A Wireless Implantable Closed-Loop Electrochemical Drug Delivery System

Abstract

Wireless implantable drug delivery systems (DDSs) enable targeted, on-demand drug release to maximize therapeutic efficacy. Ultrasound has been proposed to wirelessly power and control millimeter-sized deeply implantable DDSs, but initial demonstrations encountered challenges in power transfer and release control reliability in dynamic in vivo environments. In this work, we present a closed-loop implantable DDS using ultrasound wireless power and communication in conjunction with an electrochemical drug release mechanism. The system consists of piezoelectric transducers for wireless power and data transmission, a drug delivery module containing drug-loaded electroresponsive nanoparticles, and a custom CMOS integrated circuit for closed-loop drug release using a programmable potentiostat capable of providing potentials up to ±1.5 V and sensing current up to ±100 μA. The chip also improves power transfer robustness by enabling ultrasound power combining and rectifier voltage feedback which can be used to adapt the power transmitter and minimize misalignment. Closed-loop release control is tested in vitro using the wirelessly powered DDS at 8 cm depth by adjusting the potentiostat stimulus voltage based on feedback of redox current into fluorescein-loaded nanoparticles, resulting in consistent 2 μg release across different fluorescein loading concentrations and a 39% reduction in release amount variation. These results demonstrate the effectiveness of closed-loop release control in enabling precise and reliable drug delivery.

I. Introduction

TRADITIONAL drug delivery routes involve various considerations such as drug distribution, absorption, release profile, and patient compliance [1], [2]. Achieving optimal therapeutic outcomes, particularly for chronic diseases, often requires maintaining drug concentrations within a specific range over a period of time, which can result in complex dosing regimens [3]. Individual patient responses to drugs can also vary significantly, underscoring the need for personalized medicine to maximize efficacy [4]. Wireless implantable drug delivery systems (DDSs) offer a promising solution to address many of these challenges by enabling precise and controlled drug release at a target site. Local release enhances drug uptake in target areas, reducing the required dosage and the risk of potential side effects [5]. Furthermore, controllable release profiles can maintain drug levels within the therapeutic window without requiring active management by the patient.

On-demand DDSs mainly use actively controlled stimuli (such as electric fields, magnetic fields, electromagnetic waves, or acoustic waves) to provide temporal control over drug release [5]. A variety of controlled release mechanisms have been previously investigated, including electrothermal membrane ablation [6], electrolytic pumping [7], piezoelectric valves [8], electrochemical membrane dissolution [9], electrostatic channel modulation [10], and electrical stimulation of intrinsically conducting polymers like polypyrrole (PPy) [11]. DDSs like the ones shown in [8] and [12] are controlled by the strength of an externally generated stimuli at the site of the implant, which can be difficult to precisely control at depth. Electronically controlled DDSs like those from [6], [7], [13], [14] are usually battery-powered and/or inductively-powered and can receive commands to locally generate a drug release stimulus. Battery-powered DDSs are relatively bulky due to the amount of energy needed to power the device for extended periods of time, making them difficult to implant. Near-field inductive power transfer obviates the need for a large battery. While inductive power transfer is usually demonstrated at distances smaller than the transmit/receive coil size, adaptive and resonance-based designs have been shown to achieve larger implant depths [15]. Nonetheless, most current wireless DDSs have targeted depths less than their coil size [7], [13], [14], [16], [17].

We have previously proposed using ultrasound (US) to power and control implants for drug delivery [18]–[20] (Fig. 1(a)). US provides an efficient mechanism to wirelessly transfer power in soft tissue due to its low propagation loss, high safety limit, and focusing capabilities, meaning that devices can be millimeter-sized and deeply implanted in the body near soft tissue drug targets like cancerous tumors, the heart, or the GI tract [1], [20], [21]. Example applications include on-demand, localized drug delivery for treating heart disease/myocardial infarctions [22], preventing post-surgery infections [23], and mitigating inflammatory bowel disease flare-ups [24]. Our previous proof-of-concept ultrasonic DDS used off-the-shelf wireless power recovery and control circuits integrated in a millimeter-sized footprint and demonstrated release of fluorescein (FL) in vivo in adult mice. Release was controlled electrochemically through electrical stimulation of FL-loaded polypyrrole nanoparticles (PPy NPs) [20]. Electrochemical activation means that drug release can be controlled through charge transfer in an oxidation/reduction reaction. Using PPy NPs also supports higher drug loading compared to conventional polymer films due to the high surface area of the NPs [20]. The conductivity of PPy ensures the whole structure will be in electrical contact with an electrode and the spongelike structure allows for increased drug storage capacity and faster release (Fig. 1(b)).

Fig. 1.

(a) Conceptual diagram of a wireless closed-loop implant drug delivery system with an external ultrasound transceiver adaptively powering and communicating with an implant to optimize power transfer and control a drug delivery module (underneath). (b) Diagram of the drug release mechanism consisting of a voltage-controlled drug-loaded nanoparticulate film.

Our initial experiments revealed challenges in the wireless power and release control reliability. The received power was difficult to precisely control in vivo because it depended on implant depth and alignment which can change over the duration of the release, potentially making implant operation unpredictable. The release amount can also depend on environmental factors and the current generation of PPy NP electrodes vary sample to sample due to small variations in the fabrication process. Most drug delivery implants do not provide any feedback and programmability which is needed for closed-loop personalized release control. These issues can be addressed by incorporating more reconfigurability and feedback to make wireless implantable DDSs more versatile and reliable. In this work, we propose enhancing power transfer reliability by feeding back the implant energy level to adapt the transmitter power level and focus to ensure sufficient power is maintained during critical events. Furthermore, because PPy NP-mediated drug release is controlled through a redox reaction, we can sense and adjust the release rate using a potentiostat [25]. A potentiostat is usually used for electrochemical analysis and can maintain a voltage between a working and reference electrode (RE), by controlling and measuring the current between the working electrode (WE) and counter electrode (CE). By feeding back potentiostat current data and changing the potentiostat voltage, we can monitor and regulate drug release in real-time.

Here, we present a robust implantable DDS using US wireless power and communication links, power combining, an electrochemical drug release mechanism, and power/release feedback for closed-loop operation. This work is an extension of [19] and provides more details and new results related to the custom integrated circuit (IC) and implant system. In particular, we analyze and demonstrate techniques to increase power transfer and drug release reliability. In Section II we first present an overview of the DDS and implant components. Then we delve into the design of the circuit blocks with a focus on the power recovery and potentiostat circuits. In Section III we present electrical measurement results of the implant operation and circuit performance including adaptive wireless power transfer and programming. In Section IV we demonstrate an implantable form factor and in vitro drug release results both with and without release feedback using FL as a model drug. Finally, we conclude in Section V with a comparison to other wireless potentiostats and drug delivery implants as well as a discussion on how to further improve robustness in DDSs.

II. System Architecture

A. Implant Architecture

The closed-loop implant system block diagram is shown in Fig. 2(a) and contains three PZT4 piezoelectric transducers (piezos), an IC, an off-chip energy storage capacitor $C_{STORE}$, and a drug delivery module (DDM). Two piezos (PPZTs) are for receiving US power and downlink data at 1 MHz and the other (DPZT) is for transmitting on-off keying (OOK) modulated data at 2.5 MHz to avoid self-interference from the power harmonics [26]. The DDM includes a three-electrode system with a gold WE containing the FL-loaded PPy NPs, a silver RE, and a gold CE. The IC consists of an active rectifier, power management unit (PMU), downlink data recovery circuit to receive commands, relaxation oscillator for clocking, finite state machine (FSM) for controlling chip operation, potentiostat for controlling drug release and current sensing, analog-to-digital converter (ADC) to digitize $V_{RECT}$ or the potentiostat output, and OOK transmitter (TX) for data uplink.

Fig. 2.

(a) Block diagram of the closed-loop drug delivery implant system and custom IC. (b) System timing diagram showing power up, downlink communication, power feedback, potentiostat activation, and data uplink.

The PMU includes three low-dropout regulators (LDOs), which regulate the rectifier output ($V_{RECT}$) to 1.8 V, and a 0.5 V reference for the ADC. $V_{DD,D}$ supplies the data recovery, FSM, clock, and TX, $V_{DD,A}$ supplies the potentiostat, and $V_{DD,W}$ sets the WE voltage ($V_{WE}$). The 10-bit differential SAR ADC is adapted from [27] with a top-plate sampled, constant common-mode capacitive digital-to-analog converter (DAC). The relaxation oscillator is also adapted from [27] with increased current to generate a 5 MHz reference. The 5 MHz reference is used to generate the 1 MHz clocks for the FSM and ADC as well as the 2.5 MHz carrier frequency for uplink data.

Fig. 2(b) shows the timing diagram for typical implant operation. Operation starts once power is transmitted by the external transceiver (TRX), received by PPZT, rectified by the active rectifier, and stored on $C_{STORE}$. Once $V_{RECT}$ exceeds about 1.8 V, the bandgap reference (BGR), LDOs, and power-on reset (POR) start. In this state, the implant waits for a command to control the system and program the drug release. A command is formed from a pulse-position modulated (PPM) 36-bit packet that is amplitude-shift-keyed (ASK) onto the US power carrier. The command packet can activate the power feedback or potentiostat as well as set the potentiostat voltage, polarity, and gain. In Fig. 2(b), the first command activates the ADC and TX for power feedback, which allows for implant power monitoring and optimization of the power transfer shown by an increase in $V_{RECT}$. Once the power transfer has been optimized, the next command triggers drug release by activating the potentiostat, ADC, and TX. The potentiostat current is then recorded and fed back to monitor drug release. Uplink data is transmitted at 125 kbps on a 2.5 MHz carrier after each ADC sample. Additional commands can be sent at any time while the chip is powered to further optimize power transfer, program the potentiostat, and monitor/control drug release.

B. Drug Delivery Module

Dropsens screen printed electrodes (DRP-C220AT) were used to hold the fluorescein-loaded polypyrolle nanoparticles (FL-PPy NPs) [20]. A 3D-printed 300 μm tall PLA ring was placed on top of the WE to protect the nanoparticulate film. The ring was 3D-printed using a Micro Series M3D printer as a 6 mm outer diameter (OD), 5 mm inner diameter (ID), 300 μm tall ring with a 50 μm layer height. The ring was fused on the WE of the screen printed electrode by briefly heating the electrode to 215 ° C on a hot plate. All experiments with the DRP-C220AT electrodes in this work are run as two-electrode reactions with silver to silver chloride as the second half cell of the redox reaction.

Fluorescein at varying concentrations (17 wt%, 21 wt%, 25 wt%) was encapsulated within PPy NPs using a micelle-templated synthesis process [20]. To prepare FL-PPy NPs, 6 μL of pyrrole was added to 1 mL of a 0.1 M SDS solution in 40 mM HCl and stirred for 5 minutes. Subsequently, 24.9 μL of 14.2 μg/mL, 17.5 μg/mL, and 21 μg/mL aqueous solutions of fluorescein disodium salt were simultaneously introduced into the pyrrole solution to achieve 17 wt%, 21 wt%, and 25 wt% FL loading. After stirring for 30 minutes, 80 μL of an aqueous solution containing iron (III) chloride at a concentration of 625 mg/mL was added to the solution. The resulting mixture was stirred at room temperature for 24 hours overnight. The samples then underwent three rounds of dialysis, each lasting over 24 hours, using 300 mL of distilled water, followed by drying under vacuum. Next, the dried PPy NPs were dissolved in isopropanol to create a 5 mg/mL solution and then subjected to tip sonication for nanoparticle suspension. The suspension was aerosol spray coated onto the WE using a tape mask with a 4 mm diameter hole, at a flow rate of 5 μL/min through a silica capillary. A PDMS gel with silicone oil was prepared by mixing a precursor composed of a 1:10 ratio of potting compound to crosslinker, diluted 10 times with silicone oil. A 2 μL aliquot was placed within the ring encircling the WE. The electrodes were then subjected to a vacuum for 1.5 hours to eliminate any air bubbles. Following this, they were heated at 115 °C for another 1.5 hours to initiate gel cross-linking. Previous characterization of these DDMs in [20] showed an entrapment efficiency of 93.5%.

C. Power Recovery

The power recovery circuit rectifies the electrical signals from PPZT into $V_{RECT}$ which is used by the PMU to generate supplies for the rest of the chip. A full-wave active rectifier architecture is adapted from [28].

To optimize the power transfer, the received power needs to be measured and fed back to the external TRX. In this work, we use $V_{RECT}$ as a proxy for the received power because higher $V_{RECT}$ typically corresponds to higher available power, assuming the chip power consumption is consistent throughout the optimization duration. The sampling circuit schematic can be seen in Fig. 3(a). To fit into the ADC input range, a resistive divider ($R=1\mathrm{M}\Omega$) first divides $V_{RECT}$ by 4. Then a switched-capacitor (SC) circuit converts the single-ended signal into a differential signal centered around a common-mode voltage ($V_{CM}$) of 0.9 V. The signal is further divided by the capacitive divider between the SC circuit and the ADC capacitor DAC resulting in an overall division of $V_{RECT}∕10$ allowing for a maximum $V_{RECT}$ of 5 V with a resolution of about 10 mV. The measured voltage is then uplinked for external power transmitter adaptation. Note that the minimum voltage is limited by the minimum implant operating voltage since $V_{RECT}$ needs to be high enough to start up the implant.

Fig. 3.

(a) Schematic diagram of the $V_{RECT}$ sampling circuit. (b) Conceptual schematic diagram of (left) RF power combining and (right) DC power combining for an ultrasonically powered device.

In addition to power feedback, we propose power combining across multiple piezo power receivers to make the power transfer more robust. Power combining is a common technique in RF power transfer for increasing efficiency with rectenna arrays. Two main configurations can be used to merge the power to increase the received power and misalignment tolerance - RF and DC combining [29], [30]. In RF combining, the incoming signals are coherently combined before being delivered to a single rectifier (Fig. 3(b) left). In DC combining, each piezo has its own rectifier and the rectified DC power is combined (Fig. 3(b) right). RF combining can increase the signal power before the rectifier if the RF signals can combine in-phase, resulting in a higher achievable power than DC combining due to better rectifier efficiency. However, in-phase combining depends on the alignment of the external power transmitter with the implant power receivers. As the number of RF combining piezos increases, receiver phase shifting becomes infeasible and the piezos behave similarly to a single piezo of the same total area. The power receiver gain increases at normal incidence, and the beam pattern becomes narrower. While beneficial when perfectly aligned, as the incidence angle of the transmitted wave increases, the receiver gain and power will decrease significantly while gain and power for DC combining will remain unaffected. Hybrid techniques can also be used incorporating both RF and DC combining at different stages [30]. Here, we demonstrate and compare RF and DC combining with US wireless power across two piezos. To test RF combining, we connected two piezos in parallel to the input of the on-chip active rectifier. For DC combining, we connected each piezo to its own rectifier and connected the outputs of the rectifiers together.

D. Downlink Data Recovery

The ASK-PPM downlink scheme enables simultaneous power and data transfer up to 50 kbps, supporting continuous programming and power needed for drug release. Fig. 4 depicts schematics for the ASK demodulator, PPM clock and data recovery (CDR), and the timing diagram. The protocol and architecture is similar to [31]. The diode-connected PMOS with active body biasing ($D_1$) in the ASK demodulator passes an envelope of the received US signal ($V_{PPZT+}$) to $C_1$. The load current ($I_1$) creates a discharge path for $C_1$. The low-pass filtered (LPF) US power envelope ($V_{ENV}$) is compared to a 0.9 V reference voltage ($V_{REF}$) with comparator $A_1$, generating the PPM signal ($V_{PPM}$). In the PPM-CDR circuit, $V_{PPM}$ is converted to a clock signal ($V_{CLK}$) via a risingedge-triggered frequency divider ($DF{F}_1$). $V_{CLK}$ controls the timing of a sawtooth generator by charging and discharging $C_2$ through current sources $I_2$ and $I_3$, respectively. When the time between the first and second pulses compared to the second and third pulses is approximately 7:3, the inverted comparator ($A_2$) output ($V_{CMP}$) goes high because $V_{SAW}$ exceeds $V_{REF}$. $V_{CL{K}_B}$ then clocks in $V_{DATA}=1$ through $DF{F}_2$. If the time between the first and second pulses is shorter, $V_{CL{K}_B}$ clocks in $V_{DATA}=0$.

Fig. 4.

Downlink data recovery schematic and timing diagram (adapted from [19]).

E. Potentiostat

A potentiostat is a device that maintains a voltage across the WE and RE while measuring a current between the WE and CE. Potentiostats are typically used for electrochemical analysis. The design of a potentiostat for drug release has different specifications than one designed purely for electrochemical sensing. The first difference can be understood using the electrochemical cell equivalent circuit seen in Fig. 5. Drug loading is directly proportional to the size of the WE, which is proportional to the double-layer capacitance ($C_{DL}$). Here, we are using a WE with 4 mm diameter, corresponding to a $C_{DL}$ in the 10 μF range, which is orders of magnitude larger than most other integrated potentiostat sensing systems [32]. The large $C_{DL}$ increases the integrated noise current and can affect the control amplifier (CA) loop stability [32], [33].

Fig. 5.

Potentiostat schematic and electrochemical cell equivalent circuit (adapted from [19]).

Drug release voltages depend both on the PPy NP formulation and the drug itself. To support as broad an application space as possible, the potentiostat should be able to handle both positive and negative WE-RE potentials ($V_{WE-RE}$). In addition, because the current is correlated with drug release rate, the potentiostat needs to be able to source and sink currents on the order of 100 μA, corresponding to an $R_{CT}$ down to $\sim 10\mathrm{k}\Omega$. For sensing applications, electrode size and $C_{DL}$ is usually minimized meaning that $R_{CT}$ can be on the order of $\mathrm{M}\Omega$ to $\mathrm{G}\Omega$ and the noise current is in the pA region increasing the gain and noise requirements, but decreasing the current handling requirements.

Fig. 5 shows the potentiostat schematic. The current readout uses a current mirror topology [33] with switches that control the polarity of $V_{WE-RE}$ as well as the current direction. The WE is connected to $V_{DD,W}$ for positive $V_{WE-RE}$ and ground for negative $V_{WE-RE}$. $V_{RE}$ is set by an 8-bit R-2R DAC and a feedback loop with the CA and $M_{1∕2}$. A transimpedance amplifier (TIA) with resistive feedback ($R_{TIA}=2.5\mathrm{k}\Omega$) is used to convert the potentiostat current to a voltage. To mitigate static and dynamic errors in the current mirroring due to the large current and voltage ranges, the TIA inputs are connected to the drains of $M_{1∕2}$ and $M_{3∕4}$ which force their voltages to match. This keeps the current mirroring accurate without the need for cascode devices which reduce headroom and the achievable $V_{WE-RE}$. Using this architecture, the potentiostat can achieve a $V_{WE-RE}$ range of at least −1.5 V to 1.5 V.

Following the TIA, the current readout is further amplified, filtered, and converted into a differential signal by a SC variable gain amplifier (VGA). SC filtering is used to generate an anti-aliasing LPF before the ADC. The gain is set by $C_3∕C_4$ and the LPF cutoff frequency is set by $C_4$ and $C_5$. There are four gain settings set by two bits in the downlink packet which change $C_3$. The four gain settings allow for full-scale ranges of ±100 μA, ±50 μA, ±20 μA, and ±6 μA. Finally, the differential 10-bit SAR ADC digitizes the signal, which is then transmitted back to the external TRX. The sample rate of the ADC is 400 Hz with a corresponding SC LPF frequency of 100 Hz. A summary of the specifications can be seen in Table I.

TABLE I.

Summary of Potentiostat Specifications

Parameter	Specifications
Electrode Area	12.6 mm2
Electrode Capacitance	~10 μF
Current Range	±6, ±20, ±50, ±100 μA
Voltage Range	±1.5 V
Sample Rate	400 Hz
Bandwidth	100 Hz

Fig. 6 shows the wide-input-swing folded cascode operational transconductance amplifier (OTA) used as the core amplifier in the CA, buffer, TIA, and VGA (without current mirror). The complementary input transistors are used to achieve near rail to rail input range and the folded cascode architecture ensures high gain. The TIA adds an additional common-source output stage to accommodate the feedback resistor load.

Fig. 6.

Schematic diagram of the wide-input-swing folded cascode OTA used in the CA, TIA, buffer, and VGA (without current mirroring).

III. Circuit Measurement Results

The system was first tested with the piezos in a mineral oil tank to mimic the acoustic properties of tissue and the electronics on a separate PCB outside of the tank to allow for probing. These initial measurements were used to characterized the individual circuit blocks as well as the wireless power and communication links. Once basic functionality was confirmed, the power feedback, optimization, and combining were also demonstrated to show increased robustness. After initial electrical measurements, the electronics were packaged into a millimeter-sized footprint on one PCB which included the piezos, peripherals, and IC. In vitro wireless FL release experiments were then run with the packaged device in the oil tank. Release experiments were run both open-loop and closed-loop to show the benefits of release sensing and programmable release control for more consistent drug delivery.

Fig. 7 shows the chip micrograph and the circuit measurement setup. The 2 mm × 2 mm IC was fabricated in a TSMC 180 nm HV BCD process. Wireless powering was measured with PPZT and a single-element, unfocused US transmitter (Olympus A303S) placed 8 cm apart in a mineral oil tank. The acoustic propagation efficiency of this setup at 1 MHz was simulated with Field II to be about 1.1% [34], [35]. A US receiver (Sonic Concepts A-102) and pre-amplifier (PA-107) were used to receive data sent by DPZT. PPZT and DPZT were connected to an external PCB containing the IC, $C_{STORE}$, and equivalent lumped circuit models for the electrodes. Wireless startup and characterization of the potentiostat was done with $C_{STORE}=10\mu \mathrm{F}$, $C_{DL}=10\mu \mathrm{F}$, $R_{CT}=100\mathrm{k}\Omega$, and $R_S=30\Omega$.

Fig. 7.

Measurement setup for electrical characterization and chip micrograph.

Fig. 8 shows the measured wireless startup waveforms, activation of the potentiostat, and data feedback. As US power is sent, $V_{RECT}$ (blue) starts charging. Once $V_{RECT}$ charges enough, the LDOs start and $V_{DD,D}$ (orange) stabilizes at 1.8 V. After about 30 ms, the potentiostat activates and $V_{WE-RE}$ (yellow) goes to 0.85 V as set by the DAC. $V_{RECT}$ drops then recovers due to the large instantaneous current needed to charge $C_{DL}$. Uplink packets are observed at a rate of approximately 400 Hz. A zoomed-in view of one packet can be seen in the bottom two plots, showing the OOK modulated signal. The middle plot shows $V_{DPZT+}$ (purple) which is the OOK TX output. The bottom plot shows $V_{RX}$ (green) which is the signal recovered by the external receiver. An on-chip 17-bit pseudorandom binary sequence (PRBS) was used to test the uplink and showed a bit error rate (BER) of less than 10⁻⁵.

Fig. 8.

Measured wireless startup waveforms with potentiostat and uplink enabled.

Downlink data recovery was tested by modulating the power signal envelope and verifying that the correct bits were programmed. During these tests, we discovered a design error in comparator $A_1$. The comparator did not include hysteresis which meant that if $V_{ENV}$ crossed $V_{REF}$ too slowly, the output of $A_1$ would bounce. Since each transition of $A_1$ would switch the charge pump, any spurious transitions would lead to errors that propagated through the rest of the packet. To workaround this error, we added a few off-chip components to replicate the on-chip envelope detection as can be seen in Fig. 9(a). The envelope, $V_{ENV2}$, was generated using a half-wave rectifier and the reference, $V_{REF2}$, was generated by dividing and low-pass filtering $V_{ENV2}$ so that it would track the steady state value of $V_{ENV2}$. These two signals were fed into a low-power, off-the-shelf hysteretic comparator (TLV7042, $A_1^{\prime }$) to avoid the output bouncing issue. Since $V_{ENV}$ is accessible from off-chip, the on-chip envelope detector was left unconnected and the output of $A_1^{\prime }$ was connected to $V_{ENV}$. Because the output of $A_1^{\prime }$ is a fast digital signal, there is no issue with the output of $A_1$ bouncing. Note that the additional components are powered from $V_{DD,D}$ so the system is still wirelessly powered.

Fig. 9.

(a) Schematic of the downlink data recovery workaround. (b) Measured downlink waveforms.

The downlink waveforms were also measured using the 8 cm mineral oil link (Fig. 9(b)). The ASK-PPM signals (blue) were encoded with a 50% modulation depth. Pre-emphasis was used after each negative pulse to speed up the envelope recovery between closely spaced pulses. The middle plot shows $V_{ENV2}$ (yellow) and $V_{REF2}$ (orange) derived from the transmitted waveforms after propagating 8 cm in the tank. Each time there is a negative pulse, the output of $A_1$ ($V_{PPM}$, purple) pulses and the demodulated bits can be seen in green. In this case, the end of the packet is shown and after the last bit, the potentiostat is turned on (light blue).

Power feedback can be used to optimize wireless power transfer. The rectifier voltage, which is digitized and fed back to the external TRX, is used as an estimate for received power because higher $V_{RECT}$ corresponds to higher receiver piezo voltages. Fig. 10(a) shows the measured ADC code over the $V_{RECT}$ range. The measured $V_{RECT}$ can be used to better align the external power transmitter with the implant power receiver. We propose a powering scheme that first sweeps the power beam until the implant can just turn on. Then we use an iterative gradient ascent algorithm to steer the beam to the optimal location. For example, if this particular transmitter is 4.5 mm off from the power receiver, $V_{RECT}$ is just high enough to turn the chip on. But, if power consumption increases or the beam shifts a single millimeter, there may not be enough available power for the implant to remain functional. By using a gradient ascent algorithm, the transmitter beam can adapt can until $V_{RECT}$ is maximized, giving more margin for operation and making the system more robust to changes in alignment over time. Fig. 10(b) shows the path of the transmitter along with the associated $V_{RECT}$ over the optimization process. The transmitter location is updated according to

$$x_{TX,n+1}=x_{TX,n}+\gamma \nabla V_{RECT}(x_{TX,n})$$ (1)

where $x_{TX,n}$ is the transmitter location at step $n$ and $\gamma$ is the learning rate. In the first optimization run (blue) it took six steps to find the optimal transmitter location and increase $V_{RECT}$ to about 4.4 V. Based on the required received piezo voltage, this corresponded to an increase in available power from 0.47 mW to 3.2 mW, a 6.8× improvement from optimizing the transmitter location.

Fig. 10.

(a) Measured ADC codes over the full range of $V_{RECT}$ which is fed back for power optimization. (b) Path of the external power transmitter to optimize $V_{RECT}$ using a gradient ascent algorithm. Each step of the iteration is labeled (1) through (6). (c) Transmitter voltage and $V_{RECT}$ over the course of the optimization. After finding optimal location, $V_{TX}$ is decreased until $V_{RECT}$ settles around 2.5 V.

The power feedback does not just help to optimize the transmitter location. It can also be used to adjust the transmit power to regulate $V_{RECT}$ and reduce overall system power consumption. For example, after we found the optimal location, $V_{RECT}$ may be much higher than needed. We can now reduce the transmitter power to decrease the steady state $V_{RECT}$ to 2.5 V. This will lower the power consumption on both the implant and the external power transmitter. This process can be seen in Fig. 10(c). After optimizing the transmitter location, the transmitter voltage was lowered until $V_{RECT}$ dropped to about 2.5 V, ultimately resulting in a 54% reduction in transmit power. Depending on the situation, the transmitter power could be increased again or further reduced if needed.

Power combining is another technique that can make power transfer more robust. To test and compare RF and DC power combining we first directly drove the on-chip rectifier(s) at 1 MHz with two signal generators with $3\mathrm{k}\Omega$ source resistance to model the piezo signals (Fig. 11). In Fig. 11(a) we can see that if the two signal generators are in-phase then $V_{RECT}$ is higher for RF combining than for DC combining across the input amplitude range. In either case, $V_{RECT}$ is higher with two channels than a single channel showing that power combining can increase the received power. We can also see how the power changes with phase shifts between the two signals in Fig. 11(b). Until about a phase difference of 60°, $V_{RECT}$ for RF combining is higher than $V_{RECT}$ for DC combining. However, as the phase difference increases further (corresponding to larger transmitter angles), the power from RF combining continues to decrease until the signals perfectly cancel at 180° phase difference. Throughout the phase range, $V_{RECT}$ for DC combining remains unchanged. Finally, we verified these findings by connecting two piezos to the chip (either in parallel to one rectifier or to two separate rectifiers) at slightly varying heights. The different heights allowed for large phase differences with negligible difference in receiver amplitude. Since the US wavelength at 1 MHz in oil is about 1.45 mm, perfect cancellation should occur with about 725 μm offset. When wirelessly powered with US, the measured $V_{RECT}$ in Fig. 11(c) across three scenarios shows that DC combining can maintain consistent power regardless of piezo phase difference. Overall, these results show that power combining can increase both the power received by the implant as well as the misalignment tolerance. In this case, since we want to maximize misalignment tolerance we used the DC combining configuration.

Fig. 11.

(a) Measured $V_{RECT}$ as input voltage amplitude increases for a single input (black) and two in-phase inputs RF combined (blue) or DC combined (orange). (b) Measured $V_{RECT}$ as the phase difference between the two inputs increases for RF combining (blue) and DC combining (orange). (c) Measured $V_{RECT}$ during wireless powering of two piezos at different heights for RF combining (blue) and DC combining (orange).

The potentiostat was first tested by programming a sequence of different $V_{WE-RE}$. Fig. 12(a) shows the potentiostat voltage switching according to the wirelessly programmed commands. Fig. 12(b) shows the measured ADC output across the input current range for all gain settings. The ADC data were demodulated from the uplink packets and show that the full scale range is as expected for all gain settings. The input-referred current noise was measured to be 4.1 ${\text{nA}}_{\text{rms}}$ across a 1 Hz bandwidth. The differential nonlinearity (DNL) and integral nonlinearity (INL) were measured using the histogram method and an input current ramp (Fig. 12(c)). The overall potentiostat DNL and INL were +0.38/−0.47 and +0.77/−0.49 LSB respectively.

Fig. 12.

(a) Measured wirelessly programmed potentiostat voltage pattern. (b) Characterization of the potentiostat current readout for all gain settings. (c) Measured DNL and INL of the potentiostat front-end.

The total average power consumption was simulated to be 512 μW at a $V_{RECT}$ of 2 V. The power breakdown can be seen in Fig. 13. Table II provides a comparison with other state-of-the-art wireless potentiostats. Most other potentiostats are used solely for electrochemical sensing so the application requirements differ significantly. In this work, we designed for a much larger electrode area (corresponding to higher $C_{DL}$) and current range since drug loading and release increase with electrode size and current at the expense of higher input-referred current noise. The demonstrated operating depth is significantly larger than other works, enabled by the advantages of US power transfer as well as the use of power combining and power feedback.

Fig. 13.

Breakdown of the 512 μW simulated chip average power consumption at a $V_{RECT}$ of 2 V.

TABLE II.

Comparison with State-of-the-Art Wireless Potentiostats

	This Work	[36]	[37]	[38]	[39]	[40]	[41]
Application	Drug Delivery	Glucose/Lactate Sensing	Aptamer-based Sensing	Chemical Sensing	Chemical Sensing	Alcohol Sensing	Glucose Sensing
Technology	180 nm	40 nm	65 nm	350 nm	180 nm	65 nm	130 nm
Power Source	US	Battery	US	N/S	RF	RF Near-Field	RF Near-Field
Downlink	ASK-PPM <50 kbps	N/A	OOK-PWM 2.5 kbps	N/A	N/A	N/A	ASK 26.5 kbps
Uplink	OOK 125 kbps	OOK N/S	QPSK 625 kbps	PHM Analog	N/A	LSK Analog	LSK N/S
Distance	8 cm mineral oil	N/S	3 cm mineral oil	1 cm air	N/S	2.4 mm pork	3 cm air
Electrode Area (mm2)	12.6	N/S	0.25	N/S	N/S	0.025	0.15
IRN	$4.1{\text{nA}}_{\text{rms}}$	$14{\text{pA}}_{\text{rms}}$	$1.6{\text{nA}}_{\text{rms}}$	$10{\text{pA}}_{\text{rms}}$	$3.3{\text{pA}}_{\text{rms}}$	$1.2{\text{nA}}_{\text{rms}}$	N/S
Current Range (μA)	±100	0.12	±0.8	±1	0.8	0.1	0.04
System Power	512 μW	100 μW	6.6 mW	12 μW	4.4 μW	0.97 μW	50 μW
Dimensions (mm)	5.5×18×2*	14×9×0.8	3.6×13×3	Not packaged	Not packaged	0.85×1.5	5×17

^*Would be 5.5 × 9.2 × 2 mm³ without downlink workaround, electrodes would be underneath board and not change dimensions

N/A - not applicable, N/S - not specified, PWM - pulse width modulation, QPSK - quadrature phase shift keying, PHM - pulse harmonic modulation, LSK - load shift keying

IV. Drug Release Results

For in vitro drug release experiments, the electronics were packaged into an implantable form factor (Fig. 14). The compact version included two power piezos for DC power combining, a data piezo for uplink, the IC, and storage capacitors on the left half of the board. The additional circuits needed for downlink data recovery, including the envelope detector and comparator, were placed on the right half. The dimensions of the implant are 5.5 mm × 18 mm × 2 mm. Without the additional downlink workaround components, the longest dimension would have shrunk to 9.2 mm.

Fig. 14.

Measurement setup for in vitro release experiments with the compact form factor board. The DDM electrodes were connected separately to allow for fluorescein release measurements and larger release sample size. Closed-loop control was accomplished by demodulating RX data, reprogramming the AWG, and transmitting data to tune the potentiostat voltage.

The implant board was placed into the mineral oil tank and connected to the DDM electrodes externally. Note that the board was not encapsulated due to this external connection which allowed for more accurate drug release measurements and larger drug release sample size. There were no components placed on the bottom side of the board to allow for future integration of the DDM electrodes underneath the board as shown in Fig. 1(a). Future developments for in vivo applications could follow previously demonstrated ultrasonic implants and use a combination of PDMS, medical grade epoxy, and Parylene-C encapsulation [20], [42], [43]. Similar to the previous setup, the US transmitter and receiver were placed 8 cm away, but in this setup the RX data was demodulated and the AWG was reprogrammed in real-time for closed-loop tests. During the experiments, the DDMs were placed in 20 mL Ringer’s solution heated to 37°C. FL release was controlled by the potentiostat over a 10 min release duration and sampled at 0, 1, 2, 3, 5, 8, and 10 min. The FL samples were analyzed using a fluorometer to measure the amount released at each timepoint.

The first test used three NP formulations using different concentrations of FL (17%, 21%, and 25%). The different concentrations introduced variability in the release amounts as seen in Fig. 15(a). Even with the same stimulation voltage of −1 V, the release after 10 min varied from about 1.2 μg for 17% FL to 2.5 μg for 25% FL. Note that the results used electrodes fabricated across different days and batches. As a result, the error bars for each concentration also show large standard deviations (SD) across batches. This means that even small variations in the preparation could cause significant differences in release which is undesirable for precise drug dosing needed for certain applications.

Fig. 15.

Measured FL release across (mean ± SD) (a) different FL concentrations across multiple batches for a −1 V stimulus voltage (N = 6/group) and (b) release kinetics for 21% FL from one batch at different voltages (N = 3/group). (c) Release kinetics from a constant and pulsed stimulus.

The FL concentration in the NPs is not the only degree of freedom in controlling the release amount. The stimulation voltage can also affect the release amount [44]. In Fig. 15(b) we can see that more negative voltages increased the release amount. The release rate was highest in the first minute and slowed down as more FL was expelled into the gel and less FL remained embedded in the PPy NPs. Finally, if we change the stimulation voltage in real-time, we can change the release profile. As an example, we pulsed between negative and positive voltages at 2 min intervals, and observed a staggered release profile compared to a constant −1 V stimulus (Fig. 15(c)). To equalize the release amount in each pulse, the first pulse was at −0.8 V and the second pulse was at −0.9 V to account for the different initial conditions.

The potentiostat current sensing data also changes with the applied voltage with larger currents corresponding to higher release amounts. This is most easily seen in Fig. 16(a), which shows the results from multiple releases across different FL concentrations and stimulation voltages. We found that the integrated current over the first 2 min of the release correlated best with the total FL release with an $R^2$ of 0.864. In this case, the charge after 2 min was used because the relationship between the measured current and release is complicated by the silicone oil-PDMS gel layer placed over the NPs to prevent leakage. Because of the gel layer, once the FL is released from the PPy NPs after reduction, it still needs to travel through the gel into the Ringer’s solution to be measured. This delay is dictated by the electrophoretic forces exerted on the FL ions from the electric fields in the DDM and will be investigated further in the future. Nonetheless, there is a clear relationship between the measured potentiostat current and release amount.

Fig. 16.

(a) Measured integrated current over the first 2 min of stimulation and FL release amount after 10 min from releases using different voltages and FL concentrations. (b) Closed-loop control of the FL release by adapting potentiostat voltage to more closely match reference current (black). (c) Measured FL release across different FL concentrations without and with closed-loop adaptation showing equalized release (${\mathrm{N}}_{\text{CL}}=6∕\text{group}$).

Based on our findings about the relationship between the voltage, release rate, and current we can potentially make the release more consistent by using the feedback current data to wirelessly program the implant and adapt the stimulus voltage. We tested closed-loop release using the measured current waveform from a −1 V stimulus, 2 μg total release sample as a reference (Fig. 16(b), black). For each subsequent electrode, the stimulus started at −1 V and the measured current was compared to the reference waveform. If the current was measured to be much less than the reference current, signifying slower release, the potentiostat voltage was programmed to be more negative (blue) to increase release rate. If the current was much greater than the reference, signifying faster release, the voltage was programmed to be less negative (orange) to decrease release rate. The magnitude of the voltage change was proportional to the difference in current at each time point. In this case, the voltage was updated at 30 sec intervals in multiples of 0.1 V, though more frequent and gradual adaptation could also be used. We hypothesized that by adapting the stimulus voltage in this way, the release of all the electrodes would be closer to the 2 μg target regardless of starting FL-PPy NP formulation.

Fig. 16(c) shows the results of closed-loop release control. The adaptation equalized the release amount, which previously varied from 1.2 μg to 2.5 μg (blue bars), to about 2 μg (orange bars) across different FL concentrations. This means that even if there are large differences in the DDMs or environmental factors affecting the release rate, the release could be adapted by measuring the redox current and tuning the stimulation voltage. In addition, this adaptation could even reduce the variation within each concentration across the electrodes. For each concentration, the standard deviation decreased when using the closed-loop control compared to the constant −1 V stimulus to an average of 0.6 μg, corresponding to a 39% reduction. This can potentially be further improved using a more frequent and fine-tuned adaptation. By adding this feedback and adapting the stimulus, we could improve the reliability and consistency of the release from the PPy NPs. Table III provides a comparison with other state-of-the-art wireless drug delivery implants. Similar to above, the demonstrated operating depth is larger than most other works while also adding current-based release sensing, power combining, and voltage feedback for better reliability.

TABLE III.

Comparison with State-of-the-Art Wireless Drug Delivery Implants

	This Work	[14]	[7]	[16]	[17]	[6]	[8]
Mechanism	Electrochem. nanoparticles	Electrolysis	Electrolysis	Crevice corrosion	Electrolysis	Electrotherm. membrane	Piezoelectric valve
Power Source	US	RF Near-Field	RF Near-Field	RF Near-Field	RF Near-Field	Battery	RF Near-Field
Distance	8 cm mineral oil	17 cm air*	2 mm tissue	2 cm air	2 cm tissue phantom	N/S	2 mm air
Communication	Bidirectional	Bidirectional	N/A	N/A	Uplink	Bidirectional	N/A
Power Features	Rect. volt. feedback/Power combining	N/A	LED indicator	N/A	N/A	Battery volt. feedback	N/A
Release Control	Digital code	Digital code	Receiver coil volt.	Receiver coil volt.	Receiver coil volt.	Digital code	Receiver coil volt.
Release Sensing	Current	N/A**	LED indicator	N/A	Reservoir conductivity	Dose delivery confirmation	N/A
Dimensions (mm)	5.5×18×2	40×18×3	5×25×2	21×21×5***	20 diam.×10***	54×31×11	42×22×4

^*Distance to center of power transmitter coil (wrapped around 31 cm × 34 cm cage)

^**Neural recording features trigger release

^***Estimated from picture of device

N/A - not applicable, N/S - not specified

V. Conclusion

Next-generation implantable DDSs will require more adaptability to operate reliably especially as the systems scale and become more complex. Previous demonstrations of the feasibility of US-powered on-demand drug release revealed challenges related to power alignment and release consistency which are extremely important for reliable drug delivery. We designed a custom IC and implant which included wireless power combining, power sensing, a potentiostat, and bidirectional communication to address these challenges. By closing the loop on the power transfer and drug release, we demonstrated that we could make the overall system more robust.

In this work, we used commercial single-element US transducers as the external TRX to demonstrate the benefits of power and release feedback. Development of a portable phased array TRX would enhance peak power transfer efficiency and also enable long-term in vivo characterization to advance this DDS [45], [46]. Continued investigation is also needed to better understand the relationship between the measured current and FL release in our DDM. A better understanding combined with a more sophisticated and continuous control algorithm could further improve release reliability. While we focused on a particular drug release mechanism, this system could also be used with other electrochemical release mechanisms or other sensing applications, offering additional opportunities for feedback and release control.

Biographies

Max L. Wang (Member, IEEE) received the B.S. degree in electrical engineering from the California Institute of Technology (Caltech), Pasadena, CA, USA, in 2015, and the M.S. and Ph.D. degrees in electrical engineering from Stanford University, Stanford, CA, USA, in 2017 and 2022 respectively.

From 2014 to 2015, he worked at Caltech and contributed to the design of a radiative wireless power transfer system. Since 2022, he has worked at SiLC Technologies, Monrovia, CA, USA, where he has been developing integrated circuits and systems for frequency-modulated continuous-wave LiDARs. His current research interests include analog/mixed-signal/RF integrated circuit systems, imaging/sensor technologies, wireless power transfer, and biomedical devices.

Dr. Wang was a recipient of the Dr. Robert Noyce Stanford Graduate Fellowship and the National Science Foundation Graduate Research Fellowship.

Pyungwoo Yeon (Member, IEEE) received the B.S. degree in electrical and computer engineering from Seoul National University, Seoul, Korea, in 2010, the M.S. degree in electrical engineering and information systems from the University of Tokyo, Tokyo, Japan, in 2012, and the Ph.D. degree in electrical and computer engineering from the Georgia Institute of Technology, Atlanta, GA, USA, in 2019.

From 2012 to 2014, he was with the Power Management IC Team at Samsung Electronics, Yongin, Korea, where he contributed to prototyping an A4WP (currently AirFuel)-compatible wireless charger IC. He spent the first quarter of 2019 with the Display EE Team, Apple Inc., Cupertino, CA, USA. In Sep. 2019, he joined Stanford University as a post-doctoral fellow in electrical engineering. From 2021, he has been a Sensor Design Engineer at Apple Inc. His research interests include novel system architectures and circuit topologies for wireless implantable, wearable, and IoT medical therapeutics/diagnostics.

Dr. Yeon was the co-recipient of the IEEE Wireless Power Transfer Conference (WPTC) Best Paper Award in 2013, the Silver Award in the Samsung Electro-Mechanics Best Paper in 2016, the 3^rd IEEE Biomedical Circuits and Systems Conference (BioCAS) Best Paper Award in 2017, the Stanford RISE COVID-19 Crisis Response Trainee Seed Grant Program Award in 2020, the IEEE Solid-State Circuit Society (SSCS)-Brain Best Paper Award in 2021, the Stanford Bases Entrepreneurial Pitch Competition Winner on Bio and Health Section in 2023, and the 26^th Korean MEMS Conference Best Paper Award in 2024.

Mohmmad Mofidfar received the Ph.D. degree in Macromolecular Science and Engineering from Case Western Reserve University, Cleveland, OH, USA, in 2016.

From 2016 to 2019, he conducted his postdoctoral research in the School of Chemical and Biomolecular Engineering at Georgia Tech, Atlanta, GA, USA. In September 2020, he joined Stanford University, Stanford, CA, USA, as a Research Scientist in the Department of Chemistry. His current research, under the guidance of Professor Richard N. Zare at Stanford University focuses on elucidating the mechanism and formation of hydrogen peroxide (H₂O₂) within water microdroplets, profiling metabolites and exposomes in a silicone wristband, and developing wireless power delivery implants.

Dr. Mofidfar was a recipient of the Outstanding Graduate Students Award, CLiPS National Science Foundation’s Science & Technology Center in 2016; UCLA Clinical and Translational Science Poster Award in 2018; and Best Poster Award at the 5^th Annual Georgia Tech Postdoctoral Research Symposium in 2018.

Christian Chamberlayne received the B.S. degree in chemistry from the College of William and Mary in 2015, Williamsburg, VA, USA and the Ph.D. degree in chemistry from Stanford University, Stanford, CA, USA in 2021.

From 2021 to 2022 he continued as a postdoctoral researcher with his Ph.D. advisor Professor Richard Zare at Stanford University. He is currently a Systems Development Engineer at Picarro Inc., Santa Clara, CA, USA, working on laser cavity ringdown spectroscopy instrumentation for air quality monitoring. His research interests include conductive polymer nanoparticles, microdroplet chemistry, and laser spectroscopy.

Dr. Chamberlayne was the recipient of Stanford’s Linus Pauling Teaching Award, the Center for Molecular Analysis and Design (CMAD) Fellowship, and the Joseph, Linda, and Mark Keegan Graduate Fellowship from Stanford University.

Haixia Xu received the M.D. degree from Xi’an Jiaotong University, China, in 1998 and postdoctoral training from 2009-2011 in Hospital for Special Surgery, New York, USA.

She is a former Endocrinologist with 15 years of clinical experience at the Third Affiliated Hospital of Sun Yat-sen University, China, where she specialized in the management of diabetic patients, particularly those requiring intensive insulin therapy. She is currently a Basic Life Research Scientist at Stanford University, where she has been working since 2016. Her research focuses on innovative approaches to diabetes treatment, particularly in small molecule discovery. This includes developing targeted drug delivery strategies for islet β-cell replication and novel therapies to rescue insulin-induced hypoglycemia.

Dr. Xu was a recipient of the Young Scientist Award from the National Natural Science Foundation of China and the Ho Tim–Stanley Ho–Li Ka Shing Fellowship from Stanford University.

Justin P. Annes received the B.S. degree at Haverford College in 1996 and the M.D. Ph.D. degrees at New York University School of Medicine in 2004. During his Ph.D. studies, he researched the regulation of Transforming Growth Factor-β (TGF-β) activity under the mentorship of Dr. Daniel Rifkin.

He trained in Internal Medicine and Clinical Genetics at Brigham and Women’s Hospital (BWH) and Harvard Medical School from 2004 to 2009. His post-doctoral studies, at Harvard University, investigated the molecular pathways that control islet β-cell growth. Concurrently, he developed clinical Neuroendocrine Tumor (NETs) programs at BWH and the Dana-Farber Cancer Institute, as a Harvard Medical School Instructor from 2009 to 2012. Dr. Annes joined Stanford University in 2012, where his laboratory focuses on developing regenerative therapies for diabetes and understanding the tumorigenic mechanism of SDHx mutation-dependent cancers. His scientific contributions lie at the crossroads of endocrine pathophysiology, synthetic chemistry, and hereditary diseases. Clinically, Dr. Annes leads Stanford’s Hereditary Endocrine Neoplasia Clinic within the Endocrine Oncology Cancer Program.

Richard N. Zare received his B.A. degree from Harvard University in 1961 with a double major in chemistry and physics and his Ph.D. degree in chemical physics from Harvard University in 1964.

He is presently the Marguerite Blake Wilbur Professor in Natural Science, Department of Chemistry and Department of Physics (by courtesy), Stanford University. In his long career, he has received the National Medal of Science in 1983, the National Academy of Sciences Award in Chemical Sciences in 1991, the Wolf Prize in Chemistry in 2005, the Presidential Award for Excellence in Science, Mathematics, and Engineering Mentoring (PAESMEM), U.S. Office of Science and Technology Policy in 2009, the Priestley Medal of the American Chemical Society in 2010, the BBVA Foundation Frontiers of Knowledge Award in the Basic Sciences category in 2010, the King Faisal International Prize in Science, King Faisal Foundation in 2011, the Othmer Gold Medal from the Chemical Heritage Foundation in 2017, and the Benjamin Franklin Medal in Chemistry from the Franklin Institute in 2023.

Amin Arbabian (Senior Member, IEEE) received the Ph.D. degree in electrical engineering and computer science from the University of California at Berkeley, Berkeley, CA, USA, in 2011. From 2007 to 2008, he was a part of the Initial Engineering Team at Tagarray, Inc., Palo Alto, CA, USA (now acquired by Maxim Integrated Inc., San Jose, CA, USA). In 2010, he joined Qualcomm’s Corporate Research and Development Division, San Diego, CA, USA, where he designed circuits for nextgeneration ultra-low power wireless transceivers. In 2012, he joined Stanford University, Stanford, CA, USA, where he is currently an Associate Professor of electrical engineering and a faculty co-director of the Stanford SystemX Alliance (an industry affiliates program). He is the co-founder of Plato Systems, building a new Spatial Intelligence platform for the digital transformation of physical industries. His current technical interests include multimodality sensing and perception systems, mm-wave and high-frequency circuits and systems, the Internet-of-Everything devices, and medical implants. His group has worked on new sensing systems for programs under DARPA, ONR, NSF, DOE/ARPA-E, and NIH.

Dr. Arbabian was a recipient or a co-recipient of the 2020 IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS Best Paper Award; the 2016 Stanford University Tau Beta Pi Award for Excellence in Undergraduate Teaching; the 2015 NSF Faculty Early Career Development Program (CAREER) Award; the 2014 Defense Advanced Research Projects Agency (DARPA) Young Faculty Award (including the Director’s Fellowship in 2016); the 2013 Hellman Faculty Scholarship; the 2010–2011, 2014–2015, and 2016–2017 Qualcomm Innovation Fellowships; and best paper awards at the 2017 IEEE Biomedical Circuits and Systems Conference; the 2016 IEEE Conference on Biomedical Wireless Technologies, Networks, and Sensing Systems; the 2014 IEEE VLSI Circuits Symposium; the 2013 IEEE International Conference on Ultra-Wideband; the 2010 IEEE Jack Kilby Award for Outstanding Student Paper at the International Solid State Circuits Conference; and two-time second place best student paper awards at the 2008 and 2011 Radio Frequency Integrated Circuits (RFIC) Symposiums. He currently serves on the steering committee for the RFIC Symposium, the technical program committees of the RFIC Symposium and the VLSI Circuits Symposium, and as an Associate Editor for IEEE SOLID-STATE CIRCUITS LETTERS and the IEEE JOURNAL OF ELECTROMAGNETICS, RF AND MICROWAVES IN MEDICINE AND BIOLOGY.