Parkinson’s disease is a progressive nervous system disorder that affects more than 10 million people around the world. Most people with Parkinson’s disease start showing systems around 60 years of age, and one of the most common early symptoms is had tremors. Those can affect one or both hands, and can vary in severity from mildly annoying to nearly debilitating. A lot of research in recent years has been put into helping people with Parkinson’s disease cope with those tremors, and robot augmentation is very promising. An international team of scientists has recently developed a machine learning model that can help to control those robotic devices.
The idea behind these proposed robotic augmentations is similar to what we’ve seen with robotic exoskeletons that allow average people to dramatically increase their strength. In this case, however, the robotic devices don’t just need to improve a person’s strength—they also need to help the user remain steady. While a Parkinson’s disease sufferer’s hand might tremble, the robotic augmentation should remain stable to compensate. Without that compensation, the augmentation is, at best, limited in its usefulness. At worst, the augmentation could actually be dangerous to the user and others.
This machine learning model is able to provide that compensation by predicting the physical result of tremors. It can do that thanks to a training data set that was built by using small sensors to monitor that hand movement of 81 individual patients. Those patients were all in their ‘60s and ‘70s and living with Parkinson’s disease. With the collected data, a neural network model was able to determine what kind of physical hand tremors were likely to occur based on the user’s current movement. The system can then compensate for those hand tremors with very little lag. That could potentially allow a person with Parkinson’s disease to operate a robotic augmentation as they intend, rather than relying on the movement of their hands that they’re unable to control.