Walking Again: How AI Is Fixing Parkinson’s Gait
Gonzalo Lozano
So here's the thing about Parkinson's gait: it's basically a problem nobody's really solved yet. Patients shuffle, take these tiny steps, sometimes freeze in place. Standard deep brain stimulation (DBS) is amazing for tremor and stiffness, but for walking? Not so much.
But researchers at UC San Francisco figured out something interesting. They basically asked: what if we used AI to personalize how we stimulate the brain for walking specifically? Turns out, that might actually work.
The Problem (And Why It's Been So Hard)
"Gait has been quite difficult to treat," Dr. Doris Wang explains. She's the neurosurgeon leading all this. "Although we use continuous high-frequency DBS to treat tremor and slowness and stiffness of movement, it doesn't work well for gait."
So the standard approach just... doesn't work for walking. People with Parkinson's shuffle, take mini steps when turning, have lopsided strides, and sometimes just freeze. Which obviously makes them fall a lot. And that's the disability part.
So What Did They Do?
Wang's team basically took it as an engineering challenge. They put in bidirectional DBS devices (ones that both stimulate and listen to the brain) and had patients walk around while recording their neural data and gait. Then they'd try different stimulation settings to see what actually helped.
They created something called the Walking Performance Index (WPI), which measures arm swing, stride speed, how consistent the steps are, and whether both sides are balanced. Pretty straightforward stuff.
Then they threw the data at machine learning algorithms and asked: which DBS settings actually work for each person?
And here's where it gets interesting, different patients needed different frequencies. Some people did better with high frequency, others with lower frequency. So the ‘one size fits all’ thing everyone's been doing? Not actually optimal.
The Brain Part
When patients walked better, researchers noticed reduced beta-band brainwave activity in the globus pallidus. Which is a region involved in movement control. So there's actually a neural signature of "good walking" that they can now look for.
Where It's Going
Wang's team is working on a closed-loop system where the device automatically switches to gait-optimized settings when you're walking, then goes back to normal for everything else. UCSF started a clinical trial in October 2025 to see if this actually reduces freezing and improves walking.
Maybe, finally, there's actually an answer to this problem.
Sources:
Wang, D., et al. (2025). "Tailored Deep Brain Stimulation Improves Walking in Parkinson's." UC San Francisco. https://www.ucsf.edu/news/2025/07/430381/tailored-deep-brain-stimulation-improves-walking-parkinsons
Wang, D., & Azgomi, H. F. (2025). "Q&A: New tech helps Parkinson's patients who have trouble walking." University of California. https://medicalxpress.com/news/2025-11-qa-tech-parkinson-patients.html
University of California. (2025). "This New Tech Helps Parkinson's Patients Who Have Trouble Walking." https://www.universityofcalifornia.edu/news/new-tech-helps-parkinsons-patients-who-have-trouble-walking
UCSF Clinical Trials. (2025). "Adaptive Deep Brain Stimulation to Improve Freezing of Gait in Parkinson's Disease." https://clinicaltrials.ucsf.edu/trial/NCT06819020
