Towards guided and automated programming of subthalamic area stimulation in Parkinson's disease

Abstract

Selecting the ideal contact to apply subthalamic nucleus deep brain stimulation in Parkinson's disease can be an arduous process, with outcomes highly dependent on clinician expertise. This study aims to assess whether neuronal signals recorded intraoperatively in awake patients, and the anatomical location of contacts, can assist programming. In a cohort of 14 patients with Parkinson's disease, implanted with subthalamic nucleus deep brain stimulation, the four contacts on each lead in the 28 hemispheres were ranked according to proximity to a nominated ideal anatomical location and power of the following neuronal signals: evoked resonant neural activity, beta oscillations and high-frequency oscillations. We assessed how these rankings predicted, on each lead: (i) the motor benefit from deep brain stimulation applied through each contact and (ii) the 'ideal' contact to apply deep brain stimulation. The ranking of contacts according to each factor predicted motor benefit from subthalamic nucleus deep brain stimulation, as follows: evoked resonant neural activity; r ² = 0.50, Akaike information criterion 1039.9, beta; r ² = 0.50, Akaike information criterion 1041.6, high-frequency oscillations; r ² = 0.44, Akaike information criterion 1057.2 and anatomy; r ² = 0.49, Akaike information criterion 1048.0. Combining evoked resonant neural activity, beta and high-frequency oscillations ranking data yielded the strongest predictive model (r ² = 0.61, Akaike information criterion 1021.5). The 'ideal' contact (yielding maximal benefit) was ranked first according to each factor in the following proportion of hemispheres; evoked resonant neural activity 18/28, beta 17/28, anatomy 16/28, high-frequency oscillations 7/28. Across hemispheres, the maximal available deep brain stimulation benefit did not differ from that yielded by contacts chosen by clinicians for chronic therapy or contacts ranked first according to evoked resonant neural activity. Evoked resonant neural activity, beta oscillations and anatomy similarly predicted how motor benefit from subthalamic nucleus deep brain stimulation varied across contacts on each lead. This could assist programming by providing a probability ranking of contacts akin to a 'monopolar survey'. However, these factors identified the 'ideal' contact in only a proportion of hemispheres. More advanced signal processing and anatomical techniques may be needed for the full automation of contact selection.

Keywords: Parkinson’s disease; deep brain stimulation; evoked potentials; local field potentials; subthalamic nucleus.

Graphical Abstract

Figure 1

Neuronal signal characteristics. (A) The ERNA amplitude and power spectral density of (B) beta and (C) HFO recorded at each of the four contacts in the right hemisphere of participant 1. E0, most ventral contact in the substantia nigra pars reticulata; E1, ventral STN contact; E2, dorsal STN contact; E3, most dorsal contact in the zona incerta. (D) ERNA power [repeated-measures ANOVA, F(3,94) = 23.4, P < 0.001], (E) beta power [repeated-measure ANOVA, F(3,94) = 22.0, P < 0.001] and (F) HFO power [repeated-measures ANOVA, F(3,94) = 53.1, P < 0.001] at contacts ranked according to the neuronal signal power. Red horizontal line in D above the x-axis represents the power range for beta oscillations and HFO. *At Rankings 1–3, n = 28 hemispheres. At Ranking 4, n = 27 hemispheres, as recordings were not obtained in one ventral contact due to a technical fault. Bars represent SEMs. ERNA, evoked resonant neural activity; HFO, high-frequency oscillations.

Figure 2

Anatomical location of neuronal signals and degree of motor benefit with DBS. (A) ERNA power [repeated-measures ANOVA, F(3,94) = 11.4, P < 0.001], (B) beta power [repeated-measures ANOVA, F(3,94) = 4.9, P = 0.003] and (C) HFO power [repeated-measures ANOVA, F(3,94) = 19.3, P < 0.001] at contacts ranked according to proximity to the nominated ideal anatomical location for DBS in the STN region. (D) Hemibody UPDRS DBS benefit at contacts ranked according to the degree of motor benefit with DBS [repeated-measures ANOVA, F(3,94) = 137.6, P < 0.001]. *At Rankings 1–3, n = 28 hemispheres. At Ranking 4, n = 27 hemispheres, as recordings were not obtained in one ventral contact due to a technical fault. Bars represent SEMs. ERNA, evoked resonant neural activity; HFO, high-frequency oscillations; DBS, deep brain stimulation; UPDRS, Unified Parkinson’s Disease Rating Scale.

Figure 3

Relationship between motor benefit with DBS and neuronal signals and anatomical location of contacts. Hemibody UPDRS DBS benefit at contacts ranked according to (A) ERNA power (MEM, r² = 0.50, AIC 1039.9, P < 0.001), (B) beta power (MEM, r² = 0.50, AIC 1041.6, P < 0.001), (C) HFO power (MEM, r² = 0.44, 1057.2, P < 0.001) and (D) proximity to the nominated ideal anatomical location (MEM, r² = 0.49, AIC 1048.0, P < 0.001). (E) Hemibody UPDRS DBS benefit at the contact yielding maximal benefit, at the contact chosen by the clinician for chronic DBS and at the contact ranked first according to ERNA power, beta power, HFO power and proximity to the nominated ideal anatomical location [repeated-measures ANOVA, F(5,149) = 11.9, P < 0.001]. In A–D, raw means (dots) and standard errors (bars) are presented in the figures, whilst statistical analyses employed fitted means adjusted for fixed and random effects. *At Rankings 1–3, n = 28 hemispheres. At Ranking 4, n = 27 hemispheres, as recordings were not obtained in one ventral contact due to a technical fault. AIC, Akaike information criterion; DBS, deep brain stimulation; ERNA, evoked resonant neural activity; HFO, high-frequency oscillations; MEM, mixed-effects model; UPDRS, Unified Parkinson’s Disease Rating Scale.

Figure 4

Relationship between maximal available DBS benefit, chronic DBS location and neuronal signals and anatomical location of contacts. (A) The relationship between contacts yielding greatest motor benefit (UPDRS) with DBS and the ranking of those contacts according to the various factors. In four hemispheres, the UPDRS DBS benefit was the same in two different contacts and both contact rankings are represented. (B) The relationship between contacts selected by the clinician for chronic DBS and the ranking of those contacts according to the various factors. DBS, deep brain stimulation; ERNA, evoked resonant neural activity; HFO, high-frequency oscillations; UPDRS, Unified Parkinson’s Disease Rating Scale.