An increased interest in the investigation of the inner workings of the brain, together with recent technological advancements have been great catalysts for the development of neural stimulation and signal recording systems. These neural interfaces have enabled a better understanding of underlying neurological diseases, and provide promising therapeutic interventions for various neurological disorders.
As discoveries and technological advancements continue, new challenges and opportunities emerge. One of the major challenges is the development of small, portable, and power-efficient closed-loop neuromodulation systems. The ability to simultaneously stimulate and record is a key capability required in enabling such systems.
A closed-loop neuromodulation system is comprised of mainly four elements: (a) Stimulator: an energy-efficient and flexible stimulation engine, (b) Sensing: Low-power, high dynamic-range analog front-ends, (c) Digital Signal Processing (DSP): energy-/area-efficient digital signal processing units, and (d) Wireless transfer: an energy-efficient wireless power and data transfer unit. In summary, efficient and concurrent stimulation, sensing, processing, and transfer of neural signals are required. Design efforts are in full effect to realize leading edge stimulation, sensing, and wireless transfer technologies; however, one common difficulty in realizing concurrent stimulation and recording of neural signals is the presence of stimulation artifacts observed at the sensing end. Existing solutions (e.g., blanking the recording channel during stimulation or self-cancelling stimulation electrodes) have not answered all the challenges and lack the ability of continuous signal recording during the stimulation phase, thus rendering a critical portion of the data unusable.
In this work we propose an energy-efficient, implantable, real-time Adaptive Stimulation Artifact Rejection (ASAR) engine, capable of adaptively removing stimulation artifact for varying stimulation characteristics at multiple sites. Additionally, a blind artifact template detection technique is introduced, which in combination with the proposed ASAR algorithm, eliminates the need for any prior knowledge of the temporal and structural characteristics of the stimulation pulse; this technique also enables us to effectively battle the non-linear mapping of brain tissue, and non-idealities of electrode interfaces, with linear filtering.
Two engines, implemented in 40nm CMOS, achieve convergence of <42μs for Spike ASAR and <167μs for LFP ASAR, and can attenuate artifacts up to 100mVp-p by 49.2dB, without any prior knowledge of the stimulation pulse. The LFP and Spike ASAR designs occupy an area of 0.197mm2 and 0.209mm2, and consume 1.73μW and 3.02μW, respectively at 0.644V.
The LFP ASAR is integrated in a 64-channel sensing chip used in a state-of-the-art implantable, closed-loop neuromodulation unit (NM).
