared to a matched control group of 118 PD-DBS patients for whom 7T MRI T2-weighted classical anatomical landmarks were used for targeting. Results of this prospective comparison show that patients operated with connectivity-guided targeting showed significantly more motor improvement compared to the control group: 56 ± 15% vs. 50 ± 20% (p = 0.015), and the connectivity-guided group showed a higher optimal response rate (96% vs. 86%, p = 0.008). Connectivity analysis enabled localization of DBS electrodes in reference to the motor subdivision. The leads that were optimally located inside the motor subdivision showed an average of 60 ± 11% motor improvement. Our findings demonstrate the clinical applicability of 7T MRI probabilistic tractography for visualizing connectivity between the STN and cortical motor areas to enable electrode lead placement directed at the motor subdivision of the STN, introducing patient-specific connectivity analysis guided DBS. This proved to be an individual biomarker for implantation location that increased the effectiveness of DBS in patients with PD. Introduction The effect of deep brain stimulation (DBS) in Parkinson’s disease (PD) relies on the modulation of malfunctioning (motor) brain networks by delivering electrical pulses within the subthalamic nucleus (STN)1–4. Factors that affect the outcome after DBS in PD include accurate patient selection, correct electrode lead placement, and selection of electrical parameter programming in the basal ganglia motor network5–8. High-resolution anatomical group-level hotspots/sweetspots and normative connectomes have been used to visualize networks in great detail, identifying circuits that are associated with successful neuromodulation therapy9. However, the generalizability of motor networks based on normative connectomes to patient-specific alterations remains uncertain10. We recently demonstrated 7 Tesla (7T) MRI connectivity analysis, using the combination of T2-weighted and diffusion weighted images (DWI), for visualizing individual connections of the STN to various cortical areas11–13. We hypothesized that stimulation in an area with a high local density of probabilistic subthalamic connectivity to the primary cortical motor areas was associated with more (hemi-body) motor improvement after DBS. This connectivity-derived STN motor subdivision could serve as a possible biomarker for electrode lead placement11,12,14. Previous research has already illustrated the effectiveness of targeting that involves pathways between the STN and cortex, especially connections to the SMA and M1, both using tractography analysis and electroencephalography8,15,16. In this study, we performed 7T MRI probabilistic tractography and visualized connections between the STN and motor, associative, and limbic cortical areas, to enable electrode lead placement directed at the motor subdivision, introducing patient-specific connectivity analysis guided DBS. Results From April 2022 through October 2024, in 132 PD patients, electrode lead placement was performed using the 7T connectivity analysis, directed at the connectivity-derived motor subdivision of the STN. Thirteen patients were excluded due to participation in another prospective clinical DBS trial; six patients had a failed follow-up measurement; four patients had deviating impedances or other system failures hindering selection of optimal stimulation parameters; for three patients a postoperative CT-scan was not available; two patients had unilateral DBS placement, and two patients had insufficient 7T MRI quality for connectivity analysis, resulting in a total of 102 patients suitable for comparison. Baseline characteristics are presented in Table 1. The control group, with T2-weighted classical DBS targeting, consisted of 118 patients, who were operated from October 2019 through March 2022, when connectivity analysis was not yet available (N = 91), or between April 2022 and October 2024, in whom 7T DWI was attempted, but failed, mainly due to limited patient endurance (N = 27). In this group of patients, DBS was performed with visualization of the STN using 7T T2-weighted MRI only, and electrode placement was directed at the posterolateral part of the STN by classical anatomical landmarks, using the anterior border of the red nucleus (Fig. 1).Fig. 17T MRI connectivity analysis targeting versus 7T MRI T2 classical anatomical targeting.A–C connectivity DBS targeting and D classical T2 MRI DBS targeting. A For connectivity DBS targeting, the 7T T2 MRI is used for delineating a three-dimensional STN mask (pink). B 7T DWI is added for probabilistic tractography in which the cortical motor areas are used to visualize the connectivity-derived STN motor subdivision. C The location of the STN motor subdivision (blue) inside the STN (pink) is used as a target for lead placement. D In classical 7T T2 MRI targeting, the center of the hypointense STN is determined with the anterior border of the red nucleus and used as a target for lead placement. (Red circle: red nucleus; white dot: lead target. Top panels: axial view; bottom panels: coronal view). MRI magnetic resonance imaging, DWI diffusion-weighted imaging, DBS deep brain stimulation, STN subthalamic nucleus.Table 1Baseline characteristics of the connectivity-guided group versus the control group7T T2 group (N = 118)Connectivity analysis group (N = 102)p-valueAge (±SD)64.4 ± 8.165.1 ± 7.70.623Sex (M, %)75 (64%)52 (51%)0.060Duration of illness (years ± SD)9.7 ± 4.811.2 ± 5.00.028aDays to follow-up (±SD)233.6 ± 72.1232.3 ± 53.50.881Pre DBS MDS-UPDRS III improvement (% ± SD)55.5 ± 17.7 (N = 117)b59.1 ± 16.9 (N = 102)b0.126Pre DBS MDS-UPDRS III absolute score (± SD)45.5 ± 11.744.6 ± 14.00.602Pre DBS MDS-UPDRS I score (±SD)13.0 ± 5.7 (N = 101)b12.9 ± 5.3 (N = 94)b0.832Pre DBS MDS-UPDRS II score (±SD)15.9 ± 6.7 (N = 88)b16.7 ± 6.4 (N = 94)b0.435Pre DBS MDS-UPDRS IV score (±SD)10.5 ± 3.4 (N = 105)b10.6 ± 3.4 (N = 98)b0.730Pre DBS PDQ39 score (±SD)32.1 ± 11.8 (N = 95)b32.1 ± 11.3 (N = 95)b1.000Pre DBS SAS score (±SD)6.8 ± 5.8 (N = 104)b6.6 ± 4.5 (N = 97)b0.848Pre DBS BDI score (±SD)12.2 ± 4.2 (N = 39)b11.7 ± 6.0 (N = 91)b0.600Pre DBS LEDD score (±SD)1399.5 ± 670.31297.3 ± 582.00.232Average applied amount of current (±SD)2.2 ± 0.72.1 ± 0.60.212SD standard deviation, DBS deep brain stimulation, UPDRS Unified Parkinson’s Disease Rating Scale, PDQ Parkinson’s Disease Questionnaire, SAS Starkstein Apathy Scale, BDI Beck depression inventory, LEDD levodopa equivalent daily dose.aSignificantly different. A p-value of <0.05 is considered statistically significant.bNot all scores were available for all patients. 7T MRI connectivity analysis targeting versus 7T MRI T2 classical anatomical targeting. A–C connectivity DBS targeting and D classical T2 MRI DBS targeting. A For connectivity DBS targeting, the 7T T2 MRI is used for delineating a three-dimensional STN mask (pink). B 7T DWI is added for probabilistic tractography in which the cortical motor areas are used to visualize the connectivity-derived STN motor subdivision. C The location of the STN motor subdivision (blue) inside the STN (pink) is used as a target for lead placement. D In classical 7T T2 MRI targeting, the center of the hypointense STN is determined with the anterior border of the red nucleus and used as a target for lead placement. (Red circle: red nucleus; white dot: lead target. Top panels: axial view; bottom panels: coronal view). MRI magnetic resonance imaging, DWI diffusion-weighted imaging, DBS deep brain stimulation, STN subthalamic nucleus. Table 1Baseline characteristics of the connectivity-guided group versus the control group7T T2 group (N = 118)Connectivity analysis group (N = 102)p-valueAge (±SD)64.4 ± 8.165.1 ± 7.70.623Sex (M, %)75 (64%)52 (51%)0.060Duration of illness (years ± SD)9.7 ± 4.811.2 ± 5.00.028aDays to follow-up (±SD)233.6 ± 72.1232.3 ± 53.50.881Pre DBS MDS-UPDRS III improvement (% ± SD)55.5 ± 17.7 (N = 117)b59.1 ± 16.9 (N = 102)b0.126Pre DBS MDS-UPDRS III absolute score (± SD)45.5 ± 11.744.6 ± 14.00.602Pre DBS MDS-UPDRS I score (±SD)13.0 ± 5.7 (N = 101)b12.9 ± 5.3 (N = 94)b0.832Pre DBS MDS-UPDRS II score (±SD)15.9 ± 6.7 (N = 88)b16.7 ± 6.4 (N = 94)b0.435Pre DBS MDS-UPDRS IV score (±SD)10.5 ± 3.4 (N = 105)b10.6 ± 3.4 (N = 98)b0.730Pre DBS PDQ39 score (±SD)32.1 ± 11.8 (N = 95)b32.1 ± 11.3 (N = 95)b1.000Pre DBS SAS score (±SD)6.8 ± 5.8 (N = 104)b6.6 ± 4.5 (N = 97)b0.848Pre DBS BDI score (±SD)12.2 ± 4.2 (N = 39)b11.7 ± 6.0 (N = 91)b0.600Pre DBS LEDD score (±SD)1399.5 ± 670.31297.3 ± 582.00.232Average applied amount of current (±SD)2.2 ± 0.72.1 ± 0.60.212SD standard deviation, DBS deep brain stimulation, UPDRS Unified Parkinson’s Disease Rating Scale, PDQ Parkinson’s Disease Questionnaire, SAS Starkstein Apathy Scale, BDI Beck depression inventory, LEDD levodopa equivalent daily dose.aSignificantly different. A p-value of <0.05 is considered statistically significant.bNot all scores were available for all patients. Baseline characteristics of the connectivity-guided group versus the control group SD standard deviation, DBS deep brain stimulation, UPDRS Unified Parkinson’s Disease Rating Scale, PDQ Parkinson’s Disease Questionnaire, SAS Starkstein Apathy Scale, BDI Beck depression inventory, LEDD levodopa equivalent daily dose. aSignificantly different. A p-value of <0.05 is considered statistically significant. bNot all scores were available for all patients. Clinical improvement Follow-up measurement was conducted 6–8 months (229 ± 54 days) after surgery, and motor score improvement was assessed between baseline and follow-up measurement using the Movement Disorders Society Unified Parkinson Disease Rating Scale Part III (MDS-UPDRS-III). In the 7T MRI connectivity analysis group, a significantly higher MDS-UPDRS III improvement score was observed compared to the 7T MRI T2-weighted classical targeting control group, 56 ± 15% versus 50 ± 20%, (p = 0.015, Table 2). Also, in absolute MDS UPDRS-III score change, this difference was significant (29 ± 4 versus 25 ± 0, p = 0.012, Table 2). Furthermore, significantly fewer poor responders (<30% improvement on the MDS-UPDRS III) in the 7T MRI connectivity analysis group were observed compared to the control group (4% versus 14%, p = 0.008, Table 2). No significant differences in PD subtypes (hypokinetic-rigid symptoms (MDS-UPDRS III sub-items 3.3, 3.4–3.8), tremor (MDS-UPDRS III sub-items 3.15–3.17), and axial symptoms (MDRS-UPDRS sub-items 3.1, 3.10–3.13)) were present between the connectivity-based targeted group and the control group.Table 2Primary and secondary outcome measures of connectivity-guided group versus control group7T T2 group (N = 118)Connectivity analysis group (N = 102)p-valueMDS-UPDRS III improvement (% ± SD)49.8 ± 19.655.6 ± 15.20.015a# of non-responders (<30% UPDRS-III improvement, %)17 (14%)4 (4%)0.008aMDS-UPDRS III absolute change (±SD)25 (±0)29 (±4)0.012aMDS-UPDRS I improvement (% ± SD)7.1 ± 64.4 (N = 96)b16.7 ± 48.2 (N = 83)b0.269MDS-UPDRS II improvement (% ± SD)30.0 ± 41.8 (N = 82)b31.0 ± 48.2 (N = 83)b0.875MDS-UPDRS IV improvement (% ± SD)55.9 ± 47.9 (N = 102)b66.5 ± 32.5 (N = 88)b0.072PDQ39 improvement (% ± SD)18.3 ± 43.2 (N = 87)b10.0 ± 50.4 (N = 85)b0.247SAS improvement (absolute score ± SD)−2.2 ± 6.1 (N = 98)b−0.5 ± 4.3 (N = 88)b0.031aSAS >5 points worsening (%)29 (30) (N = 98)b13 (13) (N = 72)b0.018aBDI improvement (% ± SD)8.1 ± 46.1 (N = 37)b8.1 ± 57.1 (N = 82)b0.999LEDD score improvement (% ± SD)44.0 ± 84.954.3 ± 35.20.253UPDRS Unified Parkinson’s Disease Rating Scale, SD standard deviation, PDQ Parkinson’s Disease Questionnaire, SAS Starkstein Apathy Scale, BDI Beck Depression Inventory, LEDD levodopa equivalent daily dose.aSignificantly different. A p-value of <0.05 is considered statistically significant.bNot all scores were available for all patients. Table 2Primary and secondary outcome measures of connectivity-guided group versus control group7T T2 group (N = 118)Connectivity analysis group (N = 102)p-valueMDS-UPDRS III improvement (% ± SD)49.8 ± 19.655.6 ± 15.20.015a# of non-responders (<30% UPDRS-III improvement, %)17 (14%)4 (4%)0.008aMDS-UPDRS III absolute change (±SD)25 (±0)29 (±4)0.012aMDS-UPDRS I improvement (% ± SD)7.1 ± 64.4 (N = 96)b16.7 ± 48.2 (N = 83)b0.269MDS-UPDRS II improvement (% ± SD)30.0 ± 41.8 (N = 82)b31.0 ± 48.2 (N = 83)b0.875MDS-UPDRS IV improvement (% ± SD)55.9 ± 47.9 (N = 102)b66.5 ± 32.5 (N = 88)b0.072PDQ39 improvement (% ± SD)18.3 ± 43.2 (N = 87)b10.0 ± 50.4 (N = 85)b0.247SAS improvement (absolute score ± SD)−2.2 ± 6.1 (N = 98)b−0.5 ± 4.3 (N = 88)b0.031aSAS >5 points worsening (%)29 (30) (N = 98)b13 (13) (N = 72)b0.018aBDI improvement (% ± SD)8.1 ± 46.1 (N = 37)b8.1 ± 57.1 (N = 82)b0.999LEDD score improvement (% ± SD)44.0 ± 84.954.3 ± 35.20.253UPDRS Unified Parkinson’s Disease Rating Scale, SD standard deviation, PDQ Parkinson’s Disease Questionnaire, SAS Starkstein Apathy Scale, BDI Beck Depression Inventory, LEDD levodopa equivalent daily dose.aSignificantly different. A p-value of <0.05 is considered statistically significant.bNot all scores were available for all patients. Primary and secondary outcome measures of connectivity-guided group versus control group UPDRS Unified Parkinson’s Disease Rating Scale, SD standard deviation, PDQ Parkinson’s Disease Questionnaire, SAS Starkstein Apathy Scale, BDI Beck Depression Inventory, LEDD levodopa equivalent daily dose. aSignificantly different. A p-value of <0.05 is considered statistically significant. bNot all scores were available for all patients. The subset of 27 patients who were operated in the epoch of connectivity analysis (2022–2024) but in whom 7T DWI could not be obtained and were therefore included in the classical targeting group, had a MDS-UPDRS III improvement of 47.4 ± 19.4% with 2 (11%) non-responders, which was not a significantly different subset within the total control group of N = 118. We also compared this subset of 27 patients from the epoch of connectivity analysis, but without DWI, separately to the 7T MRI connectivity analysis group. These 27 patients by themselves also had a significantly lower MDS-UPDRS III motor improvement compared to the connectivity analysis group (47 ± 19% versus 56 ± 15%, p = 0.021, absolute score changes 24 ± 12 versus 29 ± 4, p = 0.050). The secondary outcome measures included the change from baseline to follow-up on the MDS-UPDRS I (non-motor experiences of daily living), MDS-UPDRS II (motor experiences of daily living), MDS-UPDRS IV (motor complications), Parkinson’s Disease Questionnaire (PDQ39), quality of life (QoL), Starkstein apathy scale (SAS), and Beck Depression Inventory (BDI). In the 7T MRI connectivity analysis group, significantly less worsening on the SAS rating scale was seen compared to the 7T MRI T2-weighted group after DBS (p = 0.031, Table 2). Furthermore, there is no known literature on the minimal clinically important change score (MCID) of the SAS, but there were significantly fewer patients who had an increase of ≥5 points on the SAS rating scale in the 7T MRI connectivity group, a cut-off value that is often chosen in our center (13% versus 30%, p = 0.018, Table 2). The DBS parameters (frequency, pulse width, and current amplitude) were not statistically different between the two groups. Optimal DBS location using connectivity analysis For the connectivity analysis group, the location of the three STN subdivisions in reference to each of the four electrode lead contact levels was visualized for both electrode leads (102 patients representing 204 leads) (Fig. 2). In 85 patients, both electrode leads were optimally placed in the motor subdivision as intended. Placement was considered inside the motor subdivision if at least one electrode contact level was located inside the STN motor subdivision. We found that the single active electrode levels at the time of follow-up inside the motor subdivision were, on average, surrounded by a high density (55 ± 19%) of motor connectivity. The clinical outcome of these 85 patients was also compared to the 7T MRI T2-weighted control group (See Table 3 for demographical data). Improvement on the MDS-UPDRS III after DBS for this subgroup was 60 ± 11% (p < 0.001, Table 4). Also, in the absolute MDS UPDRS-III score, this difference was significant (31 ± 6 versus 25 ± 0, p < 0.001, Table 4). Furthermore, in this subgroup, no poor responders were seen, with a range of 40%–83% improvement on the MDS-UPDRS III.Fig. 2Example of patient-specific 7T MRI connectivity analysis as applied for DBS surgery.A All three orientations (from top to bottom: axial, coronal, sagittal view) are used for STN delineation (in pink), including the thin dorsolateral tip (at crosshairs). B 7T DWI is added to visualize the STN subdivisions motor (blue), associative (green), and limbic (red), based on probabilistic tractography. C The location of the STN motor subdivision is used as target for lead placement, thereby avoiding the associative and limbic subdivisions. D Converting to DICOM was needed for importing the scan into the DBS planning Brainlab software. Electrode lead targeting and trajectory planning were aimed at the hypointense area (black voxels) inside the STN, which represented the motor subdivision. (Red circle: red nucleus; white dot: lead target). MRI magnetic resonance imaging, DWI diffusion-weighted imaging, DBS deep brain stimulation, STN subthalamic nucleus.Table 3Baseline characteristics of connectivity-guided group with both electrode leads inside the STN motor subdivision versus control group7T T2 group (N = 118)Connectivity analysis group (N = 85)p-valueAge (±SD)64.6 ± 8.164.9 ± 7.90.782Sex (M, %)75 (64%)48 (56%)0.308Duration of illness (years ± SD)9.7 ± 4.811.4 ± 5.30.021aDays to follow-up (±SD)233.6 ± 72.1234.2 ± 56.60.950Pre DBS MDS-UPDRS III improvement (% ± SD)55.5 ± 17.7 (N = 117)b59.5 ± 16.9 (N = 84)b0.107Pre DBS MDS-UPDRS III absolute score (±SD)45.5 ± 11.745.2 ± 13.80.892Pre DBS MDS-UPDRS I score (±SD)13.0 ± 5.7 (N = 101)b12.4 ± 5.1 (N = 78)b0.462Pre DBS MDS-UPDRS II score (±SD)15.9 ± 6.7 (N = 88)b16.9 ± 6.8 (N = 78)b0.329Pre DBS MDS-UPDRS IV score (±SD)10.5 ± 3.4 (N = 105)b10.6 ± 3.4 (N = 82)b0.817Pre DBS PDQ39 score (±SD)32.1 ± 11.8 (N = 95)b31.2 ± 11.5 (N = 78)b0.623Pre DBS SAS score (±SD)6.8 ± 5.8 (N = 104)b6.3 ± 4.5 (N = 80)b0.517Pre DBS BDI score (±SD)12.2 ± 4.2 (N = 39)b11.1 ± 6.1 (N = 76)b0.323Pre DBS LEDD score (±SD)1399.5 ± 670.31283.5 ± 605.40.207Average applied amount of current (±SD)2.2 ± 0.72.1 ± 0.60.210SD standard deviation, DBS deep brain stimulation, UPDRS Unified Parkinson’s Disease Rating Scale, PDQ Parkinson’s Disease Questionnaire, SAS Starkstein Apathy Scale, BDI Beck depression inventory, LEDD levodopa equivalent daily dose.aSignificantly different. A p-value of <0.05 is considered statistically significant.bNot all scores were available for all patients.Table 4Primary and secondary outcome measures of connectivity-guided group with both electrode leads inside the STN motor subdivision versus control group7T T2 group (N = 118)Connectivity analysis group (N = 85)p-valueMDS-UPDRS III improvement (% ± SD)49.8 ± 19.660.1 ± 11.1<0.001a# of non-responders (<30% UPDRS-III improvement, %)17 (14%)0 (0%)<0.001aMDS-UPDRS III absolute change (±SD)25 (±0)31 (±6)<0.001aMDS-UPDRS I improvement (% ± SD)7.1 ± 64.4 (N = 96)b15.1 ± 51.5 (N = 67)b0.398MDS-UPDRS II improvement (% ± SD)30.0 ± 41.8 (N = 82)b35.4 ± 40.7 (N = 67)b0.430MDS-UPDRS IV improvement (% ± SD)55.9 ± 47.9 (N = 102)b67.9 ± 33.0 (N = 74)b0.064PDQ39 improvement (% ± SD)18.3 ± 43.2 (N = 87)b10.8 ± 54.2 (N = 68)b0.336SAS improvement (absolute score ± SD)−2.2 ± 6.1 (N = 98)b−0.4 ± 4.0 (N = 72)b0.027aSAS > 5 points worsening (%)29 (30) (N = 98)b9 (13) (N = 72)b0.008aBDI improvement (% ± SD)8.1 ± 46.1 (N = 37)b9.0 ± 59.3 (N = 67)b0.937LEDD score improvement (% ± SD)44.0 ± 84.955.0 ± 37.50.268UPDRS Unified Parkinson’s Disease Rating Scale, SD standard deviation, PDQ Parkinson’s Disease Questionnaire, SAS Starkstein Apathy Scale, BDI Beck Depression Inventory, LEDD levodopa equivalent daily dose.aSignificantly different. A p-value of <0.05 is considered statistically significant.bNot all scores were available for all patients. Example of patient-specific 7T MRI connectivity analysis as applied for DBS surgery. A All three orientations (from top to bottom: axial, coronal, sagittal view) are used for STN delineation (in pink), including the thin dorsolateral tip (at crosshairs). B 7T DWI is added to visualize the STN subdivisions motor (blue), associative (green), and limbic (red), based on probabilistic tractography. C The location of the STN motor subdivision is used as target for lead placement, thereby avoiding the associative and limbic subdivisions. D Converting to DICOM was needed for importing the scan into the DBS planning Brainlab software. Electrode lead targeting and trajectory planning were aimed at the hypointense area (black voxels) inside the STN, which represented the motor subdivision. (Red circle: red nucleus; white dot: lead target). MRI magnetic resonance imaging, DWI diffusion-weighted imaging, DBS deep brain stimulation, STN subthalamic nucleus. Table 3Baseline characteristics of connectivity-guided group with both electrode leads inside the STN motor subdivision versus control group7T T2 group (N = 118)Connectivity analysis group (N = 85)p-valueAge (±SD)64.6 ± 8.164.9 ± 7.90.782Sex (M, %)75 (64%)48 (56%)0.308Duration of illness (years ± SD)9.7 ± 4.811.4 ± 5.30.021aDays to follow-up (±SD)233.6 ± 72.1234.2 ± 56.60.950Pre DBS MDS-UPDRS III improvement (% ± SD)55.5 ± 17.7 (N = 117)b59.5 ± 16.9 (N = 84)b0.107Pre DBS MDS-UPDRS III absolute score (±SD)45.5 ± 11.745.2 ± 13.80.892Pre DBS MDS-UPDRS I score (±SD)13.0 ± 5.7 (N = 101)b12.4 ± 5.1 (N = 78)b0.462Pre DBS MDS-UPDRS II score (±SD)15.9 ± 6.7 (N = 88)b16.9 ± 6.8 (N = 78)b0.329Pre DBS MDS-UPDRS IV score (±SD)10.5 ± 3.4 (N = 105)b10.6 ± 3.4 (N = 82)b0.817Pre DBS PDQ39 score (±SD)32.1 ± 11.8 (N = 95)b31.2 ± 11.5 (N = 78)b0.623Pre DBS SAS score (±SD)6.8 ± 5.8 (N = 104)b6.3 ± 4.5 (N = 80)b0.517Pre DBS BDI score (±SD)12.2 ± 4.2 (N = 39)b11.1 ± 6.1 (N = 76)b0.323Pre DBS LEDD score (±SD)1399.5 ± 670.31283.5 ± 605.40.207Average applied amount of current (±SD)2.2 ± 0.72.1 ± 0.60.210SD standard deviation, DBS deep brain stimulation, UPDRS Unified Parkinson’s Disease Rating Scale, PDQ Parkinson’s Disease Questionnaire, SAS Starkstein Apathy Scale, BDI Beck depression inventory, LEDD levodopa equivalent daily dose.aSignificantly different. A p-value of <0.05 is considered statistically significant.bNot all scores were available for all patients. Baseline characteristics of connectivity-guided group with both electrode leads inside the STN motor subdivision versus control group SD standard deviation, DBS deep brain stimulation, UPDRS Unified Parkinson’s Disease Rating Scale, PDQ Parkinson’s Disease Questionnaire, SAS Starkstein Apathy Scale, BDI Beck depression inventory, LEDD levodopa equivalent daily dose. aSignificantly different. A p-value of <0.05 is considered statistically significant. bNot all scores were available for all patients. Table 4Primary and secondary outcome measures of connectivity-guided group with both electrode leads inside the STN motor subdivision versus control group7T T2 group (N = 118)Connectivity analysis group (N = 85)p-valueMDS-UPDRS III improvement (% ± SD)49.8 ± 19.660.1 ± 11.1<0.001a# of non-responders (<30% UPDRS-III improvement, %)17 (14%)0 (0%)<0.001aMDS-UPDRS III absolute change (±SD)25 (±0)31 (±6)<0.001aMDS-UPDRS I improvement (% ± SD)7.1 ± 64.4 (N = 96)b15.1 ± 51.5 (N = 67)b0.398MDS-UPDRS II improvement (% ± SD)30.0 ± 41.8 (N = 82)b35.4 ± 40.7 (N = 67)b0.430MDS-UPDRS IV improvement (% ± SD)55.9 ± 47.9 (N = 102)b67.9 ± 33.0 (N = 74)b0.064PDQ39 improvement (% ± SD)18.3 ± 43.2 (N = 87)b10.8 ± 54.2 (N = 68)b0.336SAS improvement (absolute score ± SD)−2.2 ± 6.1 (N = 98)b−0.4 ± 4.0 (N = 72)b0.027aSAS > 5 points worsening (%)29 (30) (N = 98)b9 (13) (N = 72)b0.008aBDI improvement (% ± SD)8.1 ± 46.1 (N = 37)b9.0 ± 59.3 (N = 67)b0.937LEDD score improvement (% ± SD)44.0 ± 84.955.0 ± 37.50.268UPDRS Unified Parkinson’s Disease Rating Scale, SD standard deviation, PDQ Parkinson’s Disease Questionnaire, SAS Starkstein Apathy Scale, BDI Beck Depression Inventory, LEDD levodopa equivalent daily dose.aSignificantly different. A p-value of <0.05 is considered statistically significant.bNot all scores were available for all patients. Primary and secondary outcome measures of connectivity-guided group with both electrode leads inside the STN motor subdivision versus control group UPDRS Unified Parkinson’s Disease Rating Scale, SD standard deviation, PDQ Parkinson’s Disease Questionnaire, SAS Starkstein Apathy Scale, BDI Beck Depression Inventory, LEDD levodopa equivalent daily dose. aSignificantly different. A p-value of <0.05 is considered statistically significant. bNot all scores were available for all patients. In 17 patients, one or both electrode leads (twenty electrode leads, 10% of total electrode leads) were located adjacent to the motor subdivision, and connectivity analysis showed electrode leads were surrounded by a lower density of motor connectivity compared to the group with both electrode leads inside the motor subdivision (37 ± 23%, p < 0.001). The four poor responders from the total (=102) 7T MRI connectivity group were found in this subgroup. Average improvement on the MDS-UPDRS III was 33 ± 12% for this group. Connectivity versus T2-weighted targeting On average, the STN consisted of 31 ± 7 voxels with an average size of 100 ± 22 mm3. No visual differences were seen in connectivity between left- and right-sided STNs: both sides showed connectivity to motor, associative, and limbic areas. After connectivity thresholding (see Methods), the motor subdivision consisted (on average) of six voxels, and this area was located in the dorsolateral part of the STN in all patients. The associative and limbic subdivisions were located around the ventral part of the STN. However, individual differences in orientation and distribution of the regions were seen both between and within (left versus right STN) patients and affected the location of the target compared to T2-weighted classical anatomical targeting (Fig. 3).Fig. 3Patient-specific variability in location of the STN motor subdivision.The STN motor subdivision (blue) location varied between and within patients relative to the T2 classical anatomical targeting. A In this patient, the motor subdivision with leads inside is located at the anterior border of the red nucleus. The T2 MRI classical DBS target is at a comparable location. B In this patient, the STN motor subdivision is located far posterior to the anterior border of the red nucleus (classical T2 DBS location) and differs in location between left and right. The T2 MRI classical DBS target is located anterior and outside the motor subdivision. These examples illustrate that although on average the motor subdivision was located more lateral and posterior compared to classical T2 MRI DBS, the exact location varied per patient, stressing the advantage of patient-specific tractography for lead placement. (Red circle: red nucleus; white dot: lead target; white dotted line: lead trajectory; red STN subdivision: limbic area. Per patient, top panels: axial view; bottom panels: coronal view). STN subthalamic nucleus, DBS deep brain stimulation. Patient-specific variability in location of the STN motor subdivision. The STN motor subdivision (blue) location varied between and within patients relative to the T2 classical anatomical targeting. A In this patient, the motor subdivision with leads inside is located at the anterior border of the red nucleus. The T2 MRI classical DBS target is at a comparable location. B In this patient, the STN motor subdivision is located far posterior to the anterior border of the red nucleus (classical T2 DBS location) and differs in location between left and right. The T2 MRI classical DBS target is located anterior and outside the motor subdivision. These examples illustrate that although on average the motor subdivision was located more lateral and posterior compared to classical T2 MRI DBS, the exact location varied per patient, stressing the advantage of patient-specific tractography for lead placement. (Red circle: red nucleus; white dot: lead target; white dotted line: lead trajectory; red STN subdivision: limbic area. Per patient, top panels: axial view; bottom panels: coronal view). STN subthalamic nucleus, DBS deep brain stimulation. On average, the centroid of the active electrode contact level at follow-up was located slightly more lateral and posterior position in the STN compared to anatomical targeting based on the 7T T2. For the 7T MRI connectivity analysis group, the coordinates were 12.9 mm lateral (range 10.03–19.16), 2.4 mm posterior (range −0.56 to 5.47), and 2.6 mm inferior (range 0.07–6.71), relative to the midcommissural point. For the 7T MRI T2-weighted classical targeting group, the coordinates were 12.4 mm lateral (range 9.22–16.51), 1.5 mm posterior (range −1.06 to 5.63), and 3.1 mm inferior (range 0.01–7.41). The average lateral, posterior and inferior coordinates of the centroid of the active electrode contact level were statistically different between the two groups (p < 0.001 for lateral, posterior and inferior coordinates). Furthermore, in 44% of the cases in the 7T MRI connectivity analysis group, the active electrode contact level was located posterior to the anterior border of the red nucleus. The coordinates of the active electrode contact point for both groups were plotted in reference to the MCP to visualize their spatial relationship, see Fig. 4.Fig. 4The spatial relation of active electrode contact levels of the two groups compared.The coordinates of the active electrode contact levels of the 7T MRI connectivity analysis targeted group (blue dots) are shown in comparison to the coordinates of the active electrode contact levels of the 7T MRI T2-weighted classical targeting group (green crosses), in reference to the midcommissural point (at 0, 0 in the axes). Top plot) medial-lateral versus anterior-posterior; bottom plot) medial-lateral versus superior-inferior. The larger black dot and larger black cross illustrate the mean of the groups. The spatial relation of active electrode contact levels of the two groups compared. The coordinates of the active electrode contact levels of the 7T MRI connectivity analysis targeted group (blue dots) are shown in comparison to the coordinates of the active electrode contact levels of the 7T MRI T2-weighted classical targeting group (green crosses), in reference to the midcommissural point (at 0, 0 in the axes). Top plot) medial-lateral versus anterior-posterior; bottom plot) medial-lateral versus superior-inferior. The larger black dot and larger black cross illustrate the mean of the groups. Discussion In this prospective study, we use 7T MRI probabilistic tractography for DBS surgery in PD to target lead placement at the subthalamic motor network, introducing patient-specific connectivity analysis-guided DBS. This readily implementable brain network model provided an individual biomarker, the connectivity-derived motor subdivision, for implantation location that increased the effectiveness of DBS in patients with PD. The connectivity analysis visualized direct STN connections with cortical motor areas, expected to represent the hyperdirect pathway. For this, the 7T MRI T2-weighted image enabled clear visualization of distinct anatomical size and orientation, including the thin dorsolateral top, ensuring all STN voxels were included. The 7T MRI DWI, combined with probabilistic methods, was subsequently used to quantify individual connectivity levels in such a way that all voxels of the motor subdivision were visualized. The resulting connectivity-derived motor subdivision of the STN was applied as a patient-specific target for electrode lead placement in 102 DBS surgeries for PD. Compared to a control group in which only 7T MRI T2-weighted imaging (classical anatomical landmarks) for STN-DBS targeting was used, significantly more improvement in motor functioning was observed, and significantly fewer poor responders to DBS were seen. In a subgroup of patients, none of the electrode lead contact levels were placed inside the intended motor subdivision of the STN. Quantitative analysis demonstrated a lower density of motor connections surrounding these electrode contact levels. However, for the majority of patients, both electrode leads were successfully placed inside the connectivity-derived motor subdivision of the STN. This group demonstrated the most improvement in motor functioning compared to the control group, 60% versus 50%. Furthermore, no poor responders to DBS were seen in this group. Additionally, a trend of more improvement in motor fluctuations (MDS-UPDRS IV) was also seen. This prospective comparison suggests that DBS targeting using the 7T MRI connectivity-derived motor subdivision of the STN leads to more motor improvement and can function as a patient-specific biomarker. Since the introduction of MRI for STN visualization, motor improvement after DBS remained fairly stable (averaging 48%), and is highly variable across individual patients17–22. Approximately one-fifth of patients experience insufficient benefit, with improvements in motor functioning below 30%. Poor response was considered an inherent limitation of DBS therapy23–26. The inability to reduce this significance proportion was also previously reported by our group, and again confirmed in the current study in the control group23. By contrast, when applying connectivity analysis for DBS, in 96% of patients, a response of at least 30% in motor improvement was noted, and this increased further when analyzing the DBS outcome of only electrode leads located in the motor subdivision of the STN. Demonstrating less individual variation in DBS response could be considered even more important than increasing average (group) improvement, as the former ensures that more patients benefit from DBS. The comparison of the two groups was not randomized, but instead a matched control group was used to, as much as possible, prevent selection bias and to strive for balance with respect to known and unknown confounding or prognostic variables. All baseline measurements were comparable, except duration of illness due to the surgery delaying effect of the coronavirus pandemic. Our large prospective cohort comparison has the advantage of being representative of the entire source population, as all patients undergoing DBS for PD in our center underwent a 7T MRI. During the inclusion period, no changes were introduced for DBS selection criteria, DBS surgery, DBS programming, and 7T MRI parameters, except the implementation of 7T probabilistic tractography from April 2022 onwards. We did not choose to conduct a randomized design when introducing a new technique, as it was not sensible to conduct a randomized study before the method had been carefully studied in a prospective patient cohort. The current study describes the new technique and also demonstrates improved clinical outcomes in a large patient cohort when compared to a similarly large recent control group. Lead localization was only performed on intraoperative CT imaging, no delayed CT-scans were acquired. This may have resulted in not capturing potential lead migration in some patients27. Furthermore, for connectivity profiles, the Brodmann areas were visually identified and marked. Cortical targets were defined using readily applicable anatomical landmarks on the 7T T2-weighted MRI rather than a formal cytoarchitectonic parcellation scheme. While this approach was chosen to ensure complete inclusion of all visible STN connectivity projections and to remain applicable in routine clinical practice, it introduces inter-operator variability in cortical boundary delineation that more modern atlas-based parcellation systems, such as the Jülich probabilistic cytoarchitectonic atlas, could in the future reduce28. Prospective studies using automated cortical parcellation may therefore allow more standardized and reproducible definition of STN connectivity subdivisions. All patients who were operated on during the time-window of the study were included for this analysis. Potential bias between the two groups in terms of disease progression is thereby limited, also reflected by the non-significant difference in demographic data between the two groups. Potential bias may have occurred in patients who were unable to undergo DWI imaging; baseline MDS-UPDRS III scores were comparable between the two groups. Compared to targeting based solely on 7T MRI T2-weighted imaging, adaptations in the lateral and posterior direction were made in a substantial number of the patients to achieve placement in the STN motor subdivision, reflected by a statistically different anatomical location of the active electrode contact level at follow-up. Compared to 7T MRI T2-weighted imaging based on classical landmarks (i.e., the anterior border of the red nucleus), in 44% of the cases, the connectivity-derived target was located posterior to the anterior border of the red nucleus. This suggests that in 44% of the cases, classical T2-weighted targeting would have missed the tractography-derived motor subdivision of the STN. Despite these adaptations, for a select group of patients, the electrode leads were located adjacent to the motor subdivision. The motor subdivision was consistently located in the dorsolateral part of the STN, and all three orthogonal planes (axial, coronal, and sagittal) were used to determine target and trajectory angles. The dorsolateral tip was most accurately identified on the coronal and sagittal planes. This was different compared to classical T2-weighted landmarks, as these are mainly based on the axial plane. Taking these new requirements into account for directing placement at the motor subdivision, only one trajectory will fit, limiting the possibilities of adaptations when an intraoperative trajectory deviation is noted. Important to note in light of the results of this study, normative anatomical studies have yielded conflicting results regarding the concept of multiple symptom-specific sweet spots in the STN for tremor and akinesia control5,29–31. Our study does not provide additional insight into this question, as we consider the spatial precision of current techniques insufficient to discriminate and selectively target symptom-specific subregions within the motor subdivision of the STN. With increased imaging resolution and more focal stimulation, it would be interesting to study whether it is possible to identify and selectively target subregions within the motor subdivision of the STN for specific symptoms. The use of 7T MRI is also a limitation of this study. The limited availability of 7T MRI worldwide creates a challenge in transforming this individual connectivity analysis into a widely available method based on 3T technology. Connectivity analysis indicated that a number of selected active contact levels during follow-up assessment were not (ideally) located in the STN motor subdivision, both (1) for patients with both electrode leads located in the STN motor subdivision, as (2) for patients with one or two electrode leads outside the STN motor subdivision. Switching the contact level could perhaps increase the DBS effect. These patients are currently offered programming based on these findings to direct their VTA at the motor subdivision. This suggests that connectivity analysis can have a direct implication for DBS programming by accurately predicting the VTA to include the entire connectivity-derived motor subdivision of the STN, instead of the traditional (clinical guided) monopolar review14. With network analysis, associative and limbic networks are also visualized by default. This instantly shows the coherence between programming and (preventing) neuropsychological and psychiatric changes. Programming outside of the motor network more often results in these undesirable effects14. Indeed, the 7T MRI connectivity analysis group in this study showed significantly less worsening of apathy. Although increased mobility after DBS is associated with experiencing a better QoL, reports suggest that other factors, such as patient expectations, disabling PD symptoms that are not improved by DBS, and occurrence of adverse effects, determine this as well, possibly explaining the equal improvement in QoL for both groups in this study14,32–34. The probabilistically derived motor subdivision of the STN can be used as a powerful individual biomarker in DBS for PD. In neurobiological models of the nucleus, the motor subdivision is considered to be the optimal area inside the nucleus for neuromodulation7. Studies showed that this subdivision is connected directly to cortical motor areas (M1/SMA) via the hyperdirect pathway, giving it a possible modulatory advantage compared to other brain nuclei used as targets in DBS for PD6. Until now, direct patient-specific visualization and quantification for lead placement was not possible, and therefore DBS implantation has been done using awake testing, MER, local field potentials, T2-weighted MR-based landmarks, anatomical group level hotspots/sweetspots, and normative connectomes26,35. The patient group has not been studied using normative-sample-based connectivity analysis, and hence, the results cannot substantiate that individual connectivity analysis leads to better target determination than normative-sample-based connectivity analysis. Therefore, our postulated advantage of individual connectivity analysis remains a hypothesis that needs to be evaluated in a direct comparison. The presented probabilistic 7T MRI tractography method provides a usable patient-specific model in subthalamic motor subdivision visualization for DBS targeting of STN-DBS in PD. Methods Study design This was a prospective single-center study carried out at the Amsterdam University Medical Center, the Netherlands. The primary outcome was the % change of the MDS-UPDRS-III between OFF medication and follow-up in the OFF-medication-ON-DBS condition. This was compared between a 7T MRI connectivity analysis-based targeting group and a matched control group undergoing DBS with targeting based on 7T MRI T2-weighted (classical) anatomical landmarks only. The principal investigators designed the study, had full access to the data, and vouched for the completeness and accuracy of the data, the reporting of adverse events, and the fidelity of the study to the protocol. The study was performed in accordance with the Declaration of Helsinki. The study protocol was approved by our local institutional Ethics Committee (W20_540 # 20.599). Patients were included until a balanced group size compared to the historical cohort was reached. All patients were informed about the 7T MRI, and written informed consent was provided (ClinicalTrials.gov, NCT05843084). Patient selection Patients who underwent DBS surgery were included in the comparison if they had idiopathic PD with bradykinesia, tremor, and/or rigidity, and despite optimal pharmacological treatment, at least one of the following symptoms: severe motor response fluctuations, dyskinesias, or painful dystonia. Boston Scientific (Versice CartesiaTM Directional leads, Boston Scientific, Valencia, USA) or Medtronic (Sensight leads, Minneapolis, USA) leads were implanted bilaterally in the STN. 7T MRI acquisition in PD patients In all patients, both the connectivity analysis-based targeting group and a matched control group, the 7T MR-scanning procedure was performed on a 7T Achieva system (Philips Healthcare) without sedation. The magnetic field strength of the 7T MRI provided an increased signal-to-noise ratio parallelled by an improvement in image resolution compared to the typical 3T MRI used in the clinical setting. Other than the known non-compatible implants for MRI, we rarely encountered specific non-compatible implants for the high field. Before scanning, all patients were informed about the louder noises compared to lower field-strength scanners, and they were aware that they might experience nausea or vertigo temporarily during their stay in the high-magnetic field environment. 7T MRI sequences Two types of sequences were conducted: (1) a 3-dimensional (3D) sagittal T2-weighted image with Turbo spin echo (TSE) imaging, repetition time (TR) = 3000 milliseconds (ms), echo time (TE) = 324 ms, Echo Train Length (ETL) = 182, field of view (FOV) = 250 × 250 x 190, Flip angle = 100, voxel size = 0.7 × 0.7 × 0.7 millimeter (mm), (2) a diffusion weighted imaging (DWI), TR = 6084 ms, TE = 70 ms, ETL = 59, b = 1000 s/mm2, 32 directions, repeated with anteroposterior and posteroanterior phase encoding, one b0 reference, FOV = 140 × 177 × 110, voxel size = 1.4 × 1.4 × 1.5 mm, rf-power was adjusted so as to obtain sufficient signal intensity in both the STN area and cortex. MRI scanning was done using a 32-channel receive coil (Nova Medical, Wilmington, MA, USA). The sequences were not adapted or changed during the time-period. All patients (in both groups) underwent the same structural 3D sagittal T2-weighted sequence, and all patients in the connectivity-based targeted group underwent the same DWI sequence. 7T MRI T2-weighted (classical) anatomical landmarks based on DBS targeting For 7T MRI T2-weighted classical targeting, STN target planning was started on axial images using standard stereotactic coordinates relative to the mid-commissural point: 11 mm lateral, 2 mm posterior, and 4 mm ventral. Target planning was subsequently adapted to individual anatomical configuration at the level of the maximum diameter of the red nucleus on axial images, using the Bejjani line (extending laterally from the anterior borders of the red nucleus) to determine the anterior-posterior position of the target, central within the visible hypointense STN. On the coronal plane, lateral angle adjustments could then be made for entering the most dorsolateral part of the STN, thereby lengthening the trajectory within the STN36. 7T MRI connectivity analysis-based DBS targeting For 7T MRI T2 and diffusion-weighted imaging (probabilistic tractography), STN target planning was aimed at the subthalamic motor subdivision. Probabilistic patient-specific connectivity was used to visualize and identify the location of the motor subdivision, which enabled direct electrode lead placement during target planning in this area inside the STN. Processing of 7T DWI imaging The acquired Digital Imaging and Communications in Medicine (DICOM) files were converted to Neuroimaging Informatics Technology Initiative (NifTI) volumes using MRIcroGL’s dcm2niix converter. Subsequently, image denoising, to enhance the data quality, and Gibbs artefact removal were applied using the MRtrix3 software package. Afterwards, FSL was used for several processing steps. In order to correct for distortions during preprocessing, b0-reference volumes with opposite phase encoding polarity were acquired. Top-up was used to correct for susceptibility-induced distortions (geometric mismatch between the structural and the diffusion image). Eddy was used for correcting eddy currents and movement in the diffusion data, and DTIFIT was used for fitting a diffusion tensor model in each voxel. Subsequently, BedpostX was run to fit the probabilistic diffusion model on the corrected data. The different preprocessing techniques were linked in a Matlab (Matlab 2019b; Mathworks, Natick, MA) pipeline. Connectivity analysis Probabilistic tractography for visualizing the widespread connections of the STN started with manual delineating, i.e., masking, of the nucleus in native space using the ITKsnap tool (ITK-SNAP V3.8 software). The T2-weighted image was co-registered and superimposed (in ITK) to the B0. This enabled the creation of high-resolution 3-dimensional T2-weighted imaging masks directly in the diffusion-weighted space. Voxel selection inclusion criteria required a minimum of 50% STN (hypo-)intensity. Masking of all patients was done by one DBS neurosurgeon (MB). The T2-weighted MRI was registered to the B0 by using FSL’s FLIRT tool with the following parameters: 12 degrees of freedom, mutual information cost function, tri-linear interpolation, and normal search. The STN masks were applied as a seed in FSL, and the resulting cortical connectivity was used for guiding cortical masking in native space. Visualizing masks and connectivity enabled evaluation if these made sense from an anatomical point of view. Three major STN connectivity projections to the cortex were identified and masked by masking three discrete cortical targets: connectivity to the primary and supplementary motor cortex (motor subdivision, Brodmann Area (BA) 4), connectivity to the dorsolateral prefrontal cortex (associative subdivision, BA 9, 10, and 46) and connectivity to the orbitofrontal/basofrontal cortex (limbic subdivision, BA 11 and 47). Visualizing the three subdivisions resulted from running ProbtrackX (default parameters), using ipsilateral cortical masks as classification targets. Cortical areas were identified by visual inspection of the 7T T2-weighted image registered to the B0 diffusion space in native space, using readily applicable sulcal, gyral, and structural landmarks without automated parcellation. The approach was designed to capture all visible cortical connectivity projections from the STN while remaining practical for routine clinical use. On axial images, the primary motor cortex (M1, BA4) was identified using the characteristic omega-shaped hand knob on the posterior bank of the precentral gyrus as the primary landmark. BA4 then served as a reference point: the two gyri immediately anterior to the precentral gyrus were considered to represent the premotor cortex and supplementary motor area (BA6). Together, BA4 and BA6 were masked as the motor cortical target. On coronal images, cortical regions anterior to the motor areas and superior to the genu of the corpus callosum were classified as prefrontal (dorsolateral prefrontal cortex, BA9, 10, and 46), comprising the cortex along the superior and middle frontal gyri. The temporal cortex was explicitly excluded by restricting the mask to the cortex superior to the Sylvian fissure. The orbitofrontal/basofrontal cortex (BA11 and 47) was defined on coronal images as the cortex on the inferior surface of the frontal lobe below the level of the genu of the corpus callosum, comprising the orbital and rectal gyri medially and the pars orbitalis of the inferior frontal gyrus laterally. The temporal cortex was again excluded posteriorly. Seed voxels of the STN were classified according to the probability of cortical connectivity, creating areas within the STN with high or low connectivity to any of the three cortical areas. We considered an STN-voxel to be connected to a cortical area if at least 20% of its probabilistic tracks connected the two, allowing for the possibility of a voxel being connected to multiple cortical areas. As a voxel could contain connectivity to more than one cortical area, we favored this threshold approach over a winner-takes-all strategy. The three subdivisions were visualized to confirm location and orientation, and the voxels of the motor subdivision for the left and right STN were masked. Selection of the voxels was based on (1) creating a suitable target for electrode lead placement (adjacent voxels), (2) the level of individual voxel connectivity as calculated by FSL (higher connectivity was given preference), and (3) overlap of the subdivisions (voxels solely connected to the motor areas were given preference). Applying these selection criteria could result in a partial selection of motor subdivision voxels for masking. The motor subdivision masks were “punched out” in the T2-weighted (B0) MRI, and this (diffusion resolution) scan was subsequently registered to the T2-weighted MRI using FLIRT, increasing resolution. The result showed a T2-weighted MRI in which the motor subdivision voxels were “punched out”, i.e., colored black. This NifTI scan was then converted back to DICOM using the VTK DICOM tool (https://github.com/dgobbi/vtk-dicom). Converting to DICOM was needed in order to import the scan into the DBS planning Brainlab software (Brainlab AG, Munich, Germany). Electrode lead targeting and trajectory planning were aimed at the hypointense area (black voxels) inside the STN, which represented the motor subdivision. During surgery, the Brainlab software was used for assessing electrode lead placement. A difference of less than 1 mm in electrode lead location on the intraoperative CT and intended target was considered as trajectories coinciding (fully overlapping). A difference of 1 to 2 mm was closely checked for being inside the STN motor subdivision and distance relative to surrounding anatomical structures, such as the red nucleus and the (tractography-generated) corticospinal tract. Postoperative connectivity analysis After surgery, we used the preoperative 7T MRI connectivity analysis in order to visualize for each active electrode contact level in which subdivision of the STN it was located (motor, associative, or limbic). These steps were performed in the FSL (FMRIB) software by merging the MRI with postoperative CT using FLIRT with the following parameters: six degrees of freedom, mutual information cost function, tri-linear interpolation, and normal search. In FSLeyes, the T2-weighted (B0) MRI with the motor subdivisions punched out, STN subdivisions, T2-weighted MRI, and postoperative CT were all superimposed and aligned to the commissural line (using the nudge function), enabling assessment of each individual electrode contact level (four per electrode lead) for its location relative to the three subregions. Furthermore, relative volumes of the three subregions around each active electrode contact level were evaluated for each STN using FSL-stats. Surrounding connectivity was determined by examining the voxels directly adjacent to the active contact (voxel size 1.4 × 1.4 × 1.6 mm, length of electrode contact level 1.5 mm). This approach resulted in a group of nine voxels per active electrode contact, in which the active electrode contact level represented the central voxel within a nine-voxel square. In this way, all probabilistic tractography within a distance of 1.4 mm was considered, aligning with the voxel size of the diffusion-weighted MRI. This approach was chosen based on (finite element) models, which suggest that DBS almost solely affects the projection fibers near (at the same level as) the active contact point37. Furthermore, high-resolution electron microscopy studies in primates have revealed that a significant portion of the sensorimotor and associative STN surface area consists of myelinated axons with diameters predominantly less than 1 μm. These small diameters require electric field magnitudes exceeding 1.8 Volt per millimeter (V/mm) for activation, which are typically only achieved in the vicinity of the electrode contact levels with conventional stimulation settings38. In the analysis, all voxels around and including the active electrode contact level exhibiting probabilistic tracts to the three subdivisions of the STN were selected and averaged. Voxels without connectivity were not included. The averaged connectivity was divided by the maximal connectivity (the highest value) for that patient, with connectivity values as calculated by FSL. This enabled us to quantify the connectivity for each network as a percentage (0–100%). An electrode lead was considered as located inside the motor subdivision when the center of at least one electrode contact level was located inside this subdivision. General surgical procedure For both groups, the surgical procedure was performed under general anesthesia17. Single-track MER for intraoperative electrophysiological conformation of the dorsal and ventral borders was performed17. By combining MRI and a Cartesian coordinate system, electrode leads were bilaterally placed into the STN. From a central localized starting point in the brain, the midpoint of the commissural line, the anatomical optimal target in the STN was determined by moving in all three orthogonal planes of the coordinate system (medial-lateral, anterior-posterior, and inferior-superior). Electrode leads had four contact levels, with two segmented contacts in the middle (length 1.5 mm, interspace 0.5 mm). Electrode lead placement was aimed at placing the two middle electrode contact levels inside the motor subdivision of the STN, taking into consideration a safe distance (at least 3 mm) from the internal capsule, avoiding the tractography-visualized corticospinal tract and the white matter medial to the STN. Correct electrode positioning was directly assessed during surgery by using intraoperative CT. Extension wires running subcutaneously through the neck to the infraclavicular location internal pulse generator was implanted during the same surgery. After surgery, patients stayed for a short time in the postoperative care unit and were discharged the next day. Three neurosurgeons were involved in DBS lead implantations as primary surgeons and performed all surgeries in both groups during the entire study period. Besides the connectivity-based targeting technique, no other factors in the surgical procedure, including pre-operative work-up and surgical techniques, were adapted or changed throughout the study period. DBS programming For both groups, postoperative DBS programming was performed by means of standardized monopolar review (MPR). The selection of the optimal stimulation parameters (contact level and pulse settings) was achieved by MPR three to four weeks after DBS surgery. All electrode contact levels were independently tested for effect and side-effect (efficacy threshold) for left and right separately, and the most favorable contact level was chosen. All patients were stimulated using monopolar omnidirectional stimulation at one contact level on each electrode lead. The specialized nurses who performed the MPR were not aware of the location of the electrode leads relative to the STN as visualized with connectivity analysis or T2-weighted MRI. The protocol to perform MPR was kept the same for both groups and not adapted or changed within the timeframe. Patients who were included in ongoing studies for which one of the factors was changed were excluded from this analysis. Clinical assessment For both groups, all outcome scores were directly available from our electronic online DBS database (Castor EDC, Amsterdam). Patients underwent baseline and six-month follow-up evaluation as part of standard clinical care. The primary outcome was the change score in motor symptoms induced by DBS as measured by the MDS-UPDRS III after six months of DBS therapy in the OFF-medication-ON-DBS state. For calculating change scores, the highest OFF-medication score was used, either taking the baseline or six months follow-up measurement (pre-operative OFF-medication, or follow-up OFF-medication-OFF-DBS condition). Scores ranged from 0 to 132, with higher scores indicating worse function. Time constraints for clinical assessment occasionally resulted in a suboptimal OFF-medication state assessment, and this was noticed by a difference in the two OFF-medication scores, not explained by disease progression, but by other factors such as half-time of medication taken or better/worse sleep during hospital stay compared to at home. For ensuring a representative change score, the formula (([highest MDS-UPDRS III OFF-medication] – [MDS-UPDRS III OFF-medication-ON-DBS])) / (highest MDS-UPDRS III OFF-medication)*100% was used. An improvement of at least 30% on the MDS-UPDRS III score was used as criterium for clinical response26. Secondary outcome measures included the MDS-UPDRS non-motor experiences of daily living (MDS-UPDRS part I, score range from 0 to 52), MDS-UPDRS motor experiences of daily living (MDS-UPDRS part II, score range from 0-52), MDS-UPDRS motor complications (MDS-UPDRS part IV, range from 0-24, a higher score indicating worse function), QoL measured using the PDQ-39 (scores range from 0 to 100, a higher score indicating lower QoL), SAS (scores range from 0 to 42, a higher score indicating more apathy), and BDI (score range from 0 to 63, a higher score indicating more depression). The workflow and team structure of specialized DBS nurses for clinical assessments did not change during the entire study period, and all specialized DBS nurses were certified to conduct the above described rating scales. Statistical analysis The primary analysis was the between-group comparison of the mean percentage change in MDS-UPDRS III score, measured using an independent samples t-test. A p-value of less than 0.05 was considered statistically significant. All secondary outcomes and demographic data were also tested using an independent samples t-test. For sex, a Chi-squared test was performed. For statistical support and analyses, our statistical department was involved.