In Vivo Electrical Conductivity Contrast Imaging in a Mouse Model of Cancer Using High-frequency Magnetoacoustic Tomography with Magnetic Induction (hfMAT-MI)

Abstract

Cancerous tissues have electrical-conductivity signatures different from normal tissues, which contain potentially useful information for early detection. Despite recent advancements in electrical-conductivity imaging and its applications, imaging electrical conductivities with high spatial resolution remains a challenge for non-invasive diagnosis of early-stage cancer. Among the various electrical-conductivity imaging methods, magnetoacoustic tomography with magnetic induction (MAT-MI) is a promising technology for non-invasive detection of breast cancer. However, previous efforts to use MAT-MI for cancer imaging have suffered due to insufficient spatial resolution. In this work, we have developed a high-frequency MAT-MI (hfMAT-MI) system with a 2-D spatial resolution of 1 mm, a significant improvement over previous methods. Furthermore, we demonstrated the performance of this method using an in vivo cancer model in nude mice with human breast xenograft hindlimb tumors. hfMAT-MI was able to resolve not only the boundaries between cancerous and healthy tissues, but also the tumors’ internal structures. Importantly, we were able to track a growing tumor using our hfMAT-MI method for the first time in an in vivo mouse model, demonstrating the promise of this magneto-acoustic imaging system for effective detection and diagnosis of early-stage breast cancer.

I. INTRODUCTION

Electrical conductivity is one of the most important properties of biological tissues. Imaging electrical conductivities can be used to differentiate tissue types and to identify altered physiological or pathological conditions, given differences in conductivity signatures depending on the measurement frequency [1, 2]. For instance, the central part of a breast carcinoma, composed of aggregated malignant cells, collagens and elastic fibers, exhibits up to ten-times the electrical conductivity (0.3–0.9 S/m) of normal breast tissue (0.03–0.05 S/m), which consists of normal connective and glandular tissues, using frequencies between a few kilohertz to a hundred megahertz [3]. Thus, electrical conductivity can be used to screen for breast cancer in its early stage, which could significantly reduce the mortality rate. T-scan 2000, an FDA approved, multi-frequency non-tomographic scanner for breast tissues’ electrical conductivity (50–20,000 Hz), has been introduced as an adjunct to mammography, with the combined screening approach achieving 6% more sensitivity and 12% more specificity than screening using mammography alone [4]. Despite the efforts made so far, it remains a challenge to localize a non-superficial cancer lesion precisely, particularly in early stage breast cancer when the tumor size is small. Importantly, it has been demonstrated in a large-scale clinical study that the five-year relative survival rate for breast cancer can reach 96.2% if the diagnosed tumor size is less than 5 mm [5]. Thus, there is an unmet need for high-resolution conductivity imaging that can be used to identify breast cancer at its early stage.

Several noninvasive, high-spatial-resolution imaging methods have been developed for obtaining electrical conductivity measurements of biological tissues. These techniques can be categorized into three groups based on their specific mechanisms: magnetic-resonance-based, acoustic- stimulation-based and electromagnetic-stimulation-based imaging modalities.

Magnetic resonance electrical impedance tomography (MREIT) [6] is a magnetic-resonance-based method, which is a hybrid of electrical impedance tomography (EIT) [7] and magnetic resonance current density imaging (MRCDI) [8]. The MREIT can achieve higher spatial resolution and better signal-to-noise ratio (SNR) than that of the EIT, but requires high level current injection within an MRI scanner which causes safety concern. Another well-established magnetic- resonance-based tissue-impedance-imaging technique is the magnetic resonance electrical property tomography (MREPT), which works at the Larmor frequency to map both the electrical conductivity and permittivity. Taking advantage of its high imaging spatial resolution of a few millimeters or less [9, 10], MREPT has the potential to detect tumor pathology. However, both MREIT and MREPT rely on MRI, which suffer from long scanning duration and high cost. In addition, the compatibility for implanted metallic devices, such as pacemakers or nerve stimulators, is another safety concern. Hence, acoustic-stimulation-based and electromagnetic- stimulation-based studies have been explored so to achieve efficient and cost-effective imaging of electrical conductivity.

Among the acoustic-stimulation-based modalities, Hall effect imaging (HEI) [11] and the methods after which it was modeled, such as the ultrasonically induced Lorentz force method [12, 13], uses ultrasound stimulation to produce motion in a sample that is exposed to a passive magnetic field; this leads to the Lorentz force, generating a detectable electrical potential which can be measured by electrodes. However, these imaging methods have limitations, including low spatial resolution and attenuated sensitivity in the region of interest (ROI) caused by low-conductivity layers, i.e. “shielding effect” [14].

In order to achieve better SNR, electromagnetic-stimulation based modalities have been developed. He and colleagues have proposed the magnetoacoustic tomography with magnetic induction (MAT-MI) by integrating electromagnetic stimulation with ultrasound detection [15–17], through which higher resolution both laterally and axially for electrical conductivity imaging of biological tissues can be achieved while the “shielding effects” can also be overcome. In MAT-MI, a pulsed electromagnetic field is delivered by radio-frequency (RF) coils to induce eddy current in a target volume of conductive tissues. In the presence of an external static magnetic field, the induced current leads to a Lorentz force that drives mechanical vibrations within the ultrasound frequency spanning from hundreds of kilohertz to several megahertz. Such vibrations induce acoustic signals, being subsequently detected by ultrasound transducers. The conductivity distribution can be further retrieved by solving a conductivity reconstruction problem from the acoustic measurement [15, 16], including vectorizing the measurement [18], beamforming the collected ultrasound [19] or modifying the coil setup in a multi-excitation scheme [20]. A series of pilot studies have been conducted on saline phantoms [21] and biological tissues [17], as well as tumor specimens [22]. A reported ex vivo experiment on freshly-procured liver tumor specimens has demonstrated the capability of MAT-MI, for the first time, to discriminate the cancerous tissue from its surrounding tissue through conductivity contrast imaging [22]. More recently, an attempt towards in-vivo imaging of a mouse-prostate-tumor model has been reported using a MAT-MI compatible method with the aid of nanoparticles [23]. Despite significant progresses made on MAT-MI, challenges remain, including the limited spatial resolution and the background artifacts.

In this study, we have developed a system for high-frequency magneto-acoustic tomography with magnetic induction (hfMAT-MI) by redesigning the time response function of the conventional MAT-MI system. This function can be formulated as a convolution of the profile of an electromagnetic pulse and the frequency response function of an ultrasound transducer [24–26]. Based on this theory, a high-frequency electromagnetic stimulator and an RF coil with low inductance were introduced to shorten and reshape the pulse waveform. Hence, both the center frequency of the pulsed magnetic field and the −6 dB bandwidth were increased by more than three times at the emitting side; these spectral features were then matched by those of the ultrasound transducer at the receiving side. By decreasing the wavelength of both the EM stimulation and its induced ultrasound signal, the axial resolution was enhanced to one millimeter. Besides the improvement of the frequency response, a high lateral resolution was also achieved using a rotational scanning regime with an increased equivalent acoustic receiving aperture of the hfMAT-MI system for in vivo applications.

We apply this hfMAT-MI to the in vivo imaging of nude mice bearing human breast cancer xenograft hindlimb tumors. We have demonstrated that a 1-mm spatial resolution for electrical conductivity contrast imaging has been achieved by means of hfMAT-MI, largely benefiting inhomogeneous tissue discrimination and early stage tissue anomaly detection.

II. Theory

A. Imaging Theory

Fig. 1 illustrates a conceptual model of MAT-MI. To deliver the stimulation through magnetic induction, a stimulator feeds a coil with a pulsed electrical current I(t), and a pulsed magnetic field B₁(r, t) is thus induced by I(t) through the object along the z-direction. According to the Maxwell-Faraday equation, ∇ × E(r, t) = − ∂B₁(r, t)/∂t, this magnetic field leads to an electric field E(r, t) that drives a rotational eddy current J(r, t) in the object depending on the spatial distribution of electrical conductivity σ(r). According to Ohm’s law, the eddy current is calculated as J(r, t) = σ(r) E(r, t), and its spectrum features, including the center frequency and the bandwidth, are determined by those of I(t). Due to an ignorable displacement current comparing to the magnetic induction current in biological tissues, this magnetic induction process can be described by (1) for a conductive medium based on Ampère–Maxwell equation [15, 27]:

$$\nabla \cdot J(r,t)=\sigma (r)\nabla \cdot E(r,t)+E(r,t)\cdot \nabla \sigma (r)=0,$$ (1)

where ∇∙ and ∇ are the divergence and gradient operators respectively. The Lorentz force resulting from the eddy current in the object is given by F_L(r, t) = J(r, t) × [B₀(r) + B₁(r, t)], where B₀(r) is the z-directional magnetic field by a static magnet [28]. This force translates the electromagnetic pulse into detectable mechanical vibrations within an ultrasonic frequency band, which further generate acoustic pressure, p(r, t), detected by ultrasound transducers scanning around the object. In MAT-MI, due to the fact that the physical size of an imaging object (a few centimeters) is far less than the wavelength (on the scale of meters) of the pulsed magnetic field B₁(r, t), this magnetic field satisfies the quasi-static condition, allowing the separation of the spatial (r) and temporal (t) components of the aforementioned vector fields [17]. Hence, the wave propagation is described in (2) as:

$$\begin{gathered} {\nabla }^{2}p(r,t)-\frac{1}{c_s^2}\frac{{\partial }^2}{\partial t^2}p(r,t)=\nabla \cdot F_L(r,t) \\ =\nabla \cdot \left\{J(r)\times \left[B_0(r)+B_1(r)\right]\right\}f(t) \end{gathered}$$ (2)

in which the acoustic source locates at point r, and c_s is the speed of the sound. f(t) is employed here to represent the MAT-MI’s time response function. This temporal function can be considered a delta function δ(t) with an unlimited-band hypothesis [18]. In a limited-band system, however, f(t) is the convolution of the magnetic stimulation’s waveform S(t) and the ultrasound transducer’s impulse response R(t), i.e. f(t) = S(t) * R(t) [24–26]. The observed pressure signal at the detecting location r_d can be obtained by solving the differential equation (2) using the Green’s function [29] as in:

$$\begin{gathered} p(r_d,t)=-\frac{1}{4\pi }\underset{\text{V}}{\int }{\nabla }_r\cdot \left\{J(r)\times \left[B_0(r)+B_1(r)\right]\right\} \\ \cdot \frac{f(t)*\delta (t-\left|r_d-r\right|/c_s)}{\left|r_d-r\right|}dr. \end{gathered}$$ (3)

Fig. 1.

The conceptual model of MAT-MI.

As a solution for the forward problem of MAT-MI, (3) can be used to simulate the pressure temporal profiles at different detecting locations. After either calculating the pressure signals using (3) in computer simulations or collecting those signals using ultrasound transducers in experiments, the inverse problem can be solved using a time reversal approach in order to reconstruct the acoustic sources [15], as shown in:

$$\nabla \cdot F_L\approx -\frac{1}{2\pi c_s^3}\underset{\Omega }{\iint }\frac{\text{cos}\theta }{\left|r_d-r\right|}p\text{'}\text{'}(r_d,\left|r_d-r\right|/c_s)d\Omega$$ (4)

Ω is the ultrasound detection surface, θ is the angle between the normal vector of Ω at r_d and (r_d - r), and p″ is the second time derivative of acoustic pressure collected at a transducer location. With (4), the acoustic source can be reconstructed and further used to delineate the conductivity contrast in the object.

B. Spatial Resolution

For a two-dimensional image, its spatial resolution comprises of lateral and axial components. In a conventional ultrasound imaging approach, the 3 dB lateral resolution depends on the acoustic wavelength and the system’s F-number [30]. In MAT-MI, a circular scanning method is used, and when the stepping angular distance is less than or equal to the diameter of the transducer, the lateral resolution of the MAT-MI imaging system can be determined by:

$$w_{3\text{dB}}=\lambda \cdot F\#=\lambda \cdot 1/\phi$$ (5)

in which φ is the scanning view measured in radians, and λ is the acoustic wavelength. The circular scanning approach used in MAT-MI leads to a much smaller F# than that of conventional ultrasound imaging. Using a scanning angle larger than 60° in MAT-MI, the calculated 3 dB lateral resolution is better than one wavelength.

For the axial resolution, unlike the pulse-echo regime in conventional ultrasound systems, MAT-MI passively detects acoustic signals with ultrasound transducers. Thus, its axial resolution can be calculated by [30]:

$$d_{ax}=\lambda \cdot M$$ (6)

in which M is the number of oscillations of the emitting EM stimulation that contributes to the center frequency, and in pulse-echo ultrasound imaging, M would be more than one to have this axial resolution equal to several wavelengths. In order to improve the axial resolution, the system’s time response needs to be shortened. By further applying the quasi-static condition and the Biot-Savart law integrated over a circular current loop [31] with a diameter of D, for a specific location a, (7) shows a dependence of the time response function on the waveform of the pulsatile electrical current I(t) flowing through the coil and the transducer’s impulse response:

$${f(t)|}_{a,D}=\frac{{\partial B_1(t)|}_{a,D}}{\partial t}*R(t)=Q(a,D)\frac{\partial I(t)}{\partial t}*R(t)$$ (7)

where Q(a, D) incorporates the first and second kind of complete elliptic integral functions [32]. The oscillation number M should be as small as 1.

III. MAterials and methods

A. Imaging Systems

Fig. 2 presents the hardware setup for the hfMAT-MI experiments. The sample is placed on the x-y plane and is immersed in distilled water serving as the acoustic coupling medium (c_s ≈ 1.5 mm/μs). To produce the high-frequency pulsed magnetic induction for biological tissue imaging, a customized high-power high-frequency stimulator (Applied Pulsed Power, NY, USA) equipped with half-cycle stimulation technique (Fig. 3(a)) is used to feed a low-inductance coil (65 mm diameter, 800 nH inductance) up to 25 kV, and to minimize ringing effects. As presented in Fig. 3(a), the duration of each current pulse is configured to be 680 ns by the stimulator in order to have a −3 dB bandwidth of 1.3 MHz with a center frequency at 1.5 MHz for ∂B₁(t)/∂t, i.e. the induced rotational electrical field (Fig. 3(b)), in spectrum. The stimulator initiates triggering and synchronizes with a high-speed, multi-channel data acquisition card (CSE8482, Dynamic Signals LLC, IL, USA), which collects ultrasound signals at sampling frequency of 12.5 MHz to record 4096 data points at each sampling. To detect the MAT-MI signals, the ultrasound transducer of the hfMAT-MI system is a customized single-element flat-immersion type with a diameter of 14.2 mm, a nominal center frequency of 1.5 MHz, and a −6 dB fractional bandwidth of 76.85% (Olympus NDT, MA, USA). Fig. 3(c) presents the impulse response of the ultrasound transducer, and the detected acoustic signal is amplified by two-stage ultra-low noise ultrasonic preamplifiers with a −3 dB bandwidth of 0.05–2 MHz (5662, Olympus NDT, MA, USA), then is filtered by a homemade band-pass filter (1–10 MHz) and a low-pass filter (BLP-5+, Cain-Forlaw, NY, USA). The transducer mechanically scans around the sample driven by a rotary table (B5990TS, Velmex, NY, USA). This rotational actuator is fed by a programmable stepper motor controller (VXM-1, Velmex, NY, USA). In addition to steering of the rotary table in the x-y plane, the transducer’s vertical location is set by a manual adjuster to align the detecting aperture with the ROI. The hfMAT-MI system console software is programmed using LabVIEW 2013 (National Instruments, TX, USA).

Fig. 2.

The experimental system of the hfMAT-MI.

Fig. 3.

(a) The waveform of the injected pulsed current I(t) having a pulse duration of 680 ns; (b) the waveform of the induced electrical field detected by a probing coil (radius: 10 mm) at a transverse plane 30 mm away from the coil; (c) the normalized waveform of the ultrasound transducer’s impulse response; (d) the calculated time response function f(t) of the hfMAT-MI system; (e) the Welch’s power spectral density estimate of f(t), whose center frequency is located at 1.5MHz.

In Fig. 4(a), the ultrasound gain map [33] is simulated by the Field-II software [34]. In order to use a uniform ultrasound gain and also create time delay for MAT-MI signals separated from the electromagnetic pulse, the ultrasound scanning radius is set to be 160 mm in experiments, and the line profile in Fig. 4(b) depicts the gain along the lateral direction 160 mm away from the surface of the transducer. Further, using the current amplitude shown in Fig. 3(a) and the coil’s inductance value, a simulated magnetic flux density of this pulsed magnetic field 30 mm away from the coil based on the Biot-Savart law is demonstrated in Fig. 4(c) and its center line profile is shown in Fig. 4(d). With regards to the static magnetic field, a customized NdFeB permanent magnet dipole (Dexter Magnetic Technologies, IL, USA) is used to provide a static magnetic field of 0.3–0.6 Tesla in a 160 mm × 200 mm × 180 mm sample space, as shown in Fig. 4 (e)(f).

Fig. 4.

The simulated gain map of the ultrasound transducer is shown in (a) and (b); the simulated pulsed magnetic flux density B₁ 30-mm away from the coil is shown in (c) and (d); and the simulated static magnetic field B₀ formed by the dipole is presented in (e) and (f).

B. Image Co-registration

For anatomical referencing, a 64-channel pulse-echo ultrasound system (OPEN system, Lecoeur Electronique, Chuelles, France) was employed to image the region of interest, including the tumor region, at different heights within the animal. This ultrasound system incorporates a 64-element phased-array transducer (P7–4, ATL), whose center frequency and bandwidth are 5 MHz and 3 MHz respectively. The ultrasound data were then acquired with a synthetic aperture (SA) method. Using SA beamforming and time-reversal algorithms [35, 36], an ultrasound pulse-echo image of a corresponding cross section was obtained. Considering the ultrasound images as reference, feature-based method was used to register the images obtained from hfMAT-MI, and this feature was the visible animal plastic holder in both imaging modalities.

To further delineate internal structures within the tumor, histology was conducted after the in vivo imaging studies. The histology images were used to co-register the electrical conductivity contrast images obtained from the hfMAT-MI by aligning affiliated skin tissues. After the mouse was euthanized, the tumor was excised together with its surrounding tissues, and was fixed in 10% formalin solution for 48–72 hours before transferring to 70% ethanol. The sample was then sent to the Biological Materials Procurement Network at the University of Minnesota for further dehydration, paraffin embedding, and sectioning. Hematoxylin and eosin (H&E) stain was thereafter applied to blue cell nuclei and pink cytoplasm. After the preparation of multiple slides, 40× magnification images were obtained by a digital microscope (ScanScope XT digital slide scanner, Leica Biosystems, IL, USA). The histologic images were then processed by Aperio ImageScope (Leica Biosystems, IL, USA).

C. Mouse Tumor Model

Metastatic human breast carcinoma cell line MDA-MB-435A was used in this study. The cells were cultured in Dulbecco’s DMEM (Modified Eagle Medium, with 584 ml/l L-Glutamine, 4500 mg/l D-Glucose and 110 mg/l Sodium pyruvate, Corning Inc., USA), supplemented with 10% FBS (Fetal Bovine Serum, Gibco, USA), Pen Strep (100 U/ml penicillin and 100ug/ml streptomycin, Gibco, USA) and 0.0675 μg/ml human insulin (Sigma-Aldrich, USA). Cells were maintained under 37°C and 5% CO₂. Next, cells were sub-cultured by applying 0.05% trypsin-0.53 mM EDTA (Invitrogen, USA) for 5 minutes to detach the cells when reaching 70% in flasks. Cells in log phase of growth (50–60% confluent) were then harvested for tumor inoculation. These cells were rinsed, centrifuged and re-suspended by IMEM (Improved MEM without phenol red, serum or other supplements, Gibco, USA) twice, onto 2×10⁷ cells/ml cell suspensions.

All animal procedures and care were approved by the University of Minnesota Institutional Animal Care and Use Committee (IACUC) in accordance with federally approved guidelines. Female nude mice (6–8 weeks, Athymic Nude-Foxn1nu, Harlan Laboratories Inc., USA) were injected with 5×10⁶ cells subcutaneously over the dorsal flanks near the hind limbs under general anesthesia. Experiments were performed throughout the 1–8 weeks after tumor seeding when tumor diameter was between 2–15 mm.

D. In vivo Experiments

After anesthetized by intraperitoneal injection with a mixture of Ketamine and Xylazine (100 mg/kg and 10 mg/kg respectively), 1/2 original volume dose was given immediately before the hfMAT-MI imaging to extend the sedation period. The mouse was placed in a sitting posture inside an equivalent-sized plastic holder. 37℃ agar gel was then used to fill the plastic holder to provide good ultrasound coupling, maintain the animal’s position, and situate the tumor region 1.5–2 cm away from the coil. 1.5% salinity [17] was introduced into this agar gel, so the animal holder could be also seen in hfMAT-MI images, facilitating co-registration with ultrasound pulse-echo images.

During hfMAT-MI imaging, the ultrasound transducer scanned around the animal with a step size of 1.2°, covering a 180° imaging view. At each detecting channel, the signal was averaged over 175 acquisitions using a pulse repetition rate of 8.5 Hz. Thus, one in-vivo imaging trial with hfMAT-MI took approximately 60 minutes. A heating system including a submersible aquarium heater, an air vortex heater, and a thermostat was employed to achieve an ambient temperature of approximately 36 ℃, to keep the mouse from losing body temperature during anesthesia. After the in vivo imaging with hfMAT-MI, the 64-channel ultrasound system was used to conduct SA scans for the mouse in the plastic holder at a series of layers with increments of 1.5 mm along the z direction, i.e. Δz = 1.5 mm. Then, the mouse was taken out from the plastic holder, wiped dry and placed on a heater pad. Next, the tumor size was measured using a caliper before the mouse was recovered from anesthesia. This mouse was then sent back to its cage after restoring its motion capability.

E. Electrical Conductivity Measurement

A homemade device, as shown in Fig. 5, was developed to directly measure the electrical conductivity of biological tissues at the same center frequency of hfMAT-MI. The device employed four unipolar needle electrodes [17, 37] (EL450, BIOPAC Systems, CA, USA) with 1-mm tips exposed without polytetrafluoroethylene (PTFE) coatings. These electrodes were linearly and equally distributed (Spacing: 1.1 mm) on a PTFE beam. The conductivity measurement is based on Ohm’s law. After the motion system physically inserts the electrodes into a specific location of the tissue, an arbitrary waveform generator (33220A, Keysight Technologies, CA, USA) was then used to inject sinusoidal electrical current (P-P Voltage: 200 mV, frequency: 1.5 MHz) via electrodes a and d into a target location of the tissue. This current was monitored by a current-to-voltage converter, employing a JFET input operational amplifier (LF356N, Texas Instrument, TX, USA). Simultaneously, the voltage across the electrodes b and c was detected by a high-speed instrumentation amplifier incorporating a dual-channel ultralow-noise operational amplifier (AD8599, Analog Device, MA, USA) and a precision difference amplifier (AD8274, Analog Device, MA, USA). Thus, both the amplitudes of the current and voltage were displayed and measured with a digital oscilloscope (DSO7014A, Keysight Technologies, CA, USA). As a result, the local conductivity was estimated. Before measuring conductivity values in the animals of interest, the device had been calibrated using saline solutions with known conductivities. Invasive conductivity measurements were taken immediately after euthanizing the mouse through cervical dislocation.

Fig. 5.

The schematic diagram of the four-electrode device for electrical conductivity measurement of biological tissues.

IV. Results

A. Imaging Quality

The spatial resolution can be represented by the lateral and axial resolution, both of which depend on the acoustic wavelength of hfMAT-MI from equations (5) and (6). This wavelength can be calculated from λ = c_s/f₀, where f₀ is the center frequency of the acoustic pressure field p(r_d, t). This center frequency, 1.5 MHz, is determined by the Fourier transform of the time response function f(t) as shown in Fig. 3(e), leading to the acoustic wavelength λ = 1 mm.

In our in-vivo experiment using 1.5 MHz hfMAT-MI, the ultrasound transducer scanned 180° (φ = 3.14 rad) around the object of interest, resulting in a lateral resolution w_3dB of 0.32 mm, whereas from Figure 3(b), the oscillation number (M) of the EM stimulation that contributes to the center frequency is 1. This leads to an axial resolution of d_ax = 1 mm. Therefore, the overall imaging spatial resolution is determined by this 1-mm axial resolution. To compare the spatial resolutions of the 1.5 MHz hfMAT-MI and previous 500 kHz MAT-MI [16, 17, 27, 38], Fig. 6 demonstrates each method’s performance in imaging an identical agar phantom prepared in a square column. As indicated in this comparison, not only can the imaged edges of the object be reduced from 3 mm to 1 mm as shown in Fig. 6(b) and (d), but the EM artifacts were also suppressed approximately 6 dB more using the hfMAT-MI, compared with conventional MAT-MI. Both improvements are essential for reliable conductivity imaging of biological tissues. The capability of EM artifacts suppression is mainly due to engineering upgrades in ultrasound detection, magnetic stimulation, and power system design.

Fig. 6.

The comparison on imaging spatial resolutions provided by 500 kHz MAT-MI (a) and 1.5 MHz hfMAT-MI (c), when both systems image a same square-column phantom (1.5% salinity). The edges of the object are resolved with a 3-mm (b) and 1-mm (d) spatial resolutions respectively.

B. In vivo Tumor Imaging

Fig. 7 presents an in-vivo imaging study for a tumor bearing mouse. The tumor has been propagated at one of the mouse’s hindlimbs for 8 weeks, with the size reaching about 1.5 cm³. The sedated mouse maintained a sitting posture in the plastic holder as shown in Fig. 7(a) and (b). Three slices in the z-direction marked with three dashed lines were selected and imaged by the ultrasound with a phased-array probe, which is also covered by the piston transducer in the hfMAT-MI. A two-millimeter space step among those slices was achieved by controlling the translation stage. The corresponding ultrasound SA images in Fig. 7(c)(d)(e) represent cross-sectional views at different heights (z direction) of the same tumor-bearing mouse. The tumor regions are enclosed with yellow boxes, while inner boundaries (i.e. the boundaries between the tumor and internal structures of the rest of the mouse body) of the tumor are detected by the ultrasound system and indicated by red arrows. Fig. 7(f) is an electrical-conductivity contrast image using the hfMAT-MI, in which a complete contour of the same tumor is depicted. Taking the advantage of its 1-mm spatial resolution, conductivity variations inside the tumor are visualized by the hfMAT-MI image as well. These internal tumor structures can also be observed in the ultrasound images, as indicated with blue arrows. To further co-register these variations, a histology study of the tumor sample followed the hfMAT-MI and ultrasound imaging experiment. Fig. 8 shows the histological slides using the H&E stain. A necrotic core as the pink, surrounded by a “ring-shaped” dark-purple region, can be seen in Fig. 8(a).

Fig. 7.

An in-vivo imaging study of a tumor-bearing mouse. (a) The top view; (b) the side view, with three ultrasound scanning slices (c)(d)(e) indicated by three white lines; (f) the in-vivo image produced by the hfMAT-MI. The yellow boxes represent the propagated tumor region, the red arrows indicate the tumor-muscle interfaces, and the blue arrows indicate the tumor’s inner structures, e.g. necrotic core.

Fig. 8.

(a) A histological slide of the tumor using the H&E stain; (b) an enlarged view of the histological region marked by the green box in (a), in which the purple region show distinguishable features from its surrounding region in pink.

Electrical conductivity measurements at 1.5 MHz using the four-electrode device gave a mean of 0.76 S/m (five sample locations) for the necrotic core, compared to a mean of 0.62 S/m (five sample locations, measured along a transverse direction to the muscle fibers) for the tumor’s adjacent muscle, shown in Fig. 9, with a statistically significant difference between the two groups (p < 0.05) based on a paired sample t-test. As shown in the dark purple in Fig. 8(b), which is an enlarged view from an ROI marked by the green box in Fig. 8(a), the tumor’s ring-shaped peripheral solid layer, due to its limited thickness, only one sample location can be identified and measured by inserting the four electrodes. And the measured electrical conductivity for this peripheral layer is 0.44 S/m. From these measurements, the conductivity contrast inside the tumor was greater than that of the muscle-tumor interface.

Fig. 9.

The conductivity (measured by the four-electrode device at 1.5 MHz) comparison between the mouse muscles and the tumor necrosis core, and a statistical significant difference (p < 0.05) can be observed.

C. Tumor Growth Monitoring

In order to leverage its improved spatial resolution, we used hfMAT-MI to image and monitor a growing tumor in a separate mouse over the course of several weeks. Yellow boxes are used to mark the tumor regions in Fig. 10. As shown in Fig. 10 (a), the tumor grew into a “disk-like” protrusion two weeks after the transplantation of the breast cancer cells. At this early stage, the diameter of the protrusion was about 7 mm, and the height was less than 2.5 mm. Preparation of the mouse (1 week after the tumor implantation) for in-vivo imaging with hfMAT-MI is shown in Fig. 10(b), with the resulting image shown in Fig. 10(c). This image highlighted the small tumor at one hindlimb of the mouse. A significant electrical-conductivity contrast can be seen between this newly developed tumor and its adjacent muscle tissues. This distinct electrical-conductivity boundary (blue arrow) indicates that the propagated tumor is restricted subcutaneously. The 2-mm height of the tumor can be also estimated from the reconstructed image.

Fig. 10.

An in-vivo mouse model study tracking the tumor growth. (a)(b) The tumor appearance in the 1st week after a transplantation of the human cancer cell line. (d)(f)(h) are the ultrasound images of the tumor-bearing mouse in the 2nd, 4th, 6th weeks, and the insets in these three images show the tumor growth in top views. (c)(e)(g)(i) are the hfMAT-MI images of the mouse abdomen and its growing tumor in the 1st, 2nd, 4th, 6th weeks respectively, and more specifically, the sizes of the growing tumor can also be estimated. The yellow boxes represent the regions of the growing tumor, the red arrows indicate the tumor-muscle interfaces, and the blue arrows indicate the tumor’s internal structural change, e.g. necrosis.

Along with tumor growth, we started to use the ultrasound SA image to co-register the electrical-conductivity boundaries detected by hfMAT-MI for the same mouse. Fig. 10(d) and (e) display in-vivo images of the mouse tumor during the second week. An inset at the left-bottom side of Fig. 10(d) is a top-view of the mouse during imaging, and a yellow arrow indicates the tumor region. The geometry of the tumor changes and becomes a small “lump”. However, the ultrasound pulse-echo imaging appears less sensitive to the second-week tumor compared with hfMAT-MI, because the ultrasound image does not explicitly differentiate the tumor region from its surrounding soft tissues, while the image from hfMAT-MI preserves and highlights the tumor “lump”. Moreover, Fig. 10(e) shows that the tumor-muscle interface (blue arrow) starts to lose the high contrast obtained in images from the first week. When the tumor entered into its fourth week, its size had approximately doubled compared to the two week mark. In Fig. 10(f), some of the tumor boundaries start to evolve in the ultrasound image, whereas a sharp boundary is detected using hfMAT-MI, depicted in Fig. 10(g). In addition, both sets of images reveal that the tumor gets less homogeneous over time in terms of both the acoustic characteristics and the electrical properties of the tumor tissue, reflecting internal structural change related to tumor necrosis. The red arrows in Fig. 10(f) and (g) point to the suspected necrotic region of interest. At the sixth week, boundaries of the tumor were more apparent than previous observations in the ultrasound image shown in Fig. 10(h). However, similar to the weak signal from the tumor-muscle interface introduced in Fig. 7(f), the electrical-conductivity contrast was decreased on the margin (blue arrow), shown in Fig 10(i). By the fourth week, besides the expansion of the tumor’s external geometry, its internal structure had become even more complicated, as more conductivity variations are apparent in the hfMAT-MI images. At this late stage, a necrotic core (red arrow) had developed inside the tumor, which was also co-registered with histology in Fig. 11(a). The electrical conductivity of the tumor was also measured using the four-electrode device before the tumor sample was sent for the histology study. By taking three measurement sites in the tumor, the conductivity ranged 0.53 – 0.6 S/m.

Fig. 11.

(a) A histological slide of the tumor in the 6th week; (b) the measured tumor heights over 6 weeks by readings from a caliper (red) and estimations from the hfMAT-MI images (black) in Fig. 10(c)(e)(g)(i), and in results, the tracking of the tumor heights with the external measurements are consistent with the estimations from the in vivo images, and the mean relative error is 4.1%.

During this study, the sizes of the tumor were measured each week, and Fig. 11(b) presents both tumor height measurements recorded with a caliper and the in vivo electrical conductivity contrast images. The mean relative error between the readings from these two approaches was 4.1%. The measurements using the caliper were also able to validate the tumor-muscle boundaries observed in Fig. 10(c)(e)(g)(i).

V. Discussion

Tissue conductivity imaging using MAT-MI was proposed in 2005 [15] with a sonography-comparable spatial resolution. In the present study, for the first time, we present the equations for quantifying the 2-D MAT-MI imaging spatial resolution. Previously, by applying Rayleigh criterion, a spatial resolution of 1.51 mm was claimed through parallel-line-source phantom experiments, which used a high-power magneto-acoustic system with a center frequency of 460 kHz [25]. In the present study, we have demonstrated that by using high-frequency pulsed magnetic induction and its coupled acoustics, we can achieve a spatial resolution of 1-mm while obtaining reasonable contrast measurements compared with existing MAT-MI methods [17, 22, 25, 39]. To compare the spatial resolution with the reported value in [25], the minimum resolvable spatial detail using hfMAT-MI is as small as 0.53 mm if applying Rayleigh criterion. More specifically, the 2-D MAT-MI imaging spatial resolution is divided into the axial and lateral resolutions. We show that both of them depend on the frequency response of the imaging system. Furthermore, by using rotational scanning in MAT-MI, the 2-D imaging spatial resolution appears to be determined only by the axial resolution. This may not be true for applications using limited views, in which the lateral resolution may become the “bottle neck”. It is also worthwhile to note that both the element size of an ultrasound transducer and the distance from this transducer to an imaging object can affect the sensitivity and the imaging resolutions, among which the elevational resolution is another fixed property of the transducer to resolve structures in the z-direction and should be considered in a 3-D imaging case. In our 2-D imaging work, as the transducer uses a single round, flat sensing element, its elevational specification is similar to the simulated line profile in Fig. 4(b) (w_3dB = 12 mm). This 12-mm specification is used to align the transducer with the ROI, i.e. transplanted tumors. To improve this z-directional resolution, focused ultrasound transducers or an acoustic lens [40] can be used. In summary, this work addresses significant technical challenges of in vivo magneto-acoustic imaging and paves the way for future work in early detection of cancer using hfMAT-MI.

In this in-vivo, small-animal breast-cancer model study, we have shown the capability of MAT-MI in discriminating tumors from normal tissues, identifying internal tissue structures and tracking tumor growth for the first time. The imaging contrast mechanism (i.e. conductivity difference) was verified by measuring fresh tumor and its neighboring muscles with our lab-made device. The propagated tumor in Figs. 8 and 9 had been grown for 8 weeks on one mouse subject, and the resulted tumor in Fig. 10 was on another mouse, with Figs. 10(h)(i) and 11(a) showing the tumor images in week 6. Comparing these two cases, the late stage of the tumor was proved to have high inhomogeneity of electrical conductivity, and the tumor necrotic core expressed much high conductivity value due to its fluidic state with decreased cell membrane structures seen from histology results and increased ion concentration in extracellular space. Besides, we also found that when the tumor was grown for 6 weeks as imaged in Fig. 10(h)(i), the electrical conductivity contrast decreased at the tumor-muscle interface. This phenomenon may be explained by an observation when doing tumor anatomy after euthanizing the mouse subject; as a result, we noticed that the tumor had very close tissue connections to the adjacent muscle (photos not included), whereas such tissue connection was relatively loose, and gradually getting close during the initiative four weeks after tumor transplantation, and this was verified using another two mice respectively sacrificed in week 2 and week 4. Further cancer biology study is needed to understand the mechanism.

Through the in vivo experiments, as small as a 20% conductivity difference was able to be imaged with our hfMAT-MI system. In clinical breast cancer imaging, better contrast (the conductivity measured at 1.5 MHz of a developing breast tumor is 10 times higher than that of normal breast adipose tissues [3]) may be presented, which would make our magneto-acoustic imaging method more sensitive in depicting the margin of a breast tumor. However, further tests are needed to address more complicated structures in the human breast than those in an animal model. Despite this challenge, the potential of distinguishing breast cancer tissue from healthy tissues makes hfMAT-MI a promising imaging tool for assisting diagnosis of early-stage human breast cancer. One significant improvement of hfMAT-MI compared with conventional methods is its increased sample space (Fig. 2) between the magnetic dipole, allowing one human breast to be fit into the imaging setup without any difficulties. Moreover, an improved hfMAT-MI system employing multiple receiving ultrasound transducers is also being developed to dramatically shorten the 2-D imaging time down to 10 minutes, which is beneficial when considering subject’s comfort in potential clinical applications. Overall, besides the intrinsic merits of MAT-MI, such as being non-invasive, good imaging depth, and inherent immunity to the “shielding effect” [15], this work demonstrates the integration of tissue electrical-conductivity imaging with pulse-echo ultrasound imaging to provide both structural and functional information for the first time. This combination suggests the feasibility of integrating hfMAT-MI with conventional ultrasound imaging for multi-modal, early diagnosis of human breast cancer. The high-frequency technique introduced in hfMAT-MI also lays a foundation for further detecting the anisotropic property of biological tissues with the 1-mm spatial resolution. The multi-excitation MAT-MI technique have been validated through 2D experiments [20] and 3D trials [40]. Besides reconstructing the conductivity distribution, the multiple coil sets and the corresponding algorithms introduced by the multi-excitation research can also be used to compute the conductivity gradients along both x and y directions, and potentially along z direction if ultrasound elevational resolution can also be further improved.

Compared to other clinically-available imaging modalities, the imaging contrast of hfMAT-MI still needs to be further improved. One approach would be to introduce a strong static magnetic field to address this issue. In fact, several pilot imaging tests on pork phantoms have been conducted employing a 9.4T MRI magnet and a 500 kHz high-power stimulator [41]. Besides the increased costs, the system assembly of MAT-MI in an MRI machine is still challenging for in vivo experiments. Another potential approach is to redesign the RF coil that generates the pulsed magnetic field. hfMAT-MI could utilize low-inductance coils to form a large coil array, allowing an even higher current amplitude flowing through than that achieved in Fig. 3(a). This improvement on the coil engineering is believed, on the one hand, to increase the pulsed magnetic flux density, thus inducing a stronger instantaneous electrical field than that presented in Fig. 3(b). On the other hand, this would expand the effective imaging area, effectively reducing the inhomogeneity of the pulsed magnetic field. Due to the short rise time (less than 1 μs) and the low peak magnetic flux density (less than 0.1 T at 30 mm away from the coil), the current level and waveform employed in hfMAT-MI is believed to be safe avoiding nerve stimulation.

In the present study, breast cancer bearing mice were used as a cancer model. Tumors were induced by injecting MDA-MB-435A (LCC6) cells into the hindlimbs of the mice. This cell line is an ascites model of MDA-MB-435, which was derived at M.D. Anderson Cancer Center, Houston, TX, USA in 1976 from the pleural effusion of a 31-year-old female with metastatic, ductal adenocarcinoma of the breast [42]. Ever since, the MDA-MB-435 cell line has been in extensive worldwide use among established laboratories as a model for human breast cancer. The xenograft tumor formed in immune-deficient rodents (e.g. T-cell deficient athymic nu/nu mice) not only retains genetic and phenotypic properties of its human counterparts, but also serves as a cost-effective in vivo model for understanding tumor biology and mimicking critical elements of disease progression [43, 44]. However, even though breast cancer cell lines and their in vivo studies provide considerable insights into breast carcinoma, a few cell lines are not enough to characterize all types of human breast cancer with distinct features [45]. As for the MDA-MB-435 cell line used in this study, questions have been raised that gene expression analysis of the cells produced some microarrays in which MDA-MB-435 clustered with cell lines of melanoma origin [46–48]. We are aware of the debate about the origin of this cell line [49] and despite these speculations, strong evidence has been presented reiterating that MDA-MB-435 is of breast cancer origin [50] and may be used as “an excellent model for studies of highly malignant and dedifferentiated breast cancers” [47]. However, we realize that discussions concerning the cell line’s genetic origin are outside the scope of this paper, which focus more on demonstrating the capability of high-resolution in vivo imaging of cancer using conductivity contrasts. More work has been planned on testing with in-vivo models using different cancer cell lines as well as procured human breast cancer samples.

VI. Conclusions

In conclusion, we have developed a hfMAT-MI imaging system which provides electrical-conductivity imaging with a 1-mm spatial resolution, a significant improvement over conventional MAT-MI methods. We have also conducted pilot studies on in vivo tumor-bearing mouse models. Our results demonstrate the capability of hfMAT-MI for better discriminating transplanted human cancer tissue from normal surrounding tissues with internal details. To the best of our knowledge, this is the first in-vivo study using hfMAT-MI to track the tumor growth and establishes the feasibility of applying this magneto-acoustic imaging for early breast cancer detection.