Polarization optical imaging, which utilizes changes in the vector properties of light as it interacts with a sample, has emerged as a vital tool in biomedical and clinical applications [1–3]. It has proven to be particularly well suited for characterizing microstructures in fibrous constituents (e.g., muscle fibers, collagen, elastin) and effectively detecting modifications in the organization of precancerous tissue, cancerous tissue, and radio-frequency-ablated lesions within the myocardium [4–12]. Polarization imaging has also been extensively utilized for investigating structural damage caused by various pathological conditions in a wide range of clinical frameworks [13,14]. However, the strong optical scattering that occurs in biological tissue across a range of wavelengths has conventionally prevented polarization imaging at depths beyond ∼1mm, which severely limits its practical applications [1,15,16].

Microwave-induced thermoacoustic imaging (MTAI) utilizes pulsed microwaves as an excitation source for generating acoustic signals. Transient heating produces broadband acoustic frequencies on the order of several megahertz in biological tissue due to the absorption of pulsed microwaves by polar molecules or ions, which can then be detected by ultrasound transducers at the tissue surface and reconstructed to form images of the absorbed microwave energy distribution. MTAI combines the high contrast of microwave imaging with the high spatial resolution of ultrasound as an observable contrast exists between diseased tissue and normal tissue due to differences in microwave absorption coefficients [17–19]. In addition, microwave scattering in soft tissue is three orders of magnitude weaker than optical scattering due to the longer wavelengths employed for excitation. As such, MTAI can provide submillimeter resolution at centimeter depths with ultrasound as the information carrier and is a potential candidate for noninvasive disease detection in deep tissues [20–25].

MTAI uses the scalar microwave absorption coefficient as a quantitative imaging target. In practice, this coefficient varies in biological tissues and, thus, emerges as an anisotropic property that depends on both cell morphology and alignment. The basic structural composition of organisms, such as muscle and myocardial fibers, has been shown to exhibit obvious microwave absorption anisotropy [26–28]. For example, normal myocardial tissue is composed of regularly arranged muscle fibers as depicted by the Torrent-Guasp model [29–32]. If myocardial infarction (MI) occurs, the loss of cardiomyocytes (due to necrosis and a rise in collagen) decreases microwave absorption anisotropy and provides an opportunity to image microstructural changes in the tissue [33].

In this paper, a polarization microwave-induced thermoacoustic imaging (P-MTAI) technique is proposed to measure microstructural variations in biological tissues by exploiting absorption anisotropy. A theoretical model is developed and experimentally validated by applying P-MTAI to four linearly polarized microwave samples in various polarization orientations, used as excitation sources to induce thermoacoustic (TA) signals. Reconstructing these signals provided information concerning microscopic variations in the tissue, quantified by the newly proposed degree of microwave absorption anisotropy (DOMA) parameter. The practical feasibility of this approach was tested using vector microwave absorption phantoms, normal cardiac tissue, and tissues exhibiting MI. Results showed that structural variations in tissues were successfully quantified using DOMA. The low scattering characteristics of microwaves and ultrasound in soft tissues allowed P-MTAI to overcome the imaging depth limitations of polarization optical imaging, enabling submillimeter resolution over depths of several centimeters. P-MTAI was also applied to the noninvasive detection of microstructural changes in cardiomyocytes transitioning from ordered to disordered states. With further research, this may provide a potential indicator for predicting the occurrence of MI.

The operating principles underlying P-MTAI can be described as follows. The net polarization angle θ can be defined as the difference between the polarization of the excitation pulse ψ and the target’s primary axis orientation φ. The TA amplitude PTA(r⇀,θ) on the imaging plane is then a function of both the spatial position r⇀ and the polarization angle θ=ψ−φ. Mathematically, PTA(r⇀,θ) can also be represented as the product of a microwave absorption coefficient α (for biological tissue) and the local microwave fluence I(t), PTA(r⇀,θ)=ΓηthαI(t).(1)

Here, ηth and Γ are the heat conversion efficiency and the Grueneisen parameter. In this paper, the sample was assumed to be an anisotropic microwave absorption target. The direction of highest absorption defined the primary axis. As such, the interaction between the linearly polarized microwaves and an anisotropic target strongly depended on the angle between the microwave polarization orientation ψ and the target’s primary axis orientation φ. The absorption coefficient α along the polarization direction represented in Eq. (1) could then be expressed as [16,34] $$\begin{gathered} α(ψ,φ)=α‖cos2(ψ−φ)+α⊥sin2(ψ−φ) \\ =α‖+α⊥2+α‖−Aα⊥2cos(2ψ−2φ), \end{gathered}$$(2)where α‖ and α⊥ are microwave absorption coefficients along the directions parallel and perpendicular to the primary axis. The TA signal amplitude for an anisotropic target excited by a linearly polarized microwave can then be written as PTA(θ)=[α‖+α⊥2+α‖−Aα⊥2cos(2¯)]ΓηthI(t).(3)

Target heterogeneity was characterized by a new parameter, the DOMA, measured using four linearly polarized microwave excitation sources in different polarization orientations. In the Stokes formalism, the polarization state of a linearly polarized microwave can be described using a Stokes vector [1], S=[S0S1S2S3]=[IH+IVIH−IVIP−IMIR−IL],(4)where IH, IV, IP, and IM are polarization orientations of 0°, 90°, 45°, and −45°, respectively. The IR and IL terms denote right and left circular polarizations, S0 is the total detected microwave energy corresponding to the addition of two orthogonal components’ intensities, S1 is the difference in intensity between horizontal and vertical polarization states, S2 is the fraction of the intensity corresponding to a difference between 45° and −45° polarization states, and S3 is the difference between intensities for right circular and left circular polarization states.

In the Stokes formalism, the degree of linear polarization (DOLP) for a microwave is defined as DOLP=S12+S22S0.(5)

Assuming the target to be uniaxial, DOMA for a target is given by DOMA=S1TA2+S2TA2S0TA,(6)where S1TA=PH−TA−PV−TA, S2TA=PP−TA−PM−TA, and S0TA=PH−TA+PV−TA. The terms PH−TA, PV−TA, PP−TA, and PM−TA correspond to TA signal amplitudes excited by linearly polarized microwaves with orientations of 0°, 90°, 45°, and −45°, respectively. The DOMA value for uniaxial targets is normalized between 0 and 1.

A universal framework for modeling polarization in P-MTAI was developed by representing samples in spatial X–Y coordinates. In this system, the direction of high conductivity defined the principal axis, and θ denoted the angle between this axis and the electric-field E⇀. The sample was oriented along the X axis in the X–Y coordinate system and irradiated by linearly polarized microwaves as shown in Fig. 1(a). Tests showed the aspect ratio of the target conductivity exhibited basic structural anisotropy and strongly depended on the morphology and arrangement of tissue cells. In this configuration, the TA signal intensity depended on the polarization state of the incident microwave, was maximized for polarization along θ=0°, and was minimized in the orthogonal direction (θ=90°). As a result, the interaction strength between linearly polarized microwaves and heterogeneous targets depended strongly on θ. Thus, microscopic target anisotropy could be quantified using normalized DOMA ranging from 0 to 1, according to Eq. (4). Differing values were acquired over the target by varying the polarization orientation of the excitation pulsed microwaves as shown in Fig. 1(b) with higher DOMA indicating higher anisotropy. Both DOMA and fluctuations in the numerical simulation curve for peak normalized TA signal values decreased gradually as the value of α⊥α‖ increased for a given target. This simulation curve was a straight line when α⊥=α‖, indicating the maximum of the normalized thermoacoustic signal was no longer dependent on θ.

A magnetron mode pulse microwave generator (central frequency: 3 GHz, rated voltage: 10 kV, repetition frequency: 10 Hz, peak power: 70 kW) emitted a pulse microwave with a pulse width of 500 ns. Then, a pulsed microwave was coupled to a rotatable half-wave dipole antenna [size: 6cm×6cm, shown in Fig. 2(c)] via a semirigid coaxial cable (length: 1.5 m, insertion loss: 1.2 dB). The microwave energy density of the antenna port was about 7.375mW/cm2, which is lower than the safety standard (10mW/cm2 at 3 GHz [35]). By rotating the antenna, linearly polarized microwaves with the same fluence can be generated in different directions. TA signals were excited by polarized microwaves at angles of 0°, 45°, 90°, and 135° [Fig. 1(b)]. A 128-line ultrasonic transducer was adopted (central frequency: 5 MHz, active pitching angle: 0.3 mm, focal length: 50 mm, fractional bandwidth: 70%, Guangzhou Doppler Electronics Technology Co., Ltd., China) receiving TA signals. The TA signals were then amplified by a custom 128-channel low noise amplifier (voltage gain: 33 dB; −3dB bandwidth: 20 kHz–10 MHz), and acquired by a commercial U.S. imaging system (Prodigy, S-Sharp, Inc., Taiwan, China). The commercial U.S. imaging system served as a platform for data acquisition and image reconstruction. Data were digitized at a 20-MHz sampling rate in parallel. Then, a 10 times average was used to improve signal quality before reconstruction, and the DOMA value defined by Eq. (6) was calculated by a signal processor [Figs. 1(c) and 2].

To prevent the interference of a microwave to the acquisition system and improve the precision of signal acquisition and image quality, a microwave anechoic chamber was designed. The microwave anechoic chamber was a special room composed of absorbing material (polyurethane absorbing sponge) and a metal shielding body. The appearance of the absorbing material was pyramidal with a cone height of one-microwave wavelength (10 cm). The low noise amplifier was covered with a metal casing to shield electromagnetic (EM) interference [Fig. 2(b)]. A thin metal shielding shell (grounded) between the ultrasonic transducer housing and the internal components can reduce the EM interference. In addition, a copper case was added for the ultrasonic probe, which further improved the EM shielding effect.

This paper complied with the National Institutes of Health Laboratory Animal Center standards for the care and use of laboratory animals. Eight female wild-type (WT) rabbits aged 80 days (2.2–2.5 kg) were used for the study. All experiments were approved by the Ethical Committee for Animal Research at South China Normal University and were conducted following approved guidelines and regulations. Sterile/aseptic techniques were used for all surgical procedures as specified by institutional guidelines. To establish the MI model, we employed a rapid surgical approach that does not require ventilation [36,37]. In brief, rabbits were anesthetized with 2% isoflurane using a vaporized delivery system. Body temperatures were monitored using a thermal probe and maintained at 37°C using a heating pad. In the left side of the chest, a small hole was made, through which the heart protruded out. The left anterior descending artery was ligated using a 6-0 silk suture, after which the heart was immediately placed back into the chest cavity, followed by manual evacuation of air. The wound was closed using sterile medical glue to avoid infection.

The proposed P-MTAI technique was assessed using four polarization directions and four samples with anisotropy values set as shown in Fig. 3(a). The purpose of the P-MTAI technology simulation was to illustrate the rationality of the method, the simulation program was implemented by software MATLAB R2018A. Figure 3(a) shows four simulation media, whose microwave absorption coefficient parameters are set at α‖=1, α‖=0.25α⊥, α‖=0.54α⊥, and α‖=α⊥, respectively. P-MTAI simulation was carried out for these four media under the irradiation of polarized microwaves. These simulated images demonstrate that higher DOMA correlated with stronger dependence on polarization. DOMA images were acquired using the process described in Fig. 3(b) and were consistent with the theoretical values shown in Fig. 1(b).

P-MTAI was based on the use of linearly polarized microwaves, which necessitated testing the degree of linearity. First, to study the electric-field distribution of the linearly polarized microwave in the conductivity anisotropic medium, we used software CST to simulate the conductivity anisotropic sample. The conductivity anisotropic sample was set as σx=6S/m, σy=σz=2S/m [Fig. 4(a)]. Figure 4(b) shows the diagram of the medium and the antenna at different angles. The distribution of the electric field in the sample also showed angular change under the irradiation of four linearly polarized microwaves [Fig. 4(c)]. Then, a line polarization polarizer was placed in the path of the line polarization microwave irradiation, and the line polarization microwave was detected by the FST-100X E-type electric-field probe (operating frequency: 9 kHz–6 GHz; detection sensitivity: <−35dB, Shenzhen) passing through the line polarization polarizer. By rotating the linearly polarized microwave antenna, the electric-field intensity related to the linearly polarized angle was recorded, and the linearly polarized degree of microwave was calculated according to Eq. (5). Equation (5) suggests DOLP=S12+S22S0 with S1=EH−EV, S2=EP−EM, and S0=EH+EV. The terms EH, EV, EP, and EM correspond to the electric-field signal amplitudes for polarizations of 0°, 90°, 45°, and −45°, respectively. The DOLP was 0.96, which satisfied experimental requirements. The depolarization ratio, given by ρ=EHEV, is shown for different tissue thicknesses in Fig. 4(e). As seen in the figure, ρ was less than 14 at depths of up to d=60mm, indicating propagation over these distances did not significantly affect microwave polarization characteristics.

The imaging capabilities of the proposed P-MTAI system were verified using artificial imaging scenarios in which the corresponding resolution was measured experimentally. The axial and lateral resolutions of the P-MTAI system were measured using a 0.30-mm diameter capillary containing a 5% NaCl solution as a microwave absorber [Fig. 5(a)]. Gaussian fitting was performed along the horizontal and vertical tube profiles, and the full width at half maximum (FWHM) was used to measure the lateral [Fig. 5(b)] and axial resolutions [Fig. 5(c)], yielding values of 0.94 and 0.81 mm, respectively.

The viability of the proposed technique was assessed using a carbon fiber bundle exhibiting anisotropic electrical conductivity [38]. The sample was imaged through excitation by a series of linearly polarized microwaves with varying polarization orientations. A photograph of the sample and its corresponding absorption vector distribution is shown in Fig. 6(a). A carbon powder fiber exhibiting isotropic electrical conductivity was also imaged and used as a control sample. The black double arrows in Fig. 6(b) indicate the polarization orientation of the incident microwaves. Results demonstrated that the TA signal amplitude in the carbon fiber samples strongly depended on θ, the angle between the principal axis of the carbon fiber and the microwave polarization direction. In contrast, no obvious changes were observed in the carbon powder samples, which coincided with theoretical predictions. Equation (6) suggests the anisotropy of the carbon and carbon powder samples can be quantified using P-MTAI, which produced DOMA values of 0.75 and 0.0173, respectively [Fig. 6(d)]. DOMA was further acquired at depths of 0, 1, 2, 4, and 6 cm by covering the fiber with biological tissue samples [Figs. 6(i) and 6(j)]. Results indicate the proposed P-MTAI method has the ability to quantify target anisotropy in deep biological tissue.

The feasibility of using P-MTAI to directly acquire anisotropy information from biological tissue was also investigated. Skeletal muscle is composed of fibrous cells, individually arranged in the direction of muscle contraction. Electrical conductivity along the length of the fiber is, thus, significantly higher than between the fibers in the extracellular matrix, which is less conductive than individual cells. As such, skeletal muscle exhibits characteristic microwave absorption anisotropy [9]. In addition, the longitudinal conductivity is significantly higher than the transverse conductivity, even when path differences in charge transport are considered. A longitudinal section of skeletal muscle was selected as the biological tissue sample in the experiment, exhibiting distinctive conductivity anisotropy used to test imaging capabilities. A cross section of skeletal muscle exhibiting isotropic conductivity was used as the control. Figure 6(e) shows a photograph of the samples with an applied hematoxylin-eosin (H&E) stain. Corresponding results demonstrated the TA signal amplitude in the longitudinal skeletal muscle depended strongly on θ. No obvious changes were observed in the cross-sectional skeletal muscle [Figs. 6(f) and 6(g)]. Thermoacoustic signals were processed using an image reconstruction algorithm as shown in Fig. 6(h). DOMA results for the longitudinal section are also provided in the figure with a net value of ∼0.64. In contrast, the DOMA of the cross-sectional sample was 0.025. These results indicate the P-MTAI method can provide quantitative information on anisotropy in biological tissues.

We further tested P-MTAI as a potential diagnostic indicator for predicting the occurrence of MI. Cardiovascular disease is the leading cause of death worldwide, and MI (commonly referred to as a heart attack) is the leading cause of cardiovascular disease-related deaths. Healthy heart tissue is composed primarily of well-organized aligned arrays of cardiomyocytes, arranged helically through the depth of the wall. This includes a right-handed helix in the endocardium, a circumferential helix in the mesocardium, and a left-handed helix in the epicardium [Fig. 7(b)]. Thus, healthy cardiac tissue exhibits electrical conductivity anisotropy, whereas tissue experiencing an MI exhibits isotropy due to an altering of the cardiac cell sequence in the scar tissue of an infarct. As such, P-MTAI could be a useful approach for detecting changes in cardiac tissue microstructure.

P-MTAI was applied to an isolated heart rabbit model of MI. Healthy rabbit hearts were scanned in the same position using guided ultrasound for comparison purposes. Four regions of interest were then selected from the two groups of images and used for statistical analysis as shown in Fig. 7(c). The black, blue, and purple dots, respectively, represent the three independent normal regions marked in Fig. 7(c). The red dots correspond to the yellow box marked “4” in Fig. 7(d), an area where MI occurred. As hypothesized, this region exhibited the smallest differences in thermoacoustic signal amplitude. The DOMA image in Fig. 7(e) provides an intuitive quantification of the differences shown in Fig. 7(d). DOMA values for the three marked normal areas were ∼0.7, whereas those of the marked MI area were ∼0.4. All independent regions were compared between groups, and DOMA values in the hearts of healthy rabbits showed no significant differences [P=0.36; Fig. 7(f), upper row]. However, significant DOMA differences were observed in the rabbits suffering from MI [P=9.5×10−4; Fig. 7(f), bottom row]. Thus, DOMA values acquired with P-MTAI offered high distinguishability between healthy and diseased hearts. H&E staining in Fig. 7(g) further validated cardiac abnormalities with obvious signs of fibrosis in the infarcted area marked by black arrows. These experimental data suggest P-MTAI has the capacity to quantify such microstructural tissue anisotropies.

Next, we further tested the capabilities of P-MTAI for identifying MI areas in deep tissues. Actual imaging conditions were synthesized using in vivo models composed of an isolated rabbit heart and biologic tissue sections. P-MTAI was applied to MI focus areas at varying depths below normal human breast fat (1.5-cm thick) for comparison with a healthy in vivo heart model. Figures 8(a) and 8(b) reveal that P-MTAI images and thermoacoustic signals have great thermoacoustic anisotropy in normal cardiac tissue but no anisotropy in the region of origin of MI. The DOMA image in Fig. 8(c) provides an intuitive representation of the differences shown in Fig. 8(a). DOMA in the three marked normal areas was ∼0.7, and that of the marked MI area was ∼0.4. All independent regions were compared between groups. DOMA in the hearts of healthy rabbits showed no significant differences [P=0.25; Fig. 8(d), left], whereas significant variations were observed in the rabbits suffering from MI [P=9.1×10−4; Fig. 8(d), right]. Thus, P-MTAI offers good distinguishability for healthy and diseased hearts and could serve as a potential alternative for detecting focal areas of MI in deep tissue.

In this paper, targets exhibiting conductivity anisotropy displayed varying microwave absorption as a function of the polarization angle. In addition, isotropic control targets (for which α‖=α⊥) were successfully differentiated from test samples. In the experiments described above, the healthy tissue and infarcted areas were clearly identified and distinguished. In addition, upon induction of MI, the formation of fibrosis reduced the anisotropy of cardiac tissue. As a result, differences between the myocardium and the infarcted areas were evident, and the infarction lesion was clearly distinguishable using P-MTAI. The DOMA in the infarcted area was also reduced significantly (∼1.75×) compared with that of healthy tissue. Furthermore, quantitative DOMA values were found to be in excellent agreement with the results of H&E staining, illustrating the potential of P-MTAI for MI diagnosis.

P-MTAI, which relies on absorption differences for linearly polarized microwaves, is characterized by the clear visualization of lesions and healthy tissue as well as the quantification of DOMA. Images can, thus, be reconstructed by processing TA signals in four polarization directions. The “splitting” effect caused by linear polarization illumination is inevitable. The effect affects DOMA measurement. Previous studies have shown that the image splitting caused by linear polarization illumination could be suppressed by using a circularly polarized illumination [39,40]. However, we cannot use this method to reduce the impact splitting effect on the DOMA measurement because the linear polarization microwave is an indispensable part of the P-MTAI method. It may be desirable to introduce deep learning or artificial intelligence to eliminate splitting distortion.

Currently, P-MTAI requires 1 s to collect data in one polarization direction and, thus, at least, 4 s to reconstruct a full DOMA image. The average normal heart rate for a WT rabbit is between 180 and 250 beats per minute. In humans, the average healthy adult has a heart rate between 60 and 100 beats per minute. As such, the imaging rate for the current P-MTAI system is relatively small relative to the heart rate, but we believe that continuous developments in the technology could lead to higher imaging rates. As a result, this approach has considerable potential for applications in clinical cases.

The MTAI process used in this paper exhibits some limitations in both imaging range and speed. This exploratory paper used 128 line-array scans, highlighting the potential use of P-MTAI as a specific and sensitive tool for detecting underlying diseases, such as MI. Excellent anisotropic quantitative imaging capabilities as well as reproducibility and repeatability have been demonstrated. The quantitative detection of microwave absorption anisotropy in both biological and abiotic samples has been demonstrated. The application of P-MTAI to MI diagnosis in rabbits also confirmed its ability to detect deep tissue diseases. As such, P-MTAI may be able to directly image MI in vivo in a future study. Although these findings are preliminary, our approach to both abiotic imaging and MI detection supports the practical use of DOMA. This new parameter could be a possible imaging marker for monitoring disease progression in deep tissues with the potential for further application.