A flexible, low power, multi-channel, real-time BCI for seizure monitoring and Modulation

PROJECT SUMMARY Implantable neuromodulation devices for seizure control have emerged as an important treatment for medically refractory epilepsy, particularly for focal seizure disorders increasingly identified as distributed in networks not amenable for surgical resection. Current devices, however, are not suited for monitoring and modulating a brain network due to limitations in channel count, computational capabilities, programmability, and power budget. Our long-term goal is to develop a programmable, low power embedded device which can be used for the control of seizures through monitoring and modulating brain network activity. The overall objective of this application, which is the next step towards attaining our long-term goal, is to develop a programmable computational device to perform signal processing in real-time and monitor and modulate multiple brain locations within a tight power budget. In conjunction with this hardware solution, we will implement an open language and evaluate the hardware and software solutions through bench and animal studies. Our central hypothesis, based on our preliminary work, is that instead of the current approach of performing feature detection in an analog front-end or using a single application specific IC (ASIC) – which place hard constraints on programmability – a programmable and low power digital design can be achieved with asynchronous (clock- less) design. The rationale for the proposed research is that once the hardware and software solutions are developed, we will be able to more readily advance the use of devices for seizure control and for research on brain networks. We will achieve the objective of this program through three aims. In Aim 1 we will develop a processor called neuro Stream for real-time processing of streaming multichannel data. Neuro Stream will feature a reduced instruction set computer-V (RISC-V) microcontroller unit, processing elements for specific signal processing tasks, and a vector engine. Neuro Stream will be designed using asynchronous techniques, operate within a limited power budget, and support execution of a wide range of computational methods. In Aim 2 we will implement open languages neuro Octave and neuropy. Neuro Octave will be a subset of Octave, an open-source language. Neuropy will be a subset of Python. Neuro Octave and neuropy will allow users to develop their own programs for neuro Stream using familiar, commonly used, languages. In Aim 3 we will implement monitoring and control algorithms in neuro Stream and control seizures in a model of epilepsy. This contribution will be significant because it will bring the development of epilepsy devices fully into the realm of the network theory of epilepsy allowing developments in the modulation of networks to emerge and be applied for the device-based control of seizures. The proposed research is innovative, in our opinion, because in contrast to previous advances, we will develop a programmable, powerful, all-digital solution which can be extended to incorporate future research on algorithms for seizure detection, seizure forecasting and seizure control including algorithms which use modern data science methods of machine learning and deep learning.