1. Introduction

Terahertz (THz) radiation refers to electromagnetic waves propagating at frequencies in the THz range, between high-frequency edge of the microwave band, 100GHz, and the long-wavelength edge of far-infrared light, 10THz [1]. In wavelengths, this range corresponds to 0.03mm infrared to 3.0mm microwave. THz radiation has a number of properties that make it an attractive clinical imaging technique. For example, THz wavelengths are much longer than infrared and optical radiation, so scattering in biologic tissues is much reduced and could be negligible [2,3]. Contrary to X-rays, THz radiation has a relatively low photon energy and so does not ionize tissues [4,5] and DNA [6,7]. Many materials of interest have unique spectral “fingerprints” in the terahertz range [8,9]. This offers the possibility to combine spectral identification with imaging. THz radiation can also detect differences in water content and density of a tissue [2,4,5]. Such methods could allow effective detection of cancer with a safer but less invasive means [4,5]. Some frequencies of THz radiation can be used for 3D imaging of teeth [10–12] and may be more accurate than conventional X-ray imaging in dentistry [12].

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% [13]. Moreover, breast cancer is the second leading cause of cancer death in women [13]. The chance that breast cancer will be responsible for a woman's death is about 3% [13]. Recently, several studies have reported on the measurements of the dielectric properties of human breast cancerous tissues in the THz frequency ranges [14–16]. It indicates that there is a significant difference between absorption coefficient spectra of breast tumors and those of normal tissues (fatty and fibrous tissues), especially at around 0.3THz [14,16]. Previous studies indicated that the THz absorption contrast of breast tumor tissues is primarily contributed from water and cancer-induced structure change [16,17]. We have previously demonstrated a room-temperature-operated all-THz fiber-scanning near-field imaging system [18]. A preliminary study indicated that the demonstrated THz near-field transmission illumination imaging system has the potential capability to distinguish breast tumor tissues from normal tissues without any pathologic H&E staining [18]. In this work, we study the performance of the previous demonstrated all-THz fiber-scanning near-field imaging system to diagnose unstained breast tumor slices without dehydration. We chose 320GHz as the working frequency. For the system, we optimized the THz near-field imaging system by a room-temperature-operated Schoktty diode detector instead of the Golay Cell detector, which highly improved the imaging speed by 150 times. All investigated tissues were surgical removal tissues stored hydrated and frozen in Tissue Bank, Research Center of Genomic Medicine, National Taiwan University. All 48 specimens were sliced frozen from tissues of 20 female patients (mean age: 53 years; rang: 36-72 years). After the tissues sectioning, we brought the specimens to THz imaging within half an hour and imaged in a thawed state without dehydration. According to the absorption coefficient contrast, the result of the THz near-field microscopy examination agrees astonishingly well with the follow-up identification of the same sliced sample with pathologic H&E staining diagnosis. Due to the fact that all the results are easily quantifiable based on the absorption coefficient, automation and a more complete fast pre-screening is thus possible to reduce the load of extensive and time-consuming pathological examination procedure. With the help of this THz near-field microscopy, we can expect to economize the use of hospital and human resources.

2. Experimental setup of THz near-field microscopy

The system setup and corresponding experimental data analysis method have been discussed in detail in our previous report [18]. The working condition is under room temperature 23°C and with about 50% humidity. The schematic of the THz fiber-scanning near-field microscope is shown in Fig. 1 . In brief, the adopted highly flexible THz subwavelength polyethylene (PE) fiber [19,20] was with a diameter of 240μm and a length of 33cm. A CW Gunn oscillator module [21,22] emitted 320GHz waves that were collected by a pair of off-axis parabolic mirrors and focused into the PE fiber [10] with a substantially low attenuation constant (5 × 10⁻³ cm⁻¹) at 320GHz. To improve the spatial resolution, behind the fiber output end, we integrated a bull’s-eye metallic spatial filter with a subwavelength aperture (200μm-diameter) to achieve both high transmission power (10-fold higher than transmission through a single bare aperture of the same size) and near-field spatial resolution (240μm< λ/4) beyond the diffraction limit [18]. THz waves radiated from the fiber output end was normally incident onto the substrate side of the metallic spatial filter and the transmitted THz radiation from the exit side of the aperture illuminated the sliced samples, which were fixed on a 150μm-thick coverglass and held on a microscope stage. The transmitted power was then detected by a room-temperature-operated Schoktty diode detector (Virginia Diodes, Inc, model WR-2.8, response time <5μs). Compared to the Golay Cell detector (Microtech Instruments, Inc., response time = 25ms) adopted in a previous study [18], the imaging time (scanning area: 10mm × 10mm) has been improved by at least 150 times from 50mins/100 × 100pixels to 3mins/100 × 100pixels. Further improvement is possible with faster scanning motors.

 

Fig. 1 Schematic of the THz fiber-scanning near-field microscopy.

Download Full Size | PDF

3. Specimens preparation and imaging acquisition

The number of the enrolled female patient is 20 (mean age: 53 years; rang: 36-72 years), with 48 specimens under investigation in total. All the specimens used in the study were provided by the Tissue Bank of the Research Center for Genomic Medicine, National Taiwan University, and this research was approved by the Ethics Committee (IRB) of National Taiwan University Hospital. The surgically-removed tissues under investigation were stored frozen (without dehydration) right after the surgical procedure under −30 °C. All surgical procedures were performed within the recent 1 year. Before THz examination, all specimens were first frozen-sectioned from the tissues into 20μm-thick slices and placed on the selected microscope coverglasses. After the tissues sectioning, we brought the specimens to THz imaging within half an hour. During THz imaging, without any other treatment, we kept the specimens under room temperature 23°C. The laboratory was with ~50% humidity. The specimens were thus under a thawed state during imaging without dehydration. After THz near-field transmission illumination imaging, all specimens were sent for routine pathological H&E staining and examination (by H.-Y. Huang) in National Taiwan University Hospital. That is to say, before THz image, we didn’t know whether the tissues were normal or with tumor, and the THz analysis (by H. Chen) and histological identification (by H.-Y. Huang) were performed independently.

Before sectioning the breast tissues, we selected the microscope coverglasses carefully. The coverglasses were first selected with the same thickness (150μm), and then all the selected coverglasses were fixed on the microscope slice holder of the microscope stage for THz transmission measurement. The distance in the z direction between the bottom of the coverglasses and the surface of the metallic bull’s-eye structure was about 250μm. After acquiring the transmission power of the 150μm-thick coverglasses, we removed the coverglasses whose transmission power 5% lower or higher than the average value. It is important to notice that 5% transmission fluctuation will mean 2.5 mm⁻¹ error in the measured absorption coefficient for a 20μm-thick sample slice. Even thinner slice will have to endure ever greater error and is not recommended. The coverglasses with breast tissues were stabilized on a microscope slice holder of a microscope stage. The distance in the z direction between the bottom of the breast tissue and the surface of the metallic bull’s-eye structure was about 250μm. The THz near-field images were acquired by direct 2D (x-y) scanning of the scanning head (including the fiber output end and bull’s-eye structure) and the detector.

THz near-field images shown in this work were normalized to the background for correcting the angle-dependent bending loss. The background was measured by scanning one blank selected coverglass. All the THz near-field images of the breast tissue sections are shown according to the absorption coefficient (α). Most of the tissue samples were not homogeneous and contained a mixture of tissue types. To extract the properties of the constituent tissue types, the absorption coefficients were averaged linearly by assuming that any reflections and scattering caused by heterogeneities within samples were negligible [16,23]. The absorption coefficient was calculated according to the Beer-Lambert law α = ln(I_s/I_b)/d [23], where I_s is the transmitted power of THz wave through the slices and the coverglass, while I_b is the background (transmitted power of THz wave through the blank coverglass). d is the thickness of the slice, which is measured and averaged according to 3~5 different positions in the slice by an optical microscope. All samples selected were all with an average thickness d greater than 15μm. Besides, the absorption coefficients (α) we used for analyzing experimental results were taken from [15,16], in which the mean absorption coefficient of breast tumor at 320 GHz is higher than normal tissues (including fibrous and fatty tissues). The mean absorption coefficient of fatty tissues, fibrous tissues and tumor are 2~4.5mm⁻¹, 7.5~9mm⁻¹, and 9~12 mm⁻¹, respectively [15,16]. Therefore, as our THz near-field images shown, we define the color bar as follows: if the absorption coefficient inside a certain region of the THz image was greater than 9mm⁻¹, we marked the regions as tumor and showed the region in red color (9~12 mm⁻¹), while fat in blue (2~4.5mm⁻¹) and fibrous in green (7.5~9mm⁻¹) [15,16].

For the poor section quality condition induced by man-made or machine causes, the viscous fatty tissues will adhere to the blade and made the slices with uneven thickness. With higher transmission in these specific thin spots, thus obtained α value will be underestimated. We found that if the sample thickness fluctuation exceeded 5μm, the calculated α would be fallacious. Accurate diagnosis can be rendered by properly sectioning the slices before imaging. In this paper, we report our results from 46 out of the 48 sliced samples with a thickness fluctuation less than 2 μm, while two out of the total 48 sliced samples were discarded due to their large thickness fluctuation greater than 5μm.

4. Performance study

At first, aiming to investigate the reproducibility of the studied system, we obtained THz imaging on different sectioned slices out of the same specimen. Four specimens from 4 patients (Cases 1-4) were selected and imaged, and the number of sectioned slices is 10, 6, 7, and 7, respectively. The acquired THz near-field microscopic images and the corresponding pathologic photomicrograph of H&E stained sections are shown in Fig. 2 . The H&E stained pathologic diagnosis of these 4 patients were normal (Case 1), invasive ductal carcinoma (Cases 2-3), and papillary tumor (Case 4), respectively. From the corresponding THz near-field images based on the absorption coefficients as indicated by different colors, all 10 slices in Case 1 should be normal, while the rest 20 slices in Cases 2-4 are all with tumors. We (H.-Y. Huang) also marked the areas of breast tumor in the pathologic photomicrographs with a black solid boundary. Excellent agreement on the location and shape of the tumor regions can also be found. In this reproducibility study, we thus conclude that the results of the THz near-field microscopic examinations on different slices from the same patients can all match the results of the H&E examinations with a terribly high (~100%) reproducibility, both in normal tissues and tumors.

 

Fig. 2 The THz near-field microscopic images of Case 1-Case 4 and the corresponding pathologic photomicrograph of H&E stained sections. The areas of breast tumor in the pathologic photomicrographs were marked by a black solid boundary.

Download Full Size | PDF

With a high THz reproducibility between different slices out the same patient, in the following performance study, we used only one sample slice per patient for the study on the additional 16 patients (Case 5-Case 20). Thus obtained results on different patients are with a great value for future statistical analysis of the THz diagnosis specificity and sensitivity. The acquired THz near-field microscopic images and the corresponding pathologic photomicrograph of H&E stained sections are shown in Fig. 3 . Based on the pathologic photomicrograph of H&E stained sections, the 16 patients’ cases were diagnosed by the pathologist as normal (Cases 5-9), papillary tumors (Cases 10-14), and invasive ductal carcinoma (Cases 15-20). And from Fig. 3(a) to Fig. 3(c), the diagnosis results of the THz near-field microscopy examinations of the same samples from 16 patients (Cases 5-20) based on the THz absorption coefficients are normal for Cases 5-9 and with tumor for Cases 10-20, all in excellent agreement with the corresponding pathologic diagnosis by H&E staining to distinguish tumors from normal tissues. From the THz near-field microscopic images of the 46 different breast specimens, we find that in all slices with breast tumor, there are regions with absorption constants higher than 9mm⁻¹ in the acquired THz images. In all slices without breast tumor, there are no region with absorption constants higher than 9mm⁻¹ in the acquired THz images.

 

Fig. 3 The THz near-field microscopic images of Case 5 - Case 20 and the corresponding pathologic photomicrograph of H&E stained sections. The areas of breast tumor in the pathologic photomicrographs were marked by a black solid boundary.

Download Full Size | PDF

From our study on 46 specimens, we find that the THz near-field technique can all successfully distinguish slices with breast tumors from normal tissues. THz near-field microscope however cannot identify different types of breast cancers. By comparing with the pathologically marked tumor regions in the H&E stained micrographs, as shown in Figs. 2 and 3, majority THz images can find the tumor regions with a good match in term of size and shape with the corresponding pathological micrographs, but not all of them. This is partly due to the fact that during the procedure of H&E staining, the tissues, especially fatty tissue, could easily be washed out by acetone and thus cause distortion and even loss of part of the specimens. However, this kind of tissue distortion and loss doesn’t occur in our THz near-field images due to the fact that besides frozen-sectioning, THz imaging does not require any preparation on the examined specimens. After THz inspection, it is still possible to perform follow-up H&E stating and pathological analysis due to our selected compatible methodology including sample thickness and the chosen microscope coverglasses, if it is necessary. Compared with the current methods, due to the extensive processing procedure and the time-consuming analytical examination based on their staining pattern and morphological criteria of the cellular and nuclear features in the tissue sections, pathologists can only analyze a small fraction of the exercised tissues, either after biopsy or during the breast conserving surgery. This current procedure risks the sampling error as only few representative areas in a given lesion were taken for examine and only few sections were observed. By using the easily quantifiable THz absorption images to differentiate the breast tumors from normal tissues, automation on a faster and more-complete prescreening program becomes possible, which make the THz near-field microscopy a valuable addition to field surgeons so that pathologists can take advantage of the THz technology to choose the slices or reduce the number of slices to be processed for pathological analysis. Adopting the developed THz technology for a more complete and automated pre-screening, it cannot only reduce the risks on the sampling error, but could also save the valuable time to assist quick definition of the tumor margins. With the capability to reduce the load of extensive and time-consuming pathological examination procedure, this THz near-field microscopy technique can expect to economize the use of hospital and human resources.

5. Conclusion

In conclusion, based on microwave components, we developed a fast-scanning near-field THz microscopy. This microscopy is compact, low-cost, and room-temperature-operated. With two significantly developments by introducing a low-losing PE fiber and a bull’s-eye metallic spatial filter, we realized the high transmission THz power and a near-field 240μm-spatial resolution. Moreover, we selected 320GHz as the working frequency, at which the breast tumors has the most significant THz absorption contrast to normal tissues. Based on the study of 20 female patients, we demonstrate the high diagnostic performance of the THz near-field imaging on breast tumor diagnosis. Through near-field mapping of the absorption constants of breast tissue sections, the THz near-field transmission images are easily quantifiable and will make diagnosis automation possible. Adopting the developed THz technology for a more complete and automated pre-screening, it cannot only reduce the risks on the sampling error of the current pathological examination procedure, but could also save the valuable time to assist quick definition of the tumor margins during surgery. With the capability to reduce the load and sampling error of the current extensive and time-consuming pathological examination procedure, this THz near-field microscopy technique can be expected to economize the use of hospital and human resources. This studied THz near-field microscopy is thus with valuable potential applications in clinical breast tumors examination and is expected to play an important role to differentiate tumor from normal breast tissues fast, automatically, and correctly without tissue staining.