Saeed Mollahoseini Paghale, Milad Fallahzade, Mohammadreza Amirseyfadini,
Volume 23, Issue 30 (1-2026)
Background and Aims: Controlling hand tremors in neurological disorders like Parkinson's has gotten a lot of attention in recent decades. The number of theories about closed-loop deep brain stimulation is rapidly growing. The goal of this work is to offer a machine learning-based automated closed loop system for the rehabilitation of Parkinson's patients with hand tremor symptoms.
Materials and Methods: In the current study, vibration was simulated using a mathematical model that included a muscle model, basal ganglia, cortex, and supplementary motor area. To manage hand tremor, the non-integer PID proportional controller, as well as the intelligent Proximal Policy Optimization (PPO) algorithm as a subset of reinforcement learning, are employed to adapt the coefficients.
Results: One of the advantages of the proposed method, aside from reducing hand tremor and automatic learning to use at various levels of the disease, which has yielded acceptable results when compared to other control methods, is its practical implementation in the real world due to the simplicity of the controller. The automatic adjustment of artificial intelligence network coefficients in the presented strategy (PPO) makes it simple to create intelligent system.
Conclusion: The proposed intelligent system significantly reduces the side effects of continuous brain stimulation in the open-loop manner stimulation, in addition to optimizing output signals such as hand tremor compared to other controllers and being usable for all levels of the disease due to its adaptability.
