Artificial intelligence has been used to predict brain stimulation settings that could improve walking in people with Parkinson’s disease.
Researchers at the University of California, San Francisco (UCSF), developed personalised algorithms to estimate which deep brain stimulation (DBS) settings would be most effective for each individual. In early tests, participants showed improvements in walking performance.
DBS is a treatment where electrodes are implanted in the brain to deliver electrical pulses that help control movement symptoms in Parkinson’s. The study explored how AI could refine this technique by tailoring stimulation to each patient.
Walking difficulties – also known as gait dysfunction – are common in Parkinson’s, alongside tremors and bradykinesia (slowed movement).
While DBS can be effective for tremor and bradykinesia, its impact on gait has been less consistent.
Dr Doris Wang, senior author and neurosurgeon at UCSF, said: “This work not only deepens our understanding of how DBS affects movement, but also highlights the promise of personalised neuromodulation for Parkinson’s and other neurological disorders, bringing us closer to smarter, more effective neuromodulation therapies.”
Three people with DBS implants took part in the study as part of a larger clinical trial.
Their devices included electrodes in both deep brain regions and the outer layer of the brain involved in movement, allowing researchers to record signals from both areas.
Participants walked a six-metre track wearing sensors that captured details of their gait.
The team compared walking performance with recorded brain signals and tested how both changed with different DBS settings.
Dr Hamid Fekri Azgomi said: “We approached the problem of optimising DBS settings as an engineering challenge, aiming to model the relationship between stimulation parameters, brain activity, and walking performance.”
The researchers identified common features in brain activity linked to walking across all participants, along with features specific to each individual.
They used this data to create a map of how stimulation settings affect performance.
A machine learning approach – a form of AI – was used to analyse this map and predict other DBS settings that could improve walking.
“Testing the identified settings further validated the model’s efficacy, resulting in significant improvements in walking performance,” the researchers wrote.
The findings suggest that a data-driven, individualised approach could help clinicians better programme DBS devices. The team hopes to automate the process and integrate it into programming software.
“Gait dysfunction reduces mobility, increases fall risk, and significantly impacts a patient’s quality of life,” the researchers noted.
“Future work with larger, more diverse populations is needed to validate these findings and fully realise the promise of personalised DBS therapies.”
