Deep brain drug-delivery control using vagus nerve communications

Abstract

Vagus nerve stimulation (VNS) uses electrical impulses applied at the neck in order to mitigate the effects of, for example, epileptic seizures. We propose using VNS to provide data pulses to communicate with a drug-delivery system embedded near the brainstem. We model the generation of a vagus nerve compound action potential (CAP), calculating the signal attenuation and the resulting transmission range. The metabolic cost of CAP transmission in terms of the use of adenosine triphosphate (ATP) is also calculated. The channel capacity for on-off keying (OOK) is computed from the CAP characteristics, the neural refractory period and the level of background neural noise. The resulting low bit-rate, unidirectional asynchronous transmission system is analysed for the use of different methods of forward error correction (FEC) to improve bit-error rate (BER). We show a proposed data packet structure that could deliver instructions to an embedded drug-delivery system with multiple addressable drug reservoirs. We also analyse the scope for powering the drug-delivery system with energy harvested from cerebrospinal glucose.

Introduction

Communicating with powered implants embedded in the human body, such as cardiac pacemakers or defibrillators, is an increasingly important part of medical treatment. Wireless communications systems provide a bidirectional flexible link over short ranges [1], but place power requirements of hundreds of microwatts on an implant [2]. Apart from the power requirements, the transmission of commands from an external wireless electromagnetic (EM) module to an implanted device does create security implications and may be subject to hacking [3]. An alternative method is to use the peripheral or cranial nervous system in the body to transport data commands from a transmitter at one point on a nerve to an embedded receiver at another point. The use of neural communications is part of wider research into implanted devices at the micro and nano level [4] where restrictions in antenna size will limit the use of EM systems.

Data communications using action potentials (AP) along a single neuronal axon is proposed by Parcerisa-Giné and Akyildiz [5] as an option for providing a physical link between embedded micro devices. Channel models for communications using neural spikes (APs) and neurotransmitters between hippocampal neurons have been developed by Malak and Akan [6], Ramezani and Akan [7] and Veletić et al. [8]. Data communications through the single median giant axon of the earthworm was modelled by Abbasi et al. [9] who calculated a data throughput based on four different stimulus frequencies, with each frequency representing two bits of information. A neuron channel model using a sub-threshold (non-spiking) stimulus was proposed by Khodaei and Pierobon [10], [11], though sub-threshold impulses have a very short range along an axon [12]. Single neurons are, however, difficult to access and the placement of transmitters and receivers would be particularly challenging. Consequently, it could be more practical to examine the collective stimulus of multiple neurons forming a compound action potential (CAP) to provide a neural data pulse.

There is ongoing research into the use of vagus nerve stimulation (VNS) for the treatment of epileptic seizures [13], depression, heart failure [14], arthritis [15] and Crohn’s disease. At present a VNS system relies on human intervention by the physician and patient to programme the duration and intensity of the stimulus pulses. A more flexible biofeedback system is proposed by Ward et al. [16] using the degree of measured nerve activation to control stimulus delivery and provide a personalised stimulus profile. Non-stimulus based therapies include drug-delivery to the brain, although the protective blood-brain barrier (BBB) presents a challenge for the absorption of drugs [17] to treat, for example, cancer tumours. A drug-delivery system for treating epilepsy is described by Salam et al. [18] with embedded electrodes to detect seizures and a micromechanical pump to deliver the drug from a refillable reservoir located under the scalp. A neural probe with an electrophoretic microfluidic ion pump is proposed by Proctor et al. [19] to deliver variable doses of a single drug across an ion-exchange membrane. Another release mechanism is electrothermal membrane activation, first proposed by Santini et al. [20], where a metallic membrane covers a drug reservoir. An electrical current ruptures the membrane by heating and the drug reservoir releases its contents. Either of these delivery systems could be controlled by a local processor and receive release commands from an external source along a neural pathway.

Power could be delivered to an implanted device by a long-life battery [21], although this would have to be replaced at regular intervals, requiring repeated surgical intervention. The alternative is to use some form of energy harvesting to power the implant directly or to recharge a battery. Ultrasound energy harvesting operates at shallow skin depths and will not penetrate through the bone of the skull. It is possible to power implants wirelessly with (i) near-field short-range EM inductive resonant coupling using coiled antennas at frequencies up to 20 MHz, (ii) mid-field coupling (900 MHz) or (iii) far-field (2.5 GHz) EM powering [22]. The use of EM power harvesting is subject to technical constraints to meet recommended safety levels and prevent tissue damage through excessive heating [23]. The specific absorption rate (SAR) describes the quantity of EM power that can be absorbed by a tissue and is defined as:

$$SAR=\frac{\sigma E^2}{\rho }.$$

The conductivity of the tissue is σ, the density is ρ and the electric field strength is E. The SAR value is expressed in Watts per kilogram and is averaged over 1 g or 10 g of tissue. In the US the exposure limits for an unrestricted environment, set by the FCC, are 4 W/kg for 10g of tissue in the extremities (hands, wrists, feet, ankles) and 1.6 W/kg for 1 g of head, neck and trunk tissue. In other jurisdictions the equivalent ICNIRP and IEEE guidelines specify 2 W/kg for 10 g of head, neck and trunk tissue and 4 W/kg for 10 g of any other limbs [24]. The SAR limits can be converted to power intensities at different frequency ranges and a typical value is 2 W/m² for up to 200 MHz and 10 W/m² for frequencies greater than 200 MHz [24]. Powering by EM would also require the wearing of an external powering source if true mobility was required. Ideally the implanted device should have a long-life biocompatible power harvesting system that would not have the potential to cause tissue damage and would not depend on external modules.

In previous work [25] we modelled the use of ultrasound to provide harvested power for subcutaneous nanowire-based nanodevices. We then extended that work by modelling the use of arrays of coupled nanodevices for selective neural stimulation [26]. The resulting CAPs were modelled as a data communications system (200 bit/s maximum rate) using on-off keying (OOK) [27]. Our stimulus and communications system is optimised for nerves that are at a shallow depth, are not shielded by bone and can be readily accessed for device array implantation.

In this paper we model a specific potential application: the use of neural data pulses transmitted along the vagus nerve to communicate with a programmable, multi-reservoir, drug-delivery system in the brain as shown in Fig. 1. The vagus nerve is a cranial nerve extending from the brainstem and branching to thoracic, abdominal and retroperitoneal organs. The normal neuronal signals serve to moderate functions such as heart rate, breathing and rate of digestion. Two main trunks (branches), the left and the right, can be accessed either side of the neck. The left branch of the vagus nerve, where VNS electrodes are normally placed in humans, does not include cardiac branches with motor neurons and so does not cause cardiac side effects. The main side effects are hoarseness, cough or shortness of breath [13], with no interference to normal brain function [14].

The neural stimulus system delivers current pulses ( > 0.2 mA) comparable to those delivered by FDA-approved vagus nerve stimulation (VNS) systems (0.2 mA to 5 mA). Asynchronous data packets composed of CAPs could deliver instructions to an embedded device using a unidirectional neural transmission system. Detecting neural data pulses requires lower power than receiving wireless EM signals. Unidirectional transmission implies that no acknowledgement or resend messages can be sent in the reverse direction. We, therefore, analyse the use of forward error correction (FEC) in the receiver to improve performance. The main contributions of our work are as follows:

• A model of the generation and propagation of a stimulated neural CAP along the vagus nerve, which shows that it is possible to deliver a maximum OOK data rate of 200 bit/s at ranges between 60 mm and 100 mm;

• An evaluation of the metabolic energy cost of CAP generation in terms of the use of adenosine triphosphate (ATP);

• An analysis of the bit error rate (BER) and the coding gain using a selection of FEC methods;

• A proposal for simple packet structure for programmable drug-release commands;

• An assessment of the viability of using glucose energy harvesting for powering an implanted drug-delivery system.

This article is organised as follows: the activation of the vagus nerve is described in Section 2; the data link protocol and error correction in Section 3; the drug-delivery system components, data packet structure and powering in Section 4; and our conclusions are presented in Section 5.