Time domain reconstruction of basal ganglia signals in patient with Parkinson's disease

In this paper, we developed a method of simulation for intracerebral signals acquired during Deep Brain Stimulation - DBS surgery in one patient with Parkinson's disease. Based on our previous work, an auto-regressive (AR) parametric model with order 13 was used, because it generates one of the most accurate representations of basal ganglia signals in movement disorders. Then, the AR parameters were estimated in the Z transform domain with preset prediction horizon below 5 samples. Subsequently, a polynomial regression of the system parameters was performed, associated with the depth of each track of microrecording. Using these regression coefficients, a set of arbitrary signals was generated at different depths using Gaussian noise and their performance was assessed via cross-validation. Finally, we reconstructed the signals through transformation into the time domain. The proposed methodology shows mean accuracy near to 95% between the real and simulated signals. This work could contribute to the future development of a training system for stereotactic neurosurgery based on intracerebral signals.