INSTITUTE OF PHYSICS PUBLISHING HYSICS IN MEDICINE AND BIOLOGY hys.Med.Biol.52(2007)419-427 10.1088/0031-9155/52/2008 Evaluation of x-ray diffraction enhanced imaging in the diagnosis of breast cancer Chenglin Liu., Xiaohui Yan, Xinyi Zhang. 3, Wentao Yang+ Weijun Peng, Daren Shi, Peiping Zhu, Wanxia Huang and SyicharoKey Radation Rese arch enter, Physis aepatmens a pace pe res lbo at cy Physics Department of Yancheng Teachers'College, Yancheng 224002, People's Republic of 3 Shanghai Research Center of Acupuncture and Meridian, Pudong, Shanghai 201203, People's Republic of chin 4 Cancer Hospital, Medical Center of Fudan University, Shanghai 200032, People's Republic of Beijing Synchrotron Radiation Facility, Institute of High Energy Physics, Chinese Academy of E-mail: xy-zhang@fudan.edu. Received 8 March 2006, in final form 20 August 2006 Published 29 December 2006 Online at stacks. iop. org/PMB/52/419 The significance of the x-ray diffraction enhanced imaging(DEI) technique in the diagnosis of breast cancer and its feasibility in clinical medical imaging are evaluated. Different massive specimens including normal breast tissues, benign breast tumour tissues and malignant breast tumour tissues are imaged with the DEl method. The images are recorded respectively by CCD or x-ray film at different positions of the rocking curve and processed with a pixel- by-pixel algorithm. The characteristics of the DEl images about the normal and diseased tissues are compared. The rocking curves of a double-cryst diffractometer with various tissues are also studied. The differences in del images and their rocking curves are evaluated for early diagnosis of breast (Some figures in this article are in colour only in the electronic version) 1. Introduction Breast diseases are common for women and breast cancer is a dangerous killer(bothorel et al 1997). Early diagnosis and treatment are the best ways to reduce the hazard. Current procedures for the diagnosis of breast cancer mainly include self-examination, mammography (Feig 2002, Freedman et al 2003, Humphrey et al 2002), ultrasonography(US)(Shankar 0031-9155/07/020419+09S3000 2007 1OP Publishing Ltd Printed in the UK
INSTITUTE OF PHYSICS PUBLISHING PHYSICS IN MEDICINE AND BIOLOGY Phys. Med. Biol. 52 (2007) 419–427 doi:10.1088/0031-9155/52/2/008 Evaluation of x-ray diffraction enhanced imaging in the diagnosis of breast cancer Chenglin Liu1,2, Xiaohui Yan1 , Xinyi Zhang1,3, Wentao Yang4 , Weijun Peng4 , Daren Shi4 , Peiping Zhu5 , Wanxia Huang5 and Qingxi Yuan5 1 Synchrotron Radiation Research Center, Physics Department, Surface Physics Laboratory (National Key Laboratory) of Fudan University, Shanghai 200433, People’s Republic of China 2 Physics Department of Yancheng Teachers’ College, Yancheng 224002, People’s Republic of China 3 Shanghai Research Center of Acupuncture and Meridian, Pudong, Shanghai 201203, People’s Republic of China 4 Cancer Hospital, Medical Center of Fudan University, Shanghai 200032, People’s Republic of China 5 Beijing Synchrotron Radiation Facility, Institute of High Energy Physics, Chinese Academy of Sciences, Beijing 100039, People’s Republic of China E-mail: xy-zhang@fudan.edu.cn Received 8 March 2006, in final form 20 August 2006 Published 29 December 2006 Online at stacks.iop.org/PMB/52/419 Abstract The significance of the x-ray diffraction enhanced imaging (DEI) technique in the diagnosis of breast cancer and its feasibility in clinical medical imaging are evaluated. Different massive specimens including normal breast tissues, benign breast tumour tissues and malignant breast tumour tissues are imaged with the DEI method. The images are recorded respectively by CCD or x-ray film at different positions of the rocking curve and processed with a pixelby-pixel algorithm. The characteristics of the DEI images about the normal and diseased tissues are compared. The rocking curves of a double-crystal diffractometer with various tissues are also studied. The differences in DEI images and their rocking curves are evaluated for early diagnosis of breast cancers. (Some figures in this article are in colour only in the electronic version) 1. Introduction Breast diseases are common for women and breast cancer is a dangerous killer (Bothorel et al 1997). Early diagnosis and treatment are the best ways to reduce the hazard. Current procedures for the diagnosis of breast cancer mainly include self-examination, mammography (Feig 2002, Freedman et al 2003, Humphrey et al 2002), ultrasonography (US) (Shankar 0031-9155/07/020419+09$30.00 © 2007 IOP Publishing Ltd Printed in the UK 419
C Liu et al et al 1993)and magnetic resonance imaging(MRI(Orel et al 1997). Mammography is currently recognized as the gold standard for the early detection of breast cancer because it requires lower radiation doses, though the rate of false negatives is still as high as 10 per cent (Wilkinson et al 2005). Recently, some new techniques, such as digital mammography, computer aided diagnosis, synchrotron mammography and digital tomosynthesis have been developed to optimize image quality and improve the diagnostic capabilities( Giammarile and Bremond 2004, Niklason et al 1997). The synchrotron radiation, which is electromagnetic by charged particles moving with a speed very close to light speed in a circular orbit, was first found in 1947 and then widely used in different fields, including material science, biology, medicine(Margaritondo and Meuli 2003, Suortti and Thomlinson 2003), etc. Images obtained by using synchrotron radiation with higher spatial resolution and better contrast( Fiedler et al 2003)are superior to those with a conventional x-ray source. With synchrotron radiation, some microstructures inside organic tissues can be shown, which are often missed with conventional techniques Since the middle of the 1990s, x-ray phase-contrast imaging techniques, which can clearly show the microstructure of biological soft tissues under low radiation dose, have been quickly developed (Lewis et al 2002). Diffraction enhanced imaging (DED, whose contrast comes from the absorption, the refractive gradient and the small-angle scattering rejection(usually called extinction), is one of them(Hasnah et al 2002, Zhong et al 2000). The DEl method enhances the sharpness and clarity of the soft tissue image relative to the traditional method due to the combination of more contrast mechanisms(Menk 1999). In recent years, DEI has shown some latent applications in early diagnosis(Lewis 2004). The spatial resolution of DEl images is higher than that of B-mode ultrasound, MRI or CT(computer tomography), and reaches the order of micrometres (Liu et al 2005). The combination of DEl and ct (also called DEI-CT)makes medical images of tumours extraordinarily similar to pathological histology ( Fiedler et al 2004). Moreover, due to the high intensity of synchrotron radiation, only a short exposure time(see below) is needed. Therefore, synchrotron radiation based DEl has a low risk of complications due to the radiation exposure which is obviously an advantage in the early diagnosis of cancer. 6. In this paper, we compared various DEI images of the tissues for the normal breast, benign breast tumour and malignant breast tumour along with their rocking curves to distinguish the microstructures inside those tissues, and we also discussed the merits of this technique for medical applications 2. Materials and methods The specimens were prepared in the Cancer Hospital, Medical Center of Fudan University. We selected 13 of them, including four cases of normal breast tissue, four cases of benign breast tumour tissue and five cases of breast cancer tissue. Both the diseased tissues and the healthy ones were taken from mastectomy sufferers. The main part of each cancer tissue included a great mass of cancer nests, and the normal tissue was taken from a distance of about 2 cm beside the tumour. These specimens were cut into small pieces of 10 x9 mm with a thickness of approximately 2 mm and fixed in 10% buffered formalin. The DEl experiments have been carried out at the topography end station of the 4WiA beamline of the Beijing Synchrotron Radiation Facility(BSRF). The energy of the monochromatic x-rays was 9 keV with an incident angle of 12.7, and the maximum size of the light spot was approximately 12 x 10 mm" at the position of specimen. The analyser was fixed at the axis so it could be tuned to get the different positions of the rocking curve. The DEl images were recorded using a CCD with 1300 x 1030 pixels(X-ray Fast Digital Imager
420 C Liu et al et al 1993) and magnetic resonance imaging (MRI) (Orel et al 1997). Mammography is currently recognized as the gold standard for the early detection of breast cancer because it requires lower radiation doses, though the rate of false negatives is still as high as 10 per cent (Wilkinson et al 2005). Recently, some new techniques, such as digital mammography, computer aided diagnosis, synchrotron mammography and digital tomosynthesis have been developed to optimize image quality and improve the diagnostic capabilities (Giammarile and Bremond 2004, Niklason et al 1997). The synchrotron radiation, which is electromagnetic radiation emitted by charged particles moving with a speed very close to light speed in a circular orbit, was first found in 1947 and then widely used in different fields, including material science, biology, medicine (Margaritondo and Meuli 2003, Suortti and Thomlinson 2003), etc. Images obtained by using synchrotron radiation with higher spatial resolution and better contrast (Fiedler et al 2003) are superior to those with a conventional x-ray source. With synchrotron radiation, some microstructures inside organic tissues can be shown, which are often missed with conventional techniques. Since the middle of the 1990s, x-ray phase-contrast imaging techniques, which can clearly show the microstructure of biological soft tissues under low radiation dose, have been quickly developed (Lewis et al 2002). Diffraction enhanced imaging (DEI), whose contrast comes from the absorption, the refractive gradient and the small-angle scattering rejection (usually called extinction), is one of them (Hasnah et al 2002, Zhong et al 2000). The DEI method enhances the sharpness and clarity of the soft tissue image relative to the traditional method due to the combination of more contrast mechanisms (Menk 1999). In recent years, DEI has shown some latent applications in early diagnosis (Lewis 2004). The spatial resolution of DEI images is higher than that of B-mode ultrasound, MRI or CT (computer tomography), and reaches the order of micrometres (Liu et al 2005). The combination of DEI and CT (also called DEI-CT) makes medical images of tumours extraordinarily similar to pathological histology (Fiedler et al 2004). Moreover, due to the high intensity of synchrotron radiation, only a short exposure time (see below) is needed. Therefore, synchrotron radiation based DEI has a low risk of complications due to the radiation exposure which is obviously an advantage in the early diagnosis of cancer. In this paper, we compared various DEI images of the tissues for the normal breast, benign breast tumour and malignant breast tumour along with their rocking curves to distinguish the microstructures inside those tissues, and we also discussed the merits of this technique for medical applications. 2. Materials and methods The specimens were prepared in the Cancer Hospital, Medical Center of Fudan University. We selected 13 of them, including four cases of normal breast tissue, four cases of benign breast tumour tissue and five cases of breast cancer tissue. Both the diseased tissues and the healthy ones were taken from mastectomy sufferers. The main part of each cancer tissue included a great mass of cancer nests, and the normal tissue was taken from a distance of about 2 cm beside the tumour. These specimens were cut into small pieces of 10 × 9 mm with a thickness of approximately 2 mm and fixed in 10% buffered formalin. The DEI experiments have been carried out at the topography end station of the 4W1A beamline of the Beijing Synchrotron Radiation Facility (BSRF). The energy of the monochromatic x-rays was 9 keV with an incident angle of 12.7◦, and the maximum size of the light spot was approximately 12 × 10 mm2 at the position of specimen. The analyser was fixed at the axis so it could be tuned to get the different positions of the rocking curve. The DEI images were recorded using a CCD with 1300 × 1030 pixels (X-ray Fast Digital Imager
Evaluation of x-ray diffraction enhanced imaging in the diagnosis of breast cancer 18 mm system, Photonic-science Ltd, UK) or industrial x-ray film(Fuji IX80)behind the analyser crystal. The spatial resolutions of the CCD and x-ray film(pixel sizes)were 10.9 um nd 2.3 um, respectively. The distance between the specimen and CCd detector was about I m. The photon flux incident onto the specimens is proportional to the beam current. Hence we can use the product of beam current multiplied by the exposure time as the exposure dose. In our experiments the exposure dose was controlled as a constant, whose average values were about 26.4 mA s and 360 mA s for the CCD and x-ray film, respectively. The beam current was normally between 50 and 100 mA and the exposure time ranged from less than 1 s to several second For the DEl setup in the bse, two Si (111) crystals are equipped. One serves as a monochrometer and the other one serves as the analyser. These two crystals formed a double-crystal diffractometer. The rocking curve, which is the relationship between the reflected intensity and deflection angle of the analyser, can be obtained by tuning the analyser ontinuously to different angles related to the Bragg angle of Si(111). The variation of the rocking curve with different breast tissues has been carefully studied When the Ccd is used as the detector three kinds of images can be obtained. The first one is recorded at the peak position of the rocking curve, which is usually called the peak image. The other two images are taken when the analyser is tuned to the FWHM full-width of half-maximum) positions on the either side of the rocking curve. These two images contain the same absorption and little scattering information but opposite refraction information. When these two images are added pixel by pixel, the refraction information will be eliminated, thus we can obtain an image called the apparent absorption image because of the pure absorption contrast(but with weak extinction). When two images are subtracted pixel by pixel, the so-called refraction image can be obtained, in which not only the refraction but also the absorption contrast is included. And consequently the edge effect will be greatl enhanced. This image processing is very easy because only a simple pixel-by-pixel algorithm is needed In addition, we also recorded the images by x-ray films and read them out through an optical microscope. In this case two different imaging setups were used. One is to place the x-ray film about 1 cm behind the sample without using the analyser crystal, and record the image, which is called absorption image, resembling conventional mammography. The other setup is to place the x-ray film just behind the analyser crystal which is in the top position of its rocking curve, and obtain the peak image. These images will display some microstructures of various tissues after they are read out by the optical microscope 3. Result In our DEl experiments, we obtained various images from those tissues. The various normal tissue (or benign or malignant tumour tissue) cases had similar characteristics in their DEl images. Therefore, only one set of DEI images from each case is shown here. We compared these images to evaluate the significance of the DEl method and show the differences between the normal, benign and malignant tissues The peak image recorded by CCD of these three different kinds of tissues, normal benign tumour and malignant tumour, are shown in figure 1. They contain the information of absorption and small-angle scattering rejection(extinction). The apparent absorption images (figure 2) and the refraction images(figure 3)are obtained by the simple pixel-by-pixel algorithm( Chapman et al 1997)using two images obtained as the analyser is tuned to either side of the FWHM positions, i.e. the low-angle position and the high-angle position of the ocking curve, as mentioned above. The refraction image is extraordinarily sensitive to the
Evaluation of x-ray diffraction enhanced imaging in the diagnosis of breast cancer 421 18 mm system, Photonic-science Ltd, UK) or industrial x-ray film (Fuji IX80) behind the analyser crystal. The spatial resolutions of the CCD and x-ray film (pixel sizes) were 10.9 µm and 2.3 µm, respectively. The distance between the specimen and CCD detector was about 1 m. The photon flux incident onto the specimens is proportional to the beam current. Hence we can use the product of beam current multiplied by the exposure time as the exposure dose. In our experiments the exposure dose was controlled as a constant, whose average values were about 26.4 mA s and 360 mA s for the CCD and x-ray film, respectively. The beam current was normally between 50 and 100 mA and the exposure time ranged from less than 1 s to several seconds. For the DEI setup in the BSRF, two Si (1 1 1) crystals are equipped. One serves as a monochrometer and the other one serves as the analyser. These two crystals formed a double-crystal diffractometer. The rocking curve, which is the relationship between the reflected intensity and deflection angle of the analyser, can be obtained by tuning the analyser continuously to different angles related to the Bragg angle of Si (1 1 1). The variation of the rocking curve with different breast tissues has been carefully studied. When the CCD is used as the detector, three kinds of images can be obtained. The first one is recorded at the peak position of the rocking curve, which is usually called the peak image. The other two images are taken when the analyser is tuned to the FWHM (full-width of half-maximum) positions on the either side of the rocking curve. These two images contain the same absorption and little scattering information but opposite refraction information. When these two images are added pixel by pixel, the refraction information will be eliminated, thus we can obtain an image called the apparent absorption image because of the pure absorption contrast (but with weak extinction). When two images are subtracted pixel by pixel, the so-called refraction image can be obtained, in which not only the refraction but also the absorption contrast is included. And consequently the edge effect will be greatly enhanced. This image processing is very easy because only a simple pixel-by-pixel algorithm is needed. In addition, we also recorded the images by x-ray films and read them out through an optical microscope. In this case two different imaging setups were used. One is to place the x-ray film about 1 cm behind the sample without using the analyser crystal, and record the image, which is called absorption image, resembling conventional mammography. The other setup is to place the x-ray film just behind the analyser crystal which is in the top position of its rocking curve, and obtain the peak image. These images will display some microstructures of various tissues after they are read out by the optical microscope. 3. Results In our DEI experiments, we obtained various images from those tissues. The various normal tissue (or benign or malignant tumour tissue) cases had similar characteristics in their DEI images. Therefore, only one set of DEI images from each case is shown here. We compared these images to evaluate the significance of the DEI method and show the differences between the normal, benign and malignant tissues. The peak image recorded by CCD of these three different kinds of tissues, normal, benign tumour and malignant tumour, are shown in figure 1. They contain the information of absorption and small-angle scattering rejection (extinction). The apparent absorption images (figure 2) and the refraction images (figure 3) are obtained by the simple pixel-by-pixel algorithm (Chapman et al 1997) using two images obtained as the analyser is tuned to either side of the FWHM positions, i.e. the low-angle position and the high-angle position of the rocking curve, as mentioned above. The refraction image is extraordinarily sensitive to the
C Figure 1. Peak images which are recorded by CCD at the peak position of rocking curves of different breast tissues (A)normal. (B) benign tumour, and (C) malignant tumour). Figure 2. Apparent absorption images of different breast tissues. These images are obtained by adding two raw images pixel by pixel. The raw images are recorded by CCD when the analyser is tuned to either side of the FWHM positions.(A)normal, (B)benign tumour, and ( C)malignant Figure 3. Refraction images of different breast tissues. These images raw images by pixel by pixel. The raw images are recorded by CCD when the analyser is tuned to either side of the FWHM positions.(A)normal, (B)benign tumour, and(C)malignant variation of the refractive index of the sample. A few microstructures of tissues can be seen in those images. Especially, there are evidently differences in the calcification between the normal tissues, benign and malignant tumour tissues(see discussion section below). In order to compare diffraction enhanced imaging with the conventional radiographs, x-ray films were used as the detector. The peak images were recorded by x-ray films as shown in figure 4. The absorption images are shown in figure 5, and as we mentioned earlier, in this case the analyser crystal was not used. The contrast of the absorption images is lower than that of the peak images due to the strong scattering which will reduce the image's quality When the peak images are magnified by the optical microscope, some microstructures can be clearly seen. For example, we can see bundles which consist of fibres of about 30 um in diameter(denoted by the black arrow) and some reticulate structures(by the white arrow) in figure 6(A). Some small cavities are shown by arrows in figure 6(B). In the same way, the
422 C Liu et al (A) (B) (C) Figure 1. Peak images which are recorded by CCD at the peak position of rocking curves of different breast tissues (A) normal, (B) benign tumour, and (C) malignant tumour). (A) (B) (C) Figure 2. Apparent absorption images of different breast tissues. These images are obtained by adding two raw images pixel by pixel. The raw images are recorded by CCD when the analyser is tuned to either side of the FWHM positions. (A) normal, (B) benign tumour, and (C) malignant tumour). (A) (B) (C) Figure 3. Refraction images of different breast tissues. These images are obtained by subtracting two raw images by pixel by pixel. The raw images are recorded by CCD when the analyser is tuned to either side of the FWHM positions. (A) normal, (B) benign tumour, and (C) malignant tumour). variation of the refractive index of the sample. A few microstructures of tissues can be seen in those images. Especially, there are evidently differences in the calcification between the normal tissues, benign and malignant tumour tissues (see discussion section below). In order to compare diffraction enhanced imaging with the conventional radiographs, x-ray films were used as the detector. The peak images were recorded by x-ray films as shown in figure 4. The absorption images are shown in figure 5, and as we mentioned earlier, in this case the analyser crystal was not used. The contrast of the absorption images is lower than that of the peak images due to the strong scattering which will reduce the image’s quality. When the peak images are magnified by the optical microscope, some microstructures can be clearly seen. For example, we can see bundles which consist of fibres of about 30 µm in diameter (denoted by the black arrow) and some reticulate structures (by the white arrow) in figure 6(A). Some small cavities are shown by arrows in figure 6(B). In the same way, the
Evaluation of x-ray diffraction enhanced imaging in the diagnosis of breast cancer Figure 4. Peak images which are recorded by x-ray films at the peak position of rocking curves of different breast tissues(A)normal, (B)benign tumour, and(C)malignant tumour) Figure 5. Absorption images which are recorded by x-ray films of different breast tissues (A)normal. (B)benign tumour, and (C)malignant tumour). Figure 6. Peak images of various breast tissues recorded by industrial x-ray films at the peak position of the rocking curve and magnified by an optical microscope. These images can show the microstructures of various breast tissues(A) normal, (B)benign and(C) malignant irregular conglomeration(pointed out by the white arrow) and the expanded duct(by the black arrow) can be also observed, as shown in figure 6(C) On the other hand, the rocking curve can also show the differences of various tissues in the DEl method. Figure 7 shows the rocking curves of the double-crystal diffractometer itself with different breast tissues. Comparing these rocking curves, the diversities of peak displacement and FWHM are not obvious. However, the intensity and integrated intensity of reflectivity can show definite differences. These values for each case and their average values are collected in table 1 4. Discussion 4.1. Peak image, apparent absorption image and refraction image There are three kinds of images which can be used in the diagnosis of breast tumours. These images are peak images, apparent absorption images and refraction images. The peak image
Evaluation of x-ray diffraction enhanced imaging in the diagnosis of breast cancer 423 (A) (B) (C) Figure 4. Peak images which are recorded by x-ray films at the peak position of rocking curves of different breast tissues (A) normal, (B) benign tumour, and (C) malignant tumour). (A) (B) (C) Figure 5. Absorption images which are recorded by x-ray films of different breast tissues (A) normal, (B) benign tumour, and (C) malignant tumour). (A) (B) (C) Figure 6. Peak images of various breast tissues recorded by industrial x-ray films at the peak position of the rocking curve and magnified by an optical microscope. These images can show the microstructures of various breast tissues (A) normal, (B) benign tumour, and (C) malignant tumour). irregular conglomeration (pointed out by the white arrow) and the expanded duct (by the black arrow) can be also observed, as shown in figure 6(C). On the other hand, the rocking curve can also show the differences of various tissues in the DEI method. Figure 7 shows the rocking curves of the double-crystal diffractometer itself with different breast tissues. Comparing these rocking curves, the diversities of peak displacement and FWHM are not obvious. However, the intensity and integrated intensity of reflectivity can show definite differences. These values for each case and their average values are collected in table 1. 4. Discussion 4.1. Peak image, apparent absorption image and refraction image There are three kinds of images which can be used in the diagnosis of breast tumours. These images are peak images, apparent absorption images and refraction images. The peak image