AI-driven adaptive deep brain stimulation improves Parkinson’s symptoms

By Pooja Toshniwal PahariaReviewed by Benedette Cuffari, M.Sc.Aug 22 2024

In a recent study published in Nature Medicine, researchers compare the safety and feasibility of an adaptive deep brain stimulation (aDBS) system to a conventional DBS system for Parkinson's disease (PD).

Study: Chronic adaptive deep brain stimulation versus conventional stimulation in Parkinson’s disease: a blinded randomized feasibility trial. Image Credit: Max Acronym / Shutterstock.com

Current treatment options for PD

PD is a neurological disease that affects millions of people 60 years and older, many of whom reside in high-income nations. PD treatment often includes levodopa, a medication that replaces dopamine-producing neurons in deep brain areas that are important for movement control. However, excess dopamine can lead to dyskinesia, uncontrollable movements, tremors, and stiffness, which may limit the efficacy of this treatment in some patients.

As a result, some PD patients are treated with continuous DBS (cDBS) devices, which deliver continuous electrical stimulation to the brain. Nevertheless, cDBS devices often lack dynamic reactivity to changing clinical and neurological conditions.

PD patients may require a brain stimulation device that can adjust their dopamine levels and recognize brain signals associated with various symptoms. Feedback control may improve treatment efficacy; however, the optimal technique and advantages of adaptive neurostimulation remain unclear.

Although the aDBS system has been shown to improve insomnia, standard-of-care optimization for PD remains insufficient. Furthermore, motor cortical signals have the potential to affect aDBS; however, their effect on stimulation amplitude is unclear.

About the study

In the present study, researchers conducted a feasibility experiment on PD to compare the effects of aDBS and optimized cDBS during daily activities. Four individuals with PD for at least four years who took antiparkinsonian medications, experienced motor fluctuations, and had Movement Disorder Society Unified PD Rating Scale, Part III (MDS-UPDRS-III) scores between 20 and 80 that improved by 30% or more with medication were included in the study. Patients who previously had cranial surgery, implanted stimulation systems, electroconvulsive treatment, repeated transcranial magnetic stimulation, or diathermy were excluded from the study.

Between April 24, 2019, and February 26, 2022, PD patients had DBS electrode implants for aDPS. A seven-step workflow was designed to use aDBS to manage chronic motor disturbances. Low- and high-amplitude thresholds for aDBS were defined as 0.5-1.0 mA lower and higher than the optimal cDBS amplitudes, respectively.

In-clinic and at-home brain recordings were used to identify the biomarkers or neural signals closely linked to residual motor symptoms. Both medication and stimulation states were variable. The cDBS settings were adjusted for patient-tailored adaptive algorithms through supervised short- and long-term at-home testing.

All patients participated in randomized, blinded, cross-over comparisons of aDBS and cDBS for one month. This experiment was conducted during the participants' regular lifestyles, including travel and work, during which the patients continued their drug regimen while optimizing cDBS.

Both cDBS and aDBS systems use artificial intelligence (AI) to monitor brain activity for symptom changes and intervene with precisely calibrated electric pulses. The automated pipeline explored the frequency space of field potentials in the subthalamic nucleus (STN) and sensorimotor cortex for physiological signals to alleviate the most difficult symptoms.

This therapy supplements PD drugs by delivering less stimulation when the drug is active and more stimulation when the medication's effects wear off to minimize rigidity. Slower movements in PD patients indicate a shift in the neutral biomarker, which would then fall below the threshold, and stimulation amplitude would rise. Stimulation adjusts to compensate for slower movements, thereby increasing movement speed.

Study findings

The closed loop aDPS-based brain implant decreased the most unpleasant PD symptoms by 50%. In all four PD patients, stimulation-entrained gamma oscillations in the STN or motor cortex were the best indicators of high or low dopaminergic states and residual motor symptoms.
Adaptive stimulation reduced the time spent with troublesome motor symptoms as compared to clinically optimal conventional stimulation. Converging data from at-home and in-clinic recordings indicated that the stimulation-entrained gamma oscillations centered at 50% of the stimulation rates were the best predictors of medication-associated symptom status.

Levodopa-induced precisely tuned gamma oscillations between 60 and 90 Hz to subharmonic stimulation frequencies. Moreover, aDBS improved the proportion of awake time patients spent with the most troublesome symptoms as compared to cDBS while decreasing awake time percentages with opposite symptoms. The aDBS system also improved the life quality of the participants.

Conclusions

Tailored neuromodulation can improve PD patient mobility and sleep, with aDPS capable of detecting PD symptoms in real time and matching the appropriate stimulation level. This AI-powered technique provides a remarkable advancement compared to existing PD treatments and can be used to improve current knowledge on the complexities of brain signaling and PD pathophysiology. Taken together, the study findings provide the foundation for large-scale clinical studies to determine the efficacy of tailored adaptive neurostimulation in neurological diseases.