DGIST Researchers Develop NONOS to Precisely Separate Brain Rhythms

New AI Signal Processing Refines Parkinson’s Deep Brain Stimulation

Patients living with Parkinson’s disease may soon benefit from a more precise approach to deep brain stimulation (DBS), thanks to a new deep-learning technology developed by a collaborative team of researchers in South Korea. Known as “NONOS,” the analysis tool separates rhythmic brain activity from background signals with high temporal resolution. The findings, published in the journal NeuroImage, suggest a significant leap in how we might tailor electrical therapy to the specific, moment-to-moment needs of a patient’s brain.

The core challenge in DBS therapy has long been the “noise” of the brain. When clinicians monitor a patient, they look for specific rhythmic oscillations—most notably “beta waves”—that spike in intensity when the disease causes movement to slow. Traditional filters often struggle to differentiate these rhythmic bursts from the underlying, non-oscillatory background signals. It is much like trying to isolate a single voice in a crowded room; if the background noise is misidentified as a rhythmic pulse, conventional methods can mistake changes in background signals for changes in rhythms.

Overcoming the Blurred Vision of Brain Signals

Previous attempts to clean up these signals have relied on fixed time windows, a method that often creates a “motion blur” effect. Just as a long-exposure photograph obscures a fast-moving object, these older analysis techniques frequently miss the rapid, fleeting changes in brain rhythms that are critical for timely, adaptive stimulation. The research team, led by Professor Ji-Woong Choi of the Daegu Gyeongbuk Institute of Science and Technology (DGIST), focused on the mathematical nature of these background signals. Because background power naturally decreases at a predictable, exponential rate as frequency increases, the team trained the NONOS AI to recognize and subtract this specific pattern.

By effectively stripping away the background noise, the system leaves behind a pure stream of rhythmic data. This allows for the detection of rapid changes in brain activity without the lag or loss of resolution that has plagued earlier technology. In testing, the results were striking. When the team used NONOS to analyze brain signals from Parkinson’s patients, the AI’s ability to predict which foot a patient would step forward with reached an accuracy of 0.98. For perspective, conventional methods typically hover between 0.54 and 0.79 in similar predictive tasks.

Real-World Implications for Adaptive Therapy

The potential clinical relevance of this technology is personalized deep brain stimulation therapy, which adjusts electrical stimulation according to each patient’s condition. According to Professor Choi, the next phase of development involves shrinking the computational requirements of the NONOS model so it can run in real time directly on implantable hardware. If successful, this would allow for a truly personalized therapy that adjusts electrical output based on the patient’s immediate clinical state.

“The significance of this study lies in demonstrating that rhythms and background signals mixed together in brain signals can be separated without losing temporal resolution,” said Professor Ji-Woong Choi of DGIST. “If we further develop NONOS to operate in real time on implantable brain devices by reducing its computational requirements, it will contribute to improving the accuracy of personalized deep brain stimulation therapy.”

Broadening the Scope of Biological Analysis

While the immediate focus is on Parkinson’s disease, the research team suggests the utility of this technology is much broader. Professor Han-Joon Kim of the Kumoh National Institute of Technology noted that because NONOS can separate periodic and aperiodic components in various biological signals that change rapidly, the model could eventually be applied to a wide range of time-varying physiological data.

The project, which included researchers from the Kumoh National Institute of Technology, Korea University, and the Korea Brain Research Institute, was supported by the National Research Foundation of Korea and the Ministry of Education’s Glocal Lab program. By solving the fundamental problem of signal separation, the team has provided a new, high-resolution window into the complex, rhythmic world of the human brain.

Keep reading