Optimization of a Phase–to–Amplitude Coupling Algorithm for Real–Time Processing of Brain Electrical Signals

In this study a Phase–to–Amplitude cross– frequency Coupling (PAC) algorithm suitable for hardware implementation is described. The proposed architecture is aimed for real-time classification of brain states in the context of close–loop Deep Brain Stimulation (DBS) paradigm. The results obtained with the proposed PAC algorithm shown the feasibility of compute comodulograms without using the Hilbert transform. The latter yields a significant reduction in the computational complexity and is valuable for optimizing both, silicon area and power consumption related to the implementation of the PAC algorithm in implantable devices. In addition, acceleration of PAC code for off–line data processing is explored through multi–core and GPU processing. Finally, a case of study using synthetic signals and real in vivo brain recordings is presented that shows the potential of PAC phenomenon as a neurophysiological marker. Keywords—Deep Brain Stimulation, Phase–Amplitude Comodulogram, Computational complexity, Multi–core processing, FPGA.