The study was approved by the Ethics committee Research UZ/KU Leuven (S62373) and registered on Clinicaltrials.gov (NCT04658641). All participants received oral and written information and provided oral and written consent. The study was conducted in conformity with the Declaration of Helsinki, the Belgian law of May 7th 2004 on experiments on the human person and in agreement with Good Clinical Practice guidelines.

Participants that met the “UK PD Society Brain Bank Clinical Diagnostic Criteria” for the diagnosis of idiopathic PD, were included in the study (Postuma et al., 2015). Directional leads (Vercise Cartesia^®, Boston Scientific; BSC, Valencia, CA, United States) were bilaterally implanted in the STN and subcutaneously connected to the implantable pulse generator (IPG; Vercise DBS Systems, BSC, Valencia, CA, United States) that has MICC technology designed to allow for refined division of the total current over multiple DBS-contacts (Boston Scientific Corporation, 2018). The DBS-leads consist of eight DBS-contacts with a length of 1.5 mm, separated from one another by interspaces of 0.5 mm and arranged in a tip-3-3-1 configuration (Paff et al., 2020) (distal-to-proximal axis of the electrode contact numbering of left lead: C1-C8; and the right lead: C9-C16, where “C” stands for “Contact”). The surgical procedure was performed as standard-of-care at our center using the microrecording technique (Gross et al., 2006).

Patients that already participated in the previous study (Peeters et al., 2021) were now enrolled in a follow-up study, where we tested the MICC technology (see further). Four patients participated, one of which participated twice, yielding data from both hemispheres in this patient. In total, five hemispheres were tested. All participants were asked to refrain from taking their medication 12 h prior to the study visit. Demographic data and stimulation parameters used during the experiment are summarized in Supplementary Table 1.

First, stimulation was turned off in both hemispheres. One hemisphere was tested at a time, with the other hemisphere remaining off. Thereafter, the stimulation intensity was defined on the clinical contact (monopolar cathodic pulse with return on the case; 60 μs and 130 Hz) as the highest stimulation intensity without non-transient stimulation-induced side effects. For the experimental setup, stimulation was then decreased to 10 Hz. An in vitro phantom head experiment was performed as a negative control where no EPs were expected. The set-up used a head-sized watermelon, where a directional lead (Vercise Cartesia^®, BSC, Valencia, CA, United States) was positioned approximately 6.0 cm from the surface. EEG channels were then positioned on the surface and an anterior-posterior direction was appointed depending on the location of the EEG channels. All processing steps and analyses performed on real patient datasets were repeated for the phantom head dataset.

At the start of the experiment, the electric field (which had an approximately constant volume throughout the experiment as the same stimulation intensity was applied throughout the experiment) was set at the center of the most dorsal DBS-contact (i.e., 100% on C8 for the left hemisphere and C16 for the right hemisphere) for 50 s, yielding a total of 500 epochs of 100-ms duration. Then, the electric field was moved in a ventral direction in fifteen equal steps until the most ventral DBS-contact was tested. Thus, we tested sixteen incremental positions in total. The two most distant electric field locations had a distance of 6.0 mm in total, thus equating each proportional shift in the electric field was about 0.4 mm (6.0 mm/15 steps) per step. The segmented contacts were only tested in ring mode to avoid confounding the results with horizontal steering as a variable.

EEG recordings were performed with a 64-channel ActiveTwo BioSemi system with a sample rate of 16,384 Hz and a built-in low-pass filter with a cut-off frequency of 3,200 Hz. This EEG system uses active recording channels positioned according to the internationally standardized 10–20 system (Jasper, 1958) and referenced to the vertex EEG channel (Cz). One additional EEG channel (EXG1) was positioned on the skin on top of the implanted IPG to record the stimulation pulse, which served as a trigger channel to align EPs. Two additional EEG channels were positioned on the left (EXG2) and right (EXG3) mastoid to record the stimulation pulse at a cranial location with negligible neural responses. We stimulated each of the 16 depths for 50 s at 10 Hz, yielding a total of 500 epochs with a duration of 100 ms for each depth. Each epoch was baseline corrected by subtracting the average of a 1-ms period prior to stimulus onset. Then the epochs were averaged to get the averaged EP. We applied a combination of linear interpolation and template-subtraction to reduce the total stimulation-induced artifact. Template subtraction was based on the artifact recorded with EEG electrodes EXG2 and EXG3. Two bandpass 2nd-order Butterworth filters were applied to these EPs. One was designed for evaluation of short-latency responses with a high-pass cutoff frequency of 150 Hz and low-pass cutoff frequency of 1,000 Hz. The other was designed for evaluation of long-latency responses with a high-pass cutoff frequency of 1 Hz and low-pass cutoff frequency of 150 Hz. A more detailed description of the EEG protocol and artifact-reduction method can be found in Peeters et al. (2021).

All data processing and statistical analyses were done in MATLAB 2021a (Mathworks, Natick, MA, United States). A significance level of 5% was used in all tests. Based on the previous study (Peeters et al., 2021), we recorded a short-latency peak at 3 ms (P3) via the motor cortex EEG channel ipsilateral to stimulation (i.e., F3 for left hemisphere, F4 for right hemisphere) as in this EEG channel the strongest P3 was recorded. For the same reason, we recorded a long-latency peak at 10 ms (P10) via the prefrontal cortex EEG channel ipsilateral to stimulation (i.e., AF7 for left hemisphere, AF8 for right hemisphere). By central limit theorem, the individual EPs recorded conform to Gaussian assumptions so parametric statistics were used (Central Limit Theorem, 2008). Thus, we used one-way ANOVA to evaluate if the MICC depth of stimulation affected the P3 and P10 peak amplitude as measured in each individual hemisphere. Each EP consisted of more than 400 epochs, thus enough data were available to perform robust statistics at the individual hemispheric level. In the previous study, a one-way ANOVA was used to investigate if increasing stimulation intensity significantly affected P3 and P10 amplitude. If no significant effect of intensity was found on the peak amplitude, no further analysis was performed in this hemisphere (see Supplementary Table 1). For the remaining hemispheres, we used one-way ANOVA to evaluate if MICC technology significantly affected EP peak amplitude. Next, to test the separability of MICC on the P3 and P10 peak amplitude between different electric field pairs (varying from one step between two immediately adjacent electric field pairs to fifteen steps between two electric field pairs). For this, a post hoc analysis with a Bonferroni correction for multiple comparisons was applied (MATLAB, multcompare).

To investigate the relationship between the distance from each electric field center to relevant anatomical regions, we grouped all tested hemispheres (analysis on the individual hemisphere level can be found in Supplementary Figure 5). The open-source Lead-DBS image processing pipeline (version 2.5.3, Berlin, Germany) (Horn and Kühn, 2015; Horn et al., 2019) was used for postoperative lead reconstruction analyses, allowing the determination of the specific lead position and orientation on an individual hemispheric level. We then calculated the distance between the center of each electric field and the closest voxel of certain brain regions using the Distal atlas (Ewert et al., 2018).

Figure 1 shows the short- and long-latency EPs in response to DBS when using the MICC technology to vertically migrate the center of the electric field from the most dorsal DBS-contacts in 16 steps to the most ventral contact for a representative subject. Each of the 16 EPs are shown in a different color, as indicated in the legend. Figures 1A,B illustrate the short- and long-latency EPs recorded in participant 1 (left hemisphere), respectively, while Figures 1C,D the short- and long-latency EPs show recorded in a phantom head. All stimulation settings were well tolerated. In general, the EP morphology was similar to previously reported data recorded in a similar patient population (Walker et al., 2012; Miocinovic et al., 2018; Peeters et al., 2021). As expected, the P3 peak appeared strongest in the most dorsal DBS-contacts, while the P10 peak appeared strongest in the most ventral DBS-contacts (Peeters et al., 2021). Based on the analysis of the previous study (Peeters et al., 2021) we found a significant P3 peak in four out of five hemispheres and a significant P10 peak in all five hemispheres (see Supplementary Figures 1, 2). Therefore, further analysis on P3 was only performed in the four hemispheres where a significant P3 peak was detected. A summary of this analysis is provided in Supplementary Table 2.

Figure 2 illustrates the change in EP amplitude for P3 peak (left panels) and P10 peak (right panels) for participant 1 (upper panels) and the phantom head (lower panels). Each of the 16 EPs are shown in a different color (see legend). A one-way ANOVA showed that there was a significant effect of the MICC-controlled electric field depth on P3 amplitude [F_{(15, 6399)} = 36.21; p < 0.0001] for participant 1. Additionally, a significant effect of MICC-controlled electric field depth on P10 amplitude [F_{(15, 6399)} = 395.57; p < 0.0001] was also found in this participant. Importantly, control stimulation in the phantom head showed no effect of MICC-controlled electric field depth on P3 nor P10 amplitude [P3: F_{(15, 6399)} = 0.30; p = 0.956; P10: F_{(15, 6399)} = 1.13; p = 0.3259] In total, we found that the MICC-controlled electric field depth had a significant effect on P3 amplitude in all four tested hemispheres and a significant effect on P10 amplitude in all five tested hemispheres (see Table 1 and Supplementary Figures 3, 4).

Post hoc analysis was performed to investigate the separability of MICC-controlled electric field depth on P3 and P10 amplitudes. Figure 3 shows the electric field pair separation in incremental steps, varying from one step to fifteen steps on the x-axis, and the percentage of electric field pairs showing a significantly different P3 (A) or P10 (B) peak amplitude (mean ± CI) on the y-axis for all tested hemispheres after Bonferroni correction was applied. The P10 peak was significantly different on around 30% of immediately adjacent electric field pairs (1 step separation) increasing to 100% of electric field pairs when the separation was increased to 15 steps (Figure 3B). The P3 peak was only significantly different on around 5% of electric field pairs when the separation was increased to 2 steps. This percentage of significantly different electric field pairs increased steadily to around 80% of pairs as the separation was increased to 15 steps.

The above results strengthen the idea published in a previous article (Peeters et al., 2021), stating that stimulation on the different depths preferentially modulates different nuclei, thereby causing the different EP peaks. We therefore plotted the average P3 and P10 amplitudes from all tested hemispheres as a function of the distance of each of the 16 electric field depths to dorsolateral STN and to SNr, respectively (Figure 4). This indicates indeed that the closer the MICC-controlled depth is to motor STN, the stronger the P3 peak amplitude appears and P10 peak amplitude appears strongest when stimulating from an MICC-controlled depth closest to SN.