1. Introduction

The American Cancer Society (ACS) statistics show that breast cancer is one of the most common cancers among women, and the chance of developing invasive breast cancer at some time in a woman's life is about 12% [1]. Moreover, breast cancer is the second leading cause of cancer death in women [1]. The chance that breast cancer will be responsible for a woman's death is about 3% [1]. Accordingly, early high-sensitivity detection of breast cancer allows treatment at an earlier stage and would significantly reduce associated mortality. In clinical practice, the methods to detect breast cancer include physical examination, breast ultrasound, X-ray mammography, and breast MRI. X-ray mammography is the process of using low-dose amplitude X-rays to examine the human breast and is the only proven method for screening breast cancer. X-ray mammography can identify changes in the density of the breast tissue (usually caused by an increase in fibro-glandular tissue or tumors) or calcifications within the tissue that can often indicate the presence of cancer. During the procedure, the breast has to be compressed evenly with a uniform thickness in order to reduce the thickness and thickness fluctuation of tissues for improved image quality. However, there are several problems associated with X-ray mammography. Mammography utilizes ionizing X-ray, which may induce malignancy. Breast lesion may be obscured on X-ray by dense fibro-glandular structure [2, 3]. Accordingly, X-ray mammography has a false-negative rate up to 30 percent [3]. In Asian (including Taiwan), the incidence of dense breasts is higher than other Western countries. The cancer detection rate of X-ray mammography is thus even lower [4]. Besides, the average age of breast cancer patients is younger in Asia than other Western countries. Young females (especially of child-bearing age) have more dense breasts and are more susceptible to ionizing radiation. Therefore, a new noninvasive screening examination is needed for better early detection of breast cancer.

THz radiation (T-ray) has a number of properties that make it an attractive clinical in vivo imaging technique [5–10]. For example, terahertz wavelengths are much longer than infrared and optical radiation so scattering in biologic tissues is much reduced and could be negligible [11]. T-ray is non-ionizing [12], in contrast to X-ray. T-ray is very sensitive to polar substances, such as water and hydration state. For this reason, THz waves can provide a better contrast for soft tissues than X-rays. Moreover, previous T-ray studies on ex vivo human breast-cancer slices indicated that T-ray is able to provide high absorption contrast between cancers, fatty tissues and fibrous tissues with a high specificity [13–16]. Therefore, THz transmission imaging could be a valuable addition to X-ray mammography. However due to the high water absorption, previous T-ray imaging was limited in the reflection mode [7–10] which could achieve a high degree of water content sensitivity, while in vivo transmission-mode imaging was considered difficult due to the strong attenuation in most biological tissues. In order to take advantage of the high absorption contrast between breast cancers, fatty tissues and fibrous tissues, in this work, in vivo THz transmission imaging in an animal model is made possible by lowering the T-ray frequency to the low-THz regime (from 108GHz to 143GHz), for the THz absorption in living tissues tends to reduce with lower THz frequency. The possibility of performing in vivo THz transmission imaging of early human breast cancer in a subcutaneous xenograft mouse model is demonstrated. The T-ray transmission imaging system is based on all-room-temperature-operated components and a THz-fiber-scanning transmission imaging scheme [17]. The T-ray imaging system is thus compact, low-cost, but with a relatively low signal-to-noise (S/N) ratio. With our compact system, we continuously monitored the growth of breast cancer (with human breast cancer cell line MDA MB 231) in the developed subcutaneous xenograft mouse model. Our in vivo THz study not only indicates that the T-ray imaging system is able to distinguish breast cancer cells from the surrounding fatty tissue, but also with a high sensitivity. Our in vivo subcutaneous xenograft mouse model study will encourage broad and further studies on early breast cancer screening by using THz transmission or reflection imaging systems.

2. THz source and detector

According to previous THz spectroscopy studies (from 150GHz to 2THz) on ex vivo thin-sliced human breast tissues [13–15], the high tissue absorption would lead to a low T-ray penetration depth, which would render in vivo transmission measurement difficult. Since the absorption of breast tissues tends to reduce with lower THz frequency, at first, we performed ex vivo and in vivo spectroscopy experiments in the frequency range between 108GHz and 143GHz (millimeter wavelength range, MMW) by using a frequency tunable THz source-YIG oscillator module and a room-temperature-operated Schottky diode detector (Virginia Diodes, Inc, model WR-6.5). The YIG oscillator module consists of the following commercial components: a YTO (Micro Lambda Wireless, Inc, model MLOS-1826PA), an active doubler (Phase One Microwave, Inc, model SX50-217), a variable attenuator (Quinstar Technology, Inc, model QDA-Q0000), a power amplifier (Quinstar Technology, Inc, model QPW-36472335-XX), a wideband isolator (Millitech, Inc, model FBI-22), a passive tripler (Virginia diode, Inc, model WR8x3-S026) and a horn antenna (Custom Microwave, Inc, model HO8R). The adopted YIG source generates THz waves tuning from 108GHz to 143GHz with about 3mW output power. Sensitivity of the Schottky diode detector is 10⁻¹⁰W/Hz. The working condition is under room temperature 23°C and with about 50% humidity.

3. Mouse treatment and tissue preparation

All animal experiments were approved by the Institutional Animal Care and Use Committee of National Taiwan University, which permitted cancer cell and fatty tissue implantation, mouse anesthetization, and exposition of T-ray on mouse. In our experiment, the animals used were 4-to-6-month-old female BALB/cAnN.Cg-Foxnl^nu/CrlNarl mice, selected and purchased from the National Laboratory Animal Center in Taiwan. The BALB/cAnN.Cg-Foxnl^nu/CrlNarl mouse is a laboratory mouse from a strain with a genetic mutation that causes a deteriorated or absent thymus, resulting in an inhibited immune system due to a greatly reduced number of T cells. Therefore, they can’t reject cancer cells injection and xenografts of tissues from another species. The breast cancer cells used were MDA MB 231, which were cultivated in L-15 with 10% fetal bovine serum and 1% antibiotics, and the cell concentration was 5×10⁷ with 1 ml of culture media. The injection dosage for each mouse was 0.3ml. The implanted or ex vivo measured fatty tissue was aspirated from 12-week-old female B6.V-Lepob/J mice, and was rinsed three times in the transport medium (NaCl 0.9%(w/v), Glucose 56mM, HEPES 25mM, PSA 10ml, pH 7.4). The cancer cells and fatty tissue implantation as well as in vivo T-ray experiments were all conducted while the mice were anesthetized. We anesthetized the mice by intraperitoneal (IP) injection of ketamine-xylazine (50mg+15mg/kg), and all spectroscopy and imaging experiments were completed in 5 minutes. We then kept the mice warm to help them awake.

For ex vivo measurement of fatty tissue and cancerous tissue, the time from tissue aspiration to measurement was less than 12 hours. The aspirated fatty tissue was minced and filled in a 1mm-thick cell sealed by two 2mm-thick polyethylene (PE) windows, and were measured without other treatment. The grown cancerous tissue was extracted from an implanted mouse. After 2-months cancer growth, we sacrificed the mouse and extracted the cancerous tissue with a visible size of 4mm in diameter. By dissecting the cancerous tissue into a 0.82mm-thick slice, we immediately measured the absorption spectrum. For in vivo measured skin absorption spectrum, without any other treatment to the studied mouse, we stretched the anesthetized mouse dorsal skin, sandwiched it between two 2mm cover glasses, and then positioned the mouse on a platform for THz transmission measurement. The average thickness of the dorsal skin was about 0.6mm according to 10 mice’s measurement and the in vivo absorption coefficients of the dorsal skin was calculated based on the THz attenuation, after calibrating the cover glass reflection and absorption.

Since woman’s breast is a mix of fatty tissue and glandular tissue and the percentage of fatty tissue volume in the total breast volume could reach as high as 56% [18, 19], accordingly, we also implanted fatty tissue subcutaneously into the cancer area in the xenograft mouse model [20]. Current available subcutaneous xenograft animal model prevents us to implant other breast compositions like connective fibrous tissues (collagen and elastin). However these connective tissues are the main composition of the dermis layer of the mouse skin, spatially located right next to the implanted cancer cells. It is worth to mention that the specificity of T-ray to identify contrast between cancer, fatty tissue and fibrous tissue has already been demonstrated in previous ex vivo studies [13,14]. The aspirated fatty tissue, within 12 hours, was subcutaneously injected into the dorsal area of the anesthetized mouse by a syringe with an 18-gauge blunt tip cannula [20–22]. With the same method as the in vivo skin measurement, we calculated the in vivo fatty tissue absorption coefficients after considering the attenuation in skin and cover glasses. Meanwhile, the fat-implanted mouse was studied by T1-weighted magnetic resonance imaging (MRI). After the fatty tissue was subcutaneously implanted, we imaged the fat area on the 5^th day and 35^th day, and the corresponding MRI images of the mouse are shown in Fig. 1 . We marked the fat area by red boundary, and from the axial view (Fig. 1(a) and Fig. 1(c)) and the sagittal view (Fig. 1(b) and Fig. 1(d)), we found that the implanted fatty tissue has been a stable part of the mouse body and did not grow nor shrink.

 

Fig. 1 T1-weighted MRI images: (a), (c) axial view; (b), (d) sagittal view. The images were acquired on the 5^th day (a, b) and 35^th day (c, d) after fat implantation, and the fat area was marked by red boundary.

Download Full Size | PDF

4. Ex vivo and in vivo THz absorption spectra

As shown in Fig. 2 , we have successfully measured lower THz frequency (from 108 GHz to 143 GHz) absorption spectra of ex vivo fatty tissue and breast cancerous tissue, as well as the first absorption spectra of in vivo mouse skin and fatty tissue. From Fig. 2(a), it is clear that cancerous tissue absorbs much more T-ray than fatty tissue even in the studied low THz regime. The distinguishable contrast between fatty and cancerous tissue under the T-ray illumination is one of the keys to realize T-ray screen imaging in the developed mouse model. The in vivo measured T-ray absorption spectra of mouse skin and fatty tissue is shown in Fig. 2(b), and the absorption coefficients are summarized by 10 mice’s measurement. In Fig. 2, by comparing the results on skin and cancerous tissue, we found that their absorption coefficients in this studied frequency range are very close. However with a relatively uniform skin thickness which does not vary in time significantly, the T-ray attenuation due to skin can be easily calibrated by considering the skin as a uniform and position-independent attenuation background.

 

Fig. 2 (a) Ex vivo measured absorption spectra of pure water (black solid circles), breast cancerous tissues (red solid circles) and fatty tissues (blue solid circles). (b) In vivo measured absorption spectra of fatty tissues (blue solid circles) and mouse skin (green solid circles). The error bars represent the standard deviation of the mean and the test number is 10.

Download Full Size | PDF

A previous study indicated that the primary THz absorption contrast of a cancerous tissue is from water and structure change [23]. Tissues from different organs have different water contents [24] and cancerous tissues can have higher water content than normal tissues due to edema or increased vascularity [25]. Breast cancerous tissues have a high water content ~60%, while healthy breast tissues contain much less water, as little as ~40% [26]. In Fig. 2(a), we also include our measured absorption coefficient of pure liquid water in the interested frequency range. It is noticed that the absorption coefficients in cancerous tissue is slightly less than in water. Assuming that our measured breast cancer cells are with 60% water content, after comparing the measured absorption constants, we find that water contributes at least 78% of the measured absorption constant and water is thus the major contrast origin of breast cancer cells in our studied frequency range.

5. Experimental setup of the THz transmission imaging system

The significant contrast between fatty tissue and cancerous tissue under the T-ray illumination makes the application of T-ray imaging to identify breast cancer in the developed animal model possible, which is an important first step even though we have neglected the fibrous tissues in our animal model. Accordingly, we developed a fiber-scanning transmission T-ray imaging system based on the same YIG oscillator module and the Schottky diode detector. It was performed at 108GHz. Figure 3(a) shows the schematic diagram of fiber-scanning [16, 17, 27] THz imaging system. The highly flexible THz sub-wavelength PE fiber adopted was with a diameter of 600μm and a length of 40cm [27–29]. The T-ray, emitted from YIG oscillator module, was first modulated by a chopper, and then collimated by a pair of off-axis parabolic mirrors and focused into the fiber with a substantially low attenuation constant ~10⁻³cm⁻¹ [28–30]. Behind the fiber output end, a PE lens was used to focus T-ray onto the anesthetized mouse and then the transmitted power was detected by the Schottky diode detector. The power on the surface of the mouse skin is about 1mW [31] and it is safe according to the current guidelines [12, 31]. Finally, the collected signals were analyzed by a lock-in amplifier. The picture was obtained by a direct 2D (x-y) scanning of the fiber output end and detector. From Fig. 3(c), a 10mm×10mm scanning image in air, which took less than 5 minutes to acquire, we calculated that the maximum angle-induced 1D fiber bending loss in 10mm was about 10% and the S/N ratio of room-temperature imaging system was higher than 10⁵:1.

 

Fig. 3 (a) Schematic diagram of the THz fiber-scanning transmission imaging system. (b) Photograph of the T-ray imaging system. (c) A direct 2D (x-y) 10mm × 10mm scanning image of air. Power transmission was measured at different fiber bending angles, and the maximum angle-induced 1D fiber bending loss is about 10%. The S/N ratio of the imaging system is 10⁵:1.

Download Full Size | PDF

6. THz transmission imaging of a visible cancer: a proof-of-principle study

The imaging specificity and resolution were demonstrated by imaging a visible breast cancer developed in the subcutaneous xenograft mouse model and the test number was 5. Here we show one mouse’s experimental image. After the cancer cells subcutaneously implanted into a mouse’s dorsal area [32], a visible breast tumor with a size about 2.1mm×3.2mm×1.3mm was developed on the 46^th day, as photographed in Fig. 4(a) . We then injected fatty tissue to embed the tumor (Fig. 4(b)) and sandwiched the embedded dorsal area by two cover glasses (Fig. 4(e)). The thickness of the specific sandwiched tissue, including skin, fatty tissue and the hidden tumor, was 3.8mm. A 5mm×5mm transmitted THz fiber-scanning absorption image of the tumor area is shown in Fig. 4(f). The color bar is defined as absorption constant α from 1.40mm⁻¹ to 2.10mm⁻¹, while 2.10mm⁻¹ is the maximum absorption coefficient measured in the sandwiched tissue including fatty tissue and tumor, while the background of the image was defined as 1.40 mm⁻¹<α<1.45 mm⁻¹, corresponding to in vivo spectral measurement on fatty tissue. From the scanning image, it is clear that the high T-ray absorption of breast tumor is to provide endogenous contrast under T-ray imaging so that it would easily reveal itself from the background absorption. The observed high T-ray absorption area agrees well with the tumor, both in size and shape. From the scanning image, it is also clear that the image resolution is much better than λ (2.8mm), which was achieved not just by the small detection area of the Schottky detector, but also due to the nonlinear relationship between α and transmitted power. By analyzing image, the 2D resolution of our imaging system is 1.1mm × 2.4mm. Meanwhile, this test image also indicates that our THz imaging was able to measure the absorptive change ∆α as low as 0.01 mm⁻¹.

 

Fig. 4 (a)-(e) Photos of a mouse taken during preparation: (a) A visible breast tumor with a size about 2.1mm×3.2mm×1.3mm developed on the 46^th day after the cancer cell implantation; (b)-(c) Fat implantation; (d)-(e) Stretched mouse dorsal skin with area 15mm×10mm and sandwiched by two 2mm cover glasses. (f) In vivo THz imaging of a visible cancer in mouse dorsal skin. The color bar is defined as α from 1.4mm⁻¹ to 2.1mm⁻¹.

Download Full Size | PDF

7. THz imaging of early breast cancer

With a high THz absorption contrast between fatty tissue and cancer, the THz transmission imaging system should then be able to in vivo detect the early cancer before it is visible or sensitive to any other means. To this aim, we developed a different subcutaneous xenograft mouse model to test the sensitivity, and the test number was 16. After the cancer cells injection, on the 6^th day, we anesthetized the mouse and implanted mouse fatty tissue to embed the cancer cells (the place was marked after cancer cells implantation). Starting from the 13^th day, we imaged the cancer implanted area every 3~4 days until the cancer was visible. During T-ray imaging, we anesthetized the mouse and sandwiched the dorsal cancer area by two cover glasses. After finishing the imaging acquisition, we kept the mice warm to help it awake. In the following Fig. 5 and Fig. 6 , we show 2 series of experimental images to demonstrate the possibility and sensitivity of our THz imaging system to early detect breast cancer. Figure 5 shows thus acquired 10mm×10mm THz screening images on the mouse according to the imaging time from the 13^th day to the 33^th day after cancer cell implantation. For the images acquired on the 23^th day to the 33^th day, the color bar is defined according to the maximum measured T-ray absorption coefficient of the sandwiched tissues, while from the 13^th to the 17^th day, the color bar is defined as high as 1.7mm⁻¹ according to the 23^th day for comparison. It is clear that the detected absorption change in early cancer development starts on the 23^th day and the absorption change ∆α is 0.25 mm⁻¹. The early cancer development induced absorption change starts to be detected on the 18^th day in another mouse, and the corresponding 10mm×10mm THz screening images from the 18^th day to the 30^th day after cancer cell implantation are shown in Fig. 6. The color bar is defined according to the maximum measured T-ray absorption coefficient of the sandwiched tissues, and the cancer cells embedded in fat induces absorption change ∆α as small as 0.15 mm⁻¹.

 

Fig. 5 A series of THz images according to the screening time from the 13^th day to the 33^th day after the cancer cell implantation. Early cancer development was detected on the 23^th day, and the induced absorption coefficient change was 0.25mm⁻¹, corresponding to a cancer volume of 1.3mm³. For the images from the 23^th day to the 33^th day, the color bar was defined according to the maximum measured absorption coefficient, while from the 13^th day to the 17^th day, the image color bar was defined according to the 23^th day.

Download Full Size | PDF

 

Fig. 6 Another series of THz images according to the screening time from the 18^th day to the 30^th day after the cancer cell implantation in the specific mouse.

Download Full Size | PDF

In order to estimate the tumor volume, we introduce the concept of cell absorption cross section (σ). The cancer cell absorption cross section σ is defined as: σ=α/N=α×V_cell, where N is the number of absorbing cell per unit volume and V_cell is the volume of a single cancer cell. If the cancer cells embedded in fat induces absorption change ∆α, the cancer cell density N^’ can be describe as: N^’ = ∆α/σ=∆α/(α ×V_cell). The 3D volume of the total cancerous tissue V can then be evaluated. Through the THz absorption spectrum shown in Fig. 2, we calibrated the value of the absorption cross section. As shown in Fig. 5, the detected absorption change in early cancer development on the 23^th day corresponds to a total cancerous tissue volume V of 1.3mm³, while the early cancer development in another mouse shown in Fig. 6 is with 0.8mm³ in volume on the 18^th day. According to the S/N ratio of our imaging system (∆α = 0.01 mm⁻¹), the minimum detectable volume on breast cancer can thus be estimated to be 0.05mm³. By improving the system noise level (through cryogenic temperature systems) and by lengthening the data acquisition time, the minimum detectable volume could be highly improved. This high sensitivity performance of T-ray imaging on cancer cells is primarily contributed from the high absorption constant on the adopted wavelength range, which sacrifices the penetration capability. Compared with X-ray mammography, the sensitivity of X-ray mammography depends on the density of the breast [33] and current X-ray mammography can detect breast cancers as small as 2mm in diameter [34]. Compared with the state of the art clinical X-ray mammography, THz imaging is thus with a potential to serve as an alternative method to early detect breast cancer with a much smaller size.

8. Discussion

For future development of in vivo THz transmission imaging to early detect breast cancer, we did the first step to confirm the possibility in a developed subcutaneous xenograft mouse model. This result indicates the potential of T-ray imaging for noninvasive early cancer detection with a high-sensitivity and without the need of exogenous labeling. Combining with other molecular imaging means, including high-resolution optical microscopy but with a much smaller viewing area, our study indicates the capability of T-ray transmission imaging for live animal study on early cancer developments. Our study however did not prove that this current THz transmission imaging system can be directly applied to human in vivo. The most difficult tumors to spot with existing techniques are those buried in fibrous tissues in women with more dense breasts rather than fatty tissues. It is thus important to adopt or develop another animal model with fibrous tissues to test the capability of T-ray transmission imaging system, not just in terms of specificity, but also in terms of penetration. Another important issue to address in the future is gland. Due to the fact that glands are with a high content of water, it should be with a similar contrast to cancerous tissues under THz illumination, especially when we consider most human breast is originated from glandular tissue and connected duct. Combined these facts, the potential of this developed technology would be seriously limited for future in vivo human imaging. In our current study we didn’t consider the fibrous tissue for the reason that the current available subcutaneous xenograft animal model prevents us to implant fibrous tissues to mimic the human breast condition. However, the previous THz absorption spectrum study indicated that T-ray can well distinguish the difference between breast cancer and fibrous tissue [14]. Our recent study based on the statistics of 22 female patients (mean age: 54 years; rang: 36-72 years) has also indicated that that the diagnostic sensitivity and specificity of the room-temperature-operated THz fiber-scanning near-field microscopy in examining sections of human breast tissues are as high as 100% and can well distinguish the breast tumors from the fibrous tissues [35]. Due to the fast growth rate of the cancerous tissues compared to other tissues, time-lapse imaging could potentially provide the base to differentiate the cancerous tissues from other gland tissues, while multiple-frequency imaging with the assistance of numerical calculation might also provide the capability to distinguish the cancerous tissues from fibrous tissues, in animals. As for the future potential for in vivo human imaging, specificity and penetration capability are all needed to be solved and tested. A new penetration window should be explored in the MMW and the microwave regimes, which should be with an even lower absorption, spatial-resolution, and sensitivity to soft tissues than our current working frequency range. Reflectivity contrast should also be explored. Meanwhile, by applying more sensitive cryogenic temperature detection systems like superconductor-insulator-superconductor (SIS) detectors (sensitivity>10⁻¹³w) and by increasing the illumination power, the penetration capability would also be much improved.

9. Conclusions

In conclusion, in vivo T-ray breast cancer imaging is demonstrated in a subcutaneous xenograft mouse study by using a transmission-based room-temperature-operating system. Owing to the high THz absorption contrast between breast cancer cells and fatty tissues, we successfully achieved in vivo early detection of breast cancer in the developed mouse model. The detection limit of the cancerous tissue volume in our THz room-temperature-operating imaging system was 0.05mm³. Moreover, our demonstrated THz transmission imaging system is compact, noninvasive, and non-ionizing, and could provide a unique and sensitive means to image early breast cancer development in the subcutaneous xenograft mouse model without the need of exogenous contrast.