For patients with Parkinson's disease, changes in their walking ability are often dramatic. "Parkinsonian gait," as it's often called, can result in changes in stride length and asymmetry between the legs. This gait disorder reduces a person's mobility, increases the risk of falls, and significantly impacts a patient's quality of life. Researchers at the University of California, Berkeley, have now developed tailored deep brain stimulation (DBS) using machine learning to improve individualized Parkinson's gait.
Parkinson's is the fastest-growing brain disease worldwide. While high-frequency deep brain stimulation (DBS) is highly effective in reducing symptoms such as tremor, rigidity, and bradykinesia (slowing of movement), its impact on gait is more variable and less predictable in patients with advanced gait impairment. Key challenges in improving DBS outcomes in advanced gait impairment include:
- The lack of a standardized gait method that clinicians can use during programming.
- Understanding the impact of different stimulation factors on gait.
Machine learning
In a recent study, researchers at the University of California, San Francisco (UCSF) developed a systematic way to quantify key aspects of gait relevant to Parkinson's disease. They used machine learning to identify the best DBS settings for each individual. According to the researchers, these personalized settings led to meaningful improvements in gait — such as faster, steadier steps — without exacerbating other symptoms.
"We approached optimizing DBS settings as a technical challenge, with the goal of modeling the relationship between stimulation parameters, brain activity, and gait performance," says Hamid Fekri Azgomi, PhD, lead author of the study and a postdoctoral researcher in the Wang Lab at UCSF. The study results are published in npj Parkinson's Disease.
Impact on gait function
During the study, Parkinson's patients were implanted with a DBS device that both stimulates the brain and records neural activity during gait. During clinic visits, the patients' DBS settings were adjusted within safe limits to investigate the impact on gait. In response to each set of DBS settings, participants walked in a loop of approximately six meters, generating a continuous stream of their neural data and gait kinematics.
The researchers then developed a gait performance index (WPI) to assess gait metrics such as step length, stride velocity, and arm swing amplitude, providing insight into gait consistency. By combining these metrics, the WPI offered a comprehensive assessment of gait, covering multiple dimensions of motor function affected by Parkinson's disease.
According to Azgomi, the results confirmed that changes in DBS settings were effectively captured by the WPI and that the results were consistent with patient and clinician assessments at each visit. "This validation supports our conclusion that the WPI is an effective measure for assessing and guiding gait improvements in people with Parkinson's disease. Using these techniques, we were able to predict and identify personalized DBS settings that improved the WPI."
The researchers also identified brain activity patterns associated with improved gait. Using multivariate models, they then identified distinct neural dynamics that differentiate optimal gait performance from less effective patterns. For example, improved gait correlated with reduced beta-band brainwave activity during specific phases of the gait cycle in the globus pallidus, a part of the brain associated with loss of muscle movement in people with Parkinson's.
Importance of personalized interventions
These findings, along with the identified individual-specific neural biomarkers, underscore the importance of personalized, data-driven interventions for improving gait in people with Parkinson's disease, according to the study.
Co-lead author Doris Wang, a neurosurgeon and associate professor of neurosurgery at UCSF, states that the study's conclusions not only impact our understanding of how DBS affects movement but also offer the promise of personalized neuromodulation for Parkinson's disease and other neurological disorders. "This brings us closer to smarter, more effective neuromodulation therapies."
Follow-up research will include the development of automated systems for real-time gait analysis and the integration of WPI with DBS programming software. Technologies such as walking mats, wearable sensors, and advanced motion capture systems could enable continuous and accurate gait monitoring, which could enable more precise DBS adjustments.
