1.

Introduction

During thermotherapy or cryotherapy, it is necessary to monitor the temperature distribution in the tissues for the safe deposition of heat energy in the surrounding healthy tissue and efficient destruction of tumor and abnormal cells. To this end, real-time temperature monitoring with high spatial resolution $(\sim 1\mathrm{mm})$ and high temperature sensitivity ( $1°\mathrm{C}$ or better) is needed.¹ The most accurate temperature monitoring is by directly measuring the temperature with a thermocouple or thermistor. However, it is invasive, hence, generally not preferred and often not feasible. Several noninvasive temperature monitoring methods have been developed. Infrared thermography is a real-time method with $0.1°\mathrm{C}$ accuracy but is limited only to superficial temperature.² Ultrasound can be applied for real-time temperature measurements with good spatial resolution and high penetration depth, but the temperature sensitivity is low.^{3, 4, 5} Magnetic resonance imaging has the advantages of high resolution and sensitivity, but it is expensive, bulky, and slow.^{6, 7} Therefore, an accurate, noninvasive, real-time temperature measurement method needs to be developed.

The thermoacoustic (TA) and photoacoustic (PA) effects are based on the generation of pressure waves on absorption of microwave and light energy, respectively. A short microwave and laser pulse is usually used to irradiate the tissue. If thermal confinement and stress confinement conditions are met, then pressure waves are generated efficiently. The pressure rise of the generated acoustic wave is proportional to a dimensionless parameter called the Grueneisen parameter, and to the local fluence. The local fluence depends on the tissue parameters, such as the absorption coefficient, scattering coefficient, and anisotropy factor, and does not change significantly with temperature. However, the Grueneisen parameter, which depends on the isothermal compressibility, the thermal coefficient of volume expansion, the mass density, and the specific heat capacity at constant volume of the tissue, changes significantly with temperature. Thus, the generated TA/PA signal amplitude changes with temperature. Here, we show that by monitoring the change in the TA/PA signal amplitude, we were able to monitor the change in temperature of the object.

The TA/PA technique has been widely applied in biomedical imaging applications, such as breast cancer imaging, brain structural and functional imaging, blood-oxygenation and hemoglobin monitoring, tumor angiogenesis, and, recently, for molecular imaging.^{8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22} Lately, PA sensing has also been used to monitor tissue temperature.^{1, 23, 24, 25, 26, 27} However, TA sensing of temperature has never been studied. These two techniques do not interact and can be used independently. Depending on the need, one has to choose which technique to use. The main difference between these two techniques is the contrast mechanism. For example, water and ion concentrations are the main sources of contrast in TA measurements, whereas blood and melanin are the main sources of contrast in PA measurements. Therefore, if we need to monitor the temperature of a blood vessel, then the PA technique will be more useful; whereas if we need to monitor the temperature of muscles, then the TA technique will be preferred. TA/PA temperature sensing is a noninvasive, real-time method. The TA/PA technique has the ability to image deeply (up to $5\mathrm{cm}$ ) with high spatial resolution (scalable: millimeters to microns). We can monitor temperature with high temporal resolution and high temperature sensitivity (scalable with temporal resolution: $\pm 0.015$ and $\pm 0.15°\mathrm{C}$ at $200\mathrm{s}$ (2000 measurements averaged) and $2\mathrm{s}$ (20 measurements averaged) resolutions, respectively). Because microwaves penetrate more deeply into tissue than light, we can potentially monitor temperature in vivo for locations deep inside the body.

2.

Theoretical Background

If the microwave/laser excitation is much shorter than both the thermal diffusion (i.e., the excitation is in thermal confinement) and the pressure propagation (i.e., the excitation is in stress confinement) in a heated region, the fractional volume expansion $dV∕V$ can be expressed as

When the fractional change in volume is negligible under rapid heating, the local pressure rise immediately after the microwave/laser excitation pulse can be derived as

We define the Grueneisen parameter (dimensionless) as

Therefore,

3.

System Description

Figure 1a shows the combined TA and PA system used for sensing temperature. A similar concept of integrating light with microwaves was used earlier for a breast cancer imaging system.²¹ The plastic chamber containing the sample holder was filled with mineral oil, a nonmicrowave-absorbing material. Moreover, because mineral oil is visibly transparent, light absorption is negligible. Mineral oil also acts as a coupling medium for sound propagation, and thus, mineral oil was an ideal choice as a background medium for all our experiments. The microwave/laser assembly was placed under the sample holder chamber, from where it illuminated the sample by either microwave or laser, alternately, for TA/PA sensing. The microwave was delivered using a horn antenna, whereas the laser was delivered by a free-space optical assembly. The prism and ground glass of the laser illumination system were incorporated inside the microwave horn antenna. As a result, it was not necessary to mechanically switch between the microwave and laser sources: the switching was electronic and instantaneous. Light was delivered through a drilled $\sim 10\text{-}\mathrm{mm}$ -diam hole in one of the narrow walls of the horn antenna. The laser beam was broadened by a concave lens placed outside the hole in the horn antenna, then reflected by the prism and homogenized by the ground glass. This type of beam expansion scheme has been used extensively before.^{13, 15, 28} The insertion of the optical devices inside the microwave horn antenna had no significant effect on the microwave delivery.²¹

Fig. 1

(a) Schematic of the experimental setup. MO: mineral oil bath, SH: sample holder vial, UST: ultrasonic transducer, GG: ground glass, PR: prism, AN: horn antenna, CL: concave lens, TH: thermistor, (b) TA signals from a LDPE vial (i.d. $12\mathrm{mm}$ , volume $5\mathrm{cc}$ ) filled with mineral oil (MO) and DI water (DI). (c) PA signals from LDPE vial filled with ink solution $({\mu }_{\mathrm{a}}=30{\mathrm{cm}}^{-1})$ and mineral oil at $532\mathrm{nm}$ wavelength.

4.

Results and Discussions

Figure 2a shows the peak-to-peak TA signal amplitude and the actual temperature of the DI water. The TA signal decreased as the DI water cooled to room temperature with time. The TA signal follows the actual temperature profile (red line) very well. Figure 2b plots the TA signal versus the temperature of the sample. The green line shows a linear curve fit with an $R^2$ of 0.95. A linear relationship between the actual temperature and the TA signal was observed. Figure 2c shows the TA signal increased when the cold DI water temperature reached room temperature. The TA signal follows the actual temperature profile (red line) very well. Figure 2d plots the TA signal versus the temperature of DI water, and the green line shows a linear curve fit with an $R^2$ of 0.91. A linear relationship between the temperature and the TA signal was observed. For water in this temperature range, the Grueneisen parameter is a linear function of temperature^{31, 32} and, therefore, the TA signal amplitude also varies linearly with the temperature.

Fig. 2

(a) TA signal and actual temperature of DI water as heated DI water was allowed to come to room temperature, (b) TA signal versus temperature, which shows almost a linear relationship. Green line is the linear curve fitting with an $R^2$ of 0.95. (c) TA signal and the actual temperature of DI water as cold DI water was allowed to come to room temperature, (d) TA signal versus temperature, which shows almost a linear relationship. Green line is the linear curve fitting with an $R^2$ of 0.91. (Color online only.)

Similar experiments were done for PA measurements with ink solution as a sample. Figure 3a shows the PA signal generated from the ink solution and the actual temperature. Once again, the PA signal decreased as the solution temperature approached room temperature, and followed the actual temperature profile (red line) very well. Figure 3b plots the PA signal versus temperature, with the green line showing a linear curve fit with an $R^2$ of 0.98. Next, a cold ink solution was allowed to reach room temperature. The PA signal increased as the temperature of the solution increased [Fig. 3c] and followed the actual temperature profile (red line) very well. Figure 3d plots the PA signal versus the temperature with the green line showing a linear curve fit with an $R^2$ of 0.98.

Fig. 3

(a) PA signal and actual temperature of diluted black ink solution $({\mu }_{\mathrm{a}}=30{\mathrm{cm}}^{-1})$ as the heated sample was allowed to come to room temperature, (b) PA signal versus temperature, which shows almost a linear relationship. Green line is the linear curve fitting with an $R^2$ of 0.98. (c) PA signal and the actual temperature of ink solution as cold solution was allowed to come to room temperature, (d) PA signal versus temperature, which shows almost a linear relationship. Green line is the linear curve fitting with an $R^2$ of 0.98. (Color online only.)

Table 1 summarizes the TA/PA measurements. The number of measurements averaged was 20 for each data point, and as the microwave/laser was operating at $10\mathrm{Hz}$ pulse repetition rate, the temporal resolution was $2\mathrm{s}$ . We can see the change in signal per degree change in temperature is slightly higher for increasing temperature than for decreasing temperature. It was 3.6% for increasing temperature (compared to 3.0% for decreasing temperature) in the case of TA measurements, and 5.9% for increasing temperature (compared to 4.1% for decreasing temperature) in the case of PA measurements.

Table 1

TA/PA signal change per degree centigrade change in temperature of DI water/ink solution in a tube.

Figure 4a shows how the Grueneisen parameter of water varies with temperature $(\Gamma =\beta {\upsilon }_{\mathrm{s}}^2∕c_{\mathrm{p}})$ in the temperature range of interest.^{31, 32} It increases linearly with temperature, with a higher slope within the range $0–20°\mathrm{C}$ than within the range $20–100°\mathrm{C}$ (the slope in the range $0–20°\mathrm{C}$ is 1.48 times the slope in the range $20–100°\mathrm{C}$ ). This agrees with our observation in both TA and PA measurements for increasing and decreasing temperature. For the TA measurements, the slope in the range $4–22°\mathrm{C}$ was 1.16 times the slope in the range $23–58°\mathrm{C}$ , and for the PA measurements the slope in the range $13–22°\mathrm{C}$ was 1.44 times the slope in the range $24–46°\mathrm{C}$ . We can also see a slight difference in the signals, depending on the direction from which the sample comes to equilibrium. In the case of TA measurements, the equilibrium signals were 35.1 and $28.9\mathrm{mV}$ (17.6% difference) for equilibrium reached from cooling and warming, respectively. The corresponding equilibrium temperatures were 22.59 and $21.66°\mathrm{C}$ . There was an almost $1°\mathrm{C}$ difference between the two equilibrium temperatures. After accounting for this difference, we see a $\sim 14\%$ $[17.6\%-(3.0\%+3.6\%)∕2=14.3\%]$ signal discrepancy. In the case of PA measurements, the equilibrium signals were 166.7 and $140.3\mathrm{mV}$ (15.8% difference) and the equilibrium temperatures were 23.57 and $21.41°\mathrm{C}$ , respectively. There was an almost $2°\mathrm{C}$ temperature difference between the two equilibrium temperatures. After accounting for this difference, we see a $\sim 6\%$ $[15.8\%-2\times (4.1\%+5.9\%)∕2=5.8\%]$ signal discrepancy. The TA/PA signal generated from the sample is dependent on various factors, such as the orientation of the sample holder and the spatial distribution of microwaves/light on the sample surface. Therefore, the absolute signal is very sensitive to the position of the sample holder. Because the sample holder position was altered slightly between experiments, these variations in signal could arise. If the sample holder had been fixed in its position between different experiments (increasing and decreasing temperatures), the variation in the signal amplitudes would have been less. Of course, when the technology is used to monitor thermal therapy, the sample will be held stationary.

Fig. 4

(a) Variation of the Grueneisen parameter of water with temperature.^{31, 32} (b) Temperature precision versus number of measurements averaged for TA measurements in log-log scale. The plot shows a linear curve with a slope of $-0.47$ . (c) Temperature precision versus number of measurements averaged for PA measurements in log-log scale. The plot shows a linear curve with a slope of $-0.44$ .

The temperature measurement precision was $\pm 1°\mathrm{C}$ for TA measurements and $\pm 0.5°\mathrm{C}$ for PA measurements. Here, we have used the terms precision and sensitivity synonymously. The temperature precision is calculated based on the uncertainty in the mean value of the measurements (standard error). The precision would improve if we took a larger sample size. Figure 4b shows how the temperature precision typically varies with the number of measurements averaged for TA measurements. As expected, the standard error decreased as a factor of the square root of the number of measurements averaged. The greater the number of measurements averaged is, the less the standard error is, and hence, the higher the precision. The plot is on a log-log scale, yielding a linear curve with $-0.47$ slope (mean $\text{slope}=-0.40\pm 0.05$ , for seven repeated experiments). In an ideal situation, the slope should be $-0.5$ . It is observed that the temperature precision could be as high as $\pm 0.1°\mathrm{C}$ with $>2000$ measurements averaged. Of course, the increased precision comes at the expense of temporal resolution. With the current microwave source (a maximum $100\text{-}\mathrm{Hz}$ pulse repetition rate), the temporal resolution could be as long as $20\mathrm{s}$ to achieve a temperature precision of $\pm 0.1°\mathrm{C}$ . However, employing a higher repetition rate microwave source can eventually give us even higher temperature precision with practical temporal resolution. Similarly, Figure 4c shows how the temperature precision varied with the number of measurements averaged for PA measurements. As expected, the precision could be as high as $\pm 0.1°\mathrm{C}$ with $>1000$ measurements averaged. As before, the plot is on a log-log scale, with a linear curve with $-0.44$ slope (mean $\text{slope}=-0.40\pm 0.03$ , for five repeated experiments). Note that the precision varies from sample to sample depending on the signal-to-noise (SNR). We will show later that, for saline and turkey breast tissue, we obtained even higher precision in temperature based on TA measurements with 20 measurements averaged.

4.1.

Saline/Tissue Temperature Monitoring Using TA Measurements

We also used saline (0.9% w/v of NaCl in water) and turkey breast tissue as a phantom. The same experimental procedures were followed as before. Figure 5a shows that the TA signal decreased as the temperature of the saline decreased. The TA signal followed the actual temperature profile (red line) very closely. Figure 5b shows a second-order curve fit (green line) for the TA signal versus the temperature of saline with an $R^2$ of 0.99. Figure 5c shows that the TA signal decreased as the temperature of the tissue decreased. The TA signal followed the actual temperature profile (red line) very closely. Figure 5d shows a second-order curve fit (green line) for the TA signal versus the temperature with an $R^2$ of 0.99. Table 2 summarizes the TA measurements for the sensing of the saline and turkey breast tissue temperatures. As before, 20 measurements were averaged. For the saline, there was $\sim 5.9\%$ change in TA signal, whereas for tissue there was $\sim 6.1\%$ change in signal per degree centigrade change in temperature. The TA signal generated from the tissue had a very high SNR of $\sim 100$ at room temperature $(\sim 24°\mathrm{C})$ . Thus, the minimum detectable signal change (i.e., noise-equivalent signal change) is $\sim 1\%$ , which gives us a temperature sensitivity of $\sim 0.15°\mathrm{C}$ $(=1∕6.1)$ . However, if the SNR is improved, the sensitivity could be even better. The temperature precision was $\pm 0.15°\mathrm{C}$ (with 20 measurements averaged) based on the standard error. As discussed before, taking the average of a greater number of signals $(>2000)$ the precision could be improved to $\pm 0.015°\mathrm{C}$ .

Fig. 5

(a) Temperature monitoring of saline using TA measurements. TA signal followed the actual temperature profile. (b) TA signal versus temperature. A second-order relationship (green line) was observed with an $R^2$ of 0.99. (c) Temperature monitoring of turkey breast tissue using TA measurements. TA signal and actual temperature of turkey breast tissue as heated tissue was allowed to come to room temperature. (d) TA signal versus temperature. A second-order relationship was observed. The green curve shows a second-order curve fit with an $R^2$ of 0.99. (Color online only.)

Table 2

Saline and turkey breast tissue temperature monitoring using TA measurements. The precision is based on averaging 20 measurements.

	Tmax (°C)	Tmin (°C)	Smax (mV)	Smin (mV)	Signal change		Precision(°C)
					(%/°C)	(mV/°C)
Saline	49.0	21.9	196.6	75.3	5.9	4.5	$\pm 0.3$
Turkeybreast tissue	37.0	23.6	335.6	184.5	6.1	11.3	$\pm 0.15$

5.

Conclusions

A novel temperature-sensing method based on TA and PA measurements is reported. This noninvasive method has deep-tissue-sensing capabilities. Depending on the tissue types, either the TA or PA component, or both, could be used. We monitored temperature of DI water thermoacoustically within $5–60°\mathrm{C}$ , and photoacoustically monitored the temperature of diluted ink solution within $12–46°\mathrm{C}$ . We also showed that the TA-based measurements could be used for saline and turkey breast tissue temperature monitoring. The temperature sensitivity was $0.15°\mathrm{C}$ with $2\mathrm{s}$ (20 measurements averaged) temporal resolution. Measurement precision could be improved by taking more signal averages. In the future, we would like to continue the monitoring of temperature in vivo using both TA and PA sensing for various applications, such as temperature monitoring for tissue during radiofrequency ablation, radiation therapy, photothermal therapy, photodynamic therapy, cancer treatment using high intensity focuses ultrasound (HIFU), and drug delivery using HIFU.