DBS works by continuously delivering electrical stimulation to particular regions of the brain functionally associated with movement. Stimulation is currently (pun!) delivered regardless of whether or not the person with the device is experiencing symptoms or not. This can lead to undesired side effects and increased battery consumption. A better system would only deliver stimulation when symptoms are occurring. This type of system is referred to as a "closed-loop" deep brain stimulation system.
My research uses an experimental DBS device that allows us to record neural signals from the cortex (outer surface) of the brain. This part of the brain has large signals that change during movement. I use machine learning to pick out features in these signals that are predictive of symptoms, such as tremor. Then, in real time, these features are used to predict if the patient is experiencing symptoms. The system then turns DBS on or off accordingly.
The movies below show a patient with essential tremor (tremor while moving his right arm) with this DBS device implanted. He is performing a tremor assessment task that involves drawing a spiral on a piece of paper. He performs the task with his stimulation turned off, stimulation turned on, and then with our adaptive system controlling stimulation. For this task, the adaptive system alleviates tremor about as well as when his stimulation is on, which is an encouraging result.
| Stimulation off | Stimulation on | Adaptive stimulation |
The video below shows the closed-loop DBS system in action while one of the ET patients is performing a movement. The bottom three plots are a trace from a gyroscope showing movement and red ticks showing movement detection from the classifier (top), the stimulation voltage (middle), and a spectrogram showing changes in oscillatory power in the cortex (bottom). The patient starts at rest, and no movement is detected. He then begins to move, and tremor can be seen in his hand. The classifiers detect the movement (red bar in top plot) and the stimulation voltage begins to increase. After he is done moving, the classifiers detect rest, and the stimulation voltage decreases back to zero.
If you want more technical details about this project and/or related work, please see some of my publications below, or check out work from other members of my lab.
B Houston, M Thompson, J Ojemann, A Ko, H Chizeck, "Classifier-based closed-loop deep brain stimulation for essential tremor", 8th International IEEE/EMBS Conference on Neural Engineering, Shanghai, China, 2017.
B Houston, Z Blumenfeld, E Quinn, H Bronte-Stewart, H Chizeck, "Long-Term detection of Parkinsonian tremor activity from subthalamic nucleus local field potentials", 37th Annual Conference of IEEE EMBS, Milan, Italy, 2015.