Robust neuromorphic coupled oscillators for adaptive pacemakers

Abstract

Neural coupled oscillators are a useful building block in numerous models and applications. They were analyzed extensively in theoretical studies and more recently in biologically realistic simulations of spiking neural networks. The advent of mixed-signal analog/digital neuromorphic electronic circuits provides new means for implementing neural coupled oscillators on compact, low-power, spiking neural network hardware platforms. However, their implementation on this noisy, low-precision and inhomogeneous computing substrate raises new challenges with regards to stability and controllability. In this work, we present a robust, spiking neural network model of neural coupled oscillators and validate it with an implementation on a mixed-signal neuromorphic processor. We demonstrate its robustness showing how to reliably control and modulate the oscillator’s frequency and phase shift, despite the variability of the silicon synapse and neuron properties. We show how this ultra-low power neural processing system can be used to build an adaptive cardiac pacemaker modulating the heart rate with respect to the respiration phases and compare it with surface ECG and respiratory signal recordings from dogs at rest. The implementation of our model in neuromorphic electronic hardware shows its robustness on a highly variable substrate and extends the toolbox for applications requiring rhythmic outputs such as pacemakers.

Neuromorphic bioelectronic medicine for nervous system interfaces: from neural computational primitives to medical applications

“…By construction, the types of computational primitives and signal processing operations performed by mixed-signal neuromorphic circuits in hardware networks of spiking neurons are the same ones that can be observed in animal nervous systems. The neuromorphic circuits in question can carry out different types of linear, non-linear, or adaptive filtering operations [96][97][98], they can implement different types of normalizing operations on different temporal and spatial scales [99,100], they can implement delay chains [101], oscillators [35,102], resonators [103], decision-making networks [95], and importantly, they can be designed to implement SNNs with on-chip and on-line spike-based learning properties to solve classification or regression tasks [89][90][91]104].…”

“…Therefore these circuits can be of great use for the control of periodic movements [103] (e.g. in pacemakers [35]), as it is commonly done in nature by their biological equivalent CPG circuits [118]. In addition to being useful for producing periodic outputs, these circuits can also be used as 'resonators' [119], to respond robustly to periodic inputs of desired frequencies, while being less sensitive to noise and distractors in other frequency bands.…”

“…Indeed, by combining multiple computational primitives together it will be possible to design neural processing systems optimally suited to processing spatiotemporal biological signals and to solving in bioelectronic medicine applications. For example, spike frequency adaptation can be implemented in the input layer of a neuromorphic bioelectronic SNN device using adaptive silicon neurons circuits [110,129,130]; feed-forward inhibition can be accomplished by routing the input events to inhibitory synapse circuits designed using log-domain integrators [98,131]; neuromorphic coupled oscillator circuits can be implemented using hard-wired connectivity schemes [132,133], or by programming excitatory and inhibitory connections in the router lookup tables of multi-core neuromorphic systems [35,103]; representing biosignals using population coding can be achieved by using multiple neurons per input node. Population coding is particularly important, as it enables us to achieve robust computation overcoming the core-to-core and chip-to-chip device mismatch effects [134].…”

“…Similar to the neural processes they model, neuromorphic systems are adaptive, fault-tolerant, and can be flexibly configured to display complex behaviors by combining multiple instances of simpler elements and process information using energy-efficient asynchronous, event-driven approaches. For example, preliminary results on building an adaptive neuromorphic system that performs modulation of the heart chamber and restores missing biological functionalities show that is possible to reproduce the behavior of the biological respiratory central pattern generator (CPG) in silicon neurons and generate coherent heart chambers activation in medical conditions such as respiratory sinus arrhythmia (RSA) [34,35]. Another example, implement a closed loop neuromorphic decoder for intra-cortical brain-machine interface in a non-human primate with a success rates of about 96% for neural control trial.…”

Remodeling in the AV block dog is essential for tolerating moderate treadmill activity

“…Complex device technologies and algorithms developed over the last four decades focus on rate-adaptive pacing based on physiological input, including physical activity [3] , transthoracic impedance [4] , respiration [5] , [6] , and even a combination of various physiological parameters [7] . However, optimization of sensor sensitivity is challenging, and studies lack significant improvement outcome of the sensors on top of pacing at a fixed rate to treat bradycardia [8] , [9] .…”