Application of a sub–0.1-mm³ implantable mote for in vivo real-time wireless temperature sensing

Design of the monolithic, ultrasonic mote

The mote consists of a temperature sensor IC manufactured in a 0.18-μm bulk CMOS technology monolithically integrated (35) with a microscale piezoelectric transducer made of lead zirconate titanate (PZT) (36). Free from any bonding wires, solder joints, and circuit boards, this 0.3-mg mote measures only 380 μm by 300 μm by 570 μm (0.065 mm³) with direct mechanical and electrical connections between the transducer (300 μm by 300 μm by 267 μm) and the IC. A deconstructed view of the mote is shown in Fig. 1A. The CMOS IC contains two exposed aluminum (Al) pads connected to the input and ground nodes. The input pad is connected to the top side of the transducer, which is covered in 50 nm of electron beam–deposited gold (Au) and connected to the Al input pad with a 1.2-μm-thick conformal layer of sputter-deposited copper (Cu). The Al ground pad (250 μm by 250 μm, which defines the effective area of the transducer) is connected to the 50-nm-thick Au-covered bottom side of the transducer using an anisotropic conductive film (ACF). A layer of 8-μm-thick parylene is uniformly deposited over the mote with complete coverage (not shown) to provide biocompatible passivation (2, 37, 38) (see Materials and Methods for a more detailed fabrication protocol and fig. S1 for a detailed fabrication schematic). When ultrasound is applied, an output voltage is generated across the two sides of the transducer to power the IC. Figure 1B is a scanning electron microscope (SEM) image of the mote, also shown in scale in Fig. 1C. Figure 1D shows a mote placed at the tip of an 18–gauge (G) needle (0.84-mm inner diameter), highlighting the highly miniaturized form factor of the device. Figure 1E depicts seven motes loaded into a 1-ml syringe filled with phosphate-buffered saline (PBS) ready for an injection delivery.

• Download high-res image
• Open in new tab
• Download Powerpoint

Figure 1F shows the system architecture and block diagram of the mote. A customized and pulsed ultrasound signal (lower left corner of Fig. 1F) is generated with a commercial ultrasound system (Vantage 256, Verasonics) via an ultrasound imaging probe (L12-3v, Verasonics). In addition to providing power, the imager communicates to the mote with acoustic backscattering (21) through modulation of the on-chip PZT transducer electrical load. The IC itself [see note S1 for detailed circuit schematics and elaboration on circuit design can be found elsewhere (36)] leverages low-power design approaches by operating most of its transistors in the subthreshold regime [with a gate-source voltage of around 0 V (39)], reducing power consumption to only ~0.8 nW at physiological temperature (37°C). To overcome the variations of the incoming ultrasound power, the IC features a power-regulating block at the front end for converting the harvested energy from the PZT transducer to a regulated, constant voltage with a conversion efficiency of better than 71%. This voltage serves as the power supply for a temperature sensor based on a relaxation oscillator. The oscillation frequency of the oscillator is designed to vary with temperature by exploiting the temperature dependence of the leakage current flowing through a subthreshold-biased transistor. This temperature-dependent oscillator controls a modulator, which places either a high impedance (open) or a low impedance (short) across the transducer periodically, resulting in the modulation of the power of the backscattered ultrasound. Such modulation causes the amplitude of the reflected ultrasound echo to vary in the same frequency as the oscillator. At the ultrasound imager, temperature measurements are made by measuring this oscillation frequency, which shows an exponential relationship with temperatures in the form of f(T)=AT−0.5(eBT)$\mathit{f}(\mathit{T})=\mathit{A}{\mathit{T}}^{-0.5}({\mathit{e}}^{\frac{\mathit{B}}{\mathit{T}}})$, where f is the frequency, T is absolute temperature, and A and B are two process-related, temperature-independent constants (36). A fitting curve can be obtained by calibrating at two different temperatures, and any temperature can be calculated using this fitting curve and the detected oscillation frequency. In addition, the front-end power regulation results in a temperature error of only 0.088°C (36) within a wide input power range from 0.07 to 0.25 mW/mm², expressed as spatial-peak temporal-average intensity (I_SPTA).

In vitro characterization of mote functions

We first characterized the temperature sensing functionality of the motes with an in vitro tissue model. Motes were loaded on a piece of biocompatible polyimide (PI) substrate (inset in Fig. 2A) for handling and separately embedded in a piece of chicken muscle tissue, which was placed on a hot plate for temperature control. The L12-3v ultrasound probe was placed directly above the mote through degassed ultrasound gel. An ultrasound image (Fig. 2A) was formed first to locate the mote, which was ~3.5 mm deep in the tissue. A customized ultrasound signal with a duty cycle of ~0.048% and an I_SPTA of ~0.044 mW/mm² [much lower than 7.2 mW/mm², the safe exposure limit (20, 21)] successfully powered up the mote, which translates to a theoretical maximum implantation depth of ~5.7 cm assuming an applied energy of 7.2 mW/mm² and a tissue attenuation of 0.5 dB/(cm·MHz). In practice, we have only been able to demonstrate implantation depths up to 2 cm with a reasonable output power of ~1.85 mW/mm² from the probe primarily because of practical limitations in focusing the incident ultrasound energy and capturing the backscattered signal with a linear-array imaging transducer (see Discussion). This signal, composed of ultrasound pulses separated by 1 ms (Fig. 2B) and configured with a graphical user interface (see note S2), has a pulse repetition rate of 1 kHz, leading to a minimum detectable change of 1 ms in the oscillation period at a sampling rate of 1 kHz. Each pulse is four cycles at the 8.3-MHz ultrasound carrier frequency (inset in Fig. 2B) with a focal spot size of ~350 μm, large enough to accommodate motion in the probe while delivering a focused ultrasound power of approximately 2.75 μW at 3.5-mm depth.

• Download high-res image
• Open in new tab
• Download Powerpoint

The temperature of the mote was varied from 25° to 50°C (measured with a reference thermometer placed near the mote), and the acoustic backscattering data was recorded for each swept temperature. A 1.28-s period of backscattering data (1280 data points) measured at 37°C is shown in Fig. 2C. The signal-to-noise ratio (SNR) is ~56.6, large enough to identify the frequency of this signal. Figure 2D plots the temperature-to-frequency curves for two different motes processed from the same wafer, which show similar temperature dependence with a correlation coefficient of 0.9998. The average frequency at 37°C is ~7.72 Hz, and the average temperature sensitivity is ~0.45 Hz/°C (5.83%/°C) at this temperature. The average temperature resolution was measured to be ~0.049°C [root mean square (RMS)] at 37°C (see note S3), and the average temperature error was found to be approximately +0.044/−0.040°C from 31° to 35°C (see note S4), both sufficient for physiological temperature measurements (40–42). If a higher temperature error is tolerable for certain applications, then the mote can operate from 25° to 50°C, which covers all the relevant physiological temperatures (5) with a maximum temperature error of approximately ±0.2°C.

With an 8.3-MHz operating frequency, the lateral resolution of the imager is ∼280 μm and the focal spot size is ∼350 μm, both of which are smaller than the physical device size of the motes. The spatial resolution of the sensor is given by the device’s linear dimension, which is ~380 μm. In in vitro experiments, we have been able to demonstrate performance of two motes spaced less than 10 μm apart.

The maximum temporal resolution to measure changes in temperature for these sensors is determined by the oscillation frequency with a minimum of one oscillation cycle required for measurement. Therefore, the temporal resolution is 130 ms at 37°C as determined from an oscillation frequency of ~7.7 Hz. Such resolution, comparable to other biomedical sensors (5, 6), is sufficient for most sensing application with slowly varying physiological temperature changes.

In vivo evaluation in mice

Motes were then deployed in laboratory mice to evaluate their temperature sensing performance in several anatomical locations and as a tool to measure local heating in the context of FUS neuromodulation of the sciatic nerve. In all these experiments, degassed and warmed ultrasound gel (see Materials and Methods) was used as the coupling medium for transmitting ultrasound, and a heating pad was used to maintain the animal’s proper body temperature.

One of the concerns with using motes for temperature measurement is that the motes themselves will result in local heating at the incident ultrasound pressure required for operation. To confirm that this is not the case, motes were calibrated for temperature measurement (see note S4) and implanted in two representative anatomical locations, the brain (see Materials and Methods and fig. S6) and the hindlimb (see Materials and Methods and fig. S7). Figure 3A shows the experimental setup for testing on the brain. The L12-3v probe was positioned at a 5-mm distance from the mouse head to deliver power and communicate to the implanted mote. An external temperature probe was inserted subdermally in the neck area for reference temperature recording. An ultrasound power density of 0.032 mW/mm² at the probe was used in this case. The temperature recorded from the mote compares well with the reference temperature, indicating that the device is not artificially heating the surrounding tissue (Fig. 3B). The body temperature in the hindlimb was measured in a similar way (Fig. 3C). In this case, we also removed the heating pad at the start of the measurement, resulting in a decrease in the measured body temperature (Fig. 3D).

• Download high-res image
• Open in new tab
• Download Powerpoint

We further evaluated the use of the motes for temperature measurement during FUS neuromodulation of the sciatic nerve. Neuromodulatory effects were evoked by applying ultrasonic pulse trains of 1-ms pulse duration and 10-Hz pulse repetition frequency directly at the hindlimb to induce muscle contractions in the gastrocnemius muscle of the lower hindlimb (Fig. 4A), as confirmed with electromyography (EMG) recordings with a needle electrode (43). A single 3.57-MHz ultrasound transducer and a coupling cone for focusing are used for this stimulation (see Materials and Methods). Figure 4B shows the EMG responses with four different stimulation intensities (expressed as I_SPTA), where higher ultrasound intensities lead to higher excitation peak amplitudes in the EMG recordings. One of the key mechanistic questions for FUS neuromodulation is the relative role of mechanical and thermal effects (30); therefore, monitoring local heating at the site of stimulation is important. For temperature monitoring during FUS, two motes were implanted (see Materials and Methods), one at each side of the nerve at the size of stimulation (Fig. 4C). The same four stimulation intensities used for the EMG recordings were again applied directly at the sciatic nerve where the motes were implanted to induce neuromodulation (Fig. 4D).

• Download high-res image
• Open in new tab
• Download Powerpoint

The temperature profile during stimulation was also recorded using the L12-3v probe and the implanted motes (Fig. 4E). Temperature increases at the sciatic nerve were calculated by subtracting the temperature measured immediately before FUS exposure from the value measured immediately after FUS exposure. As shown in Fig. 4F, both motes show higher levels of heating with higher ultrasound stimulation intensity, with heating as high as ~0.77°C observed at an ultrasound intensity of 1745.83 mW/cm². To control for the possibility that the implanted motes were resulting in additional heating under FUS stimulation, the same FUS pulses were directly applied in vitro to the motes embedded in ultrasound gel (see fig. S8), and negligible temperature increases were observed (Fig. 4F). This control experiment confirms that the temperature measured from the implants originates from FUS effects on the tissue rather than the mote itself.