Radiological Physics and Technology

, Volume 2, Issue 1, pp 77–86

Embossed radiography utilizing energy subtraction


    • The 3rd Department of SurgeryToho University School of Medicine
  • Manabu Watanabe
    • The 3rd Department of SurgeryToho University School of Medicine
  • Eiichi Sato
    • Department of PhysicsIwate Medical University
  • Hiroshi Matsukiyo
    • The 3rd Department of SurgeryToho University School of Medicine
  • Toshiyuki Enomoto
    • The 3rd Department of SurgeryToho University School of Medicine
  • Jiro Nagao
    • The 3rd Department of SurgeryToho University School of Medicine
  • Purkhet Abderyim
    • Department of Computer and Information Sciences, Faculty of EngineeringIwate University
  • Katsuo Aizawa
    • Tokyo Medical University
  • Etsuro Tanaka
    • Department of Nutritional Science, Faculty of Applied Bio-scienceTokyo University of Agriculture
  • Hidezo Mori
    • Department of Cardiac PhysiologyNational Cardiovascular Center Research Institute
  • Toshiaki Kawai
    • Electron Tube Division #2Hamamatsu Photonics K.K
  • Shigeru Ehara
    • Department of RadiologySchool of Medicine, Iwate Medical University
  • Shigehiro Sato
    • Department of MicrobiologySchool of Medicine, Iwate Medical University
  • Akira Ogawa
    • Department of NeurosurgerySchool of Medicine, Iwate Medical University
  • Jun Onagawa
    • Department of Electronics, Faculty of EngineeringTohoku Gakuin University

DOI: 10.1007/s12194-008-0048-8

Cite this article as:
Osawa, A., Watanabe, M., Sato, E. et al. Radiol Phys Technol (2009) 2: 77. doi:10.1007/s12194-008-0048-8


Currently, it is difficult to carry out refraction-contrast radiography by using a conventional X-ray generator. Thus, we developed an embossed radiography system utilizing dual-energy subtraction for decreasing the absorption contrast in unnecessary regions, and the contrast resolution of a target region was increased by use of image-shifting subtraction and a linear-contrast system in a flat panel detector (FPD). The X-ray generator had a 100-μm-focus tube. Energy subtraction was performed at tube voltages of 45 and 65 kV, a tube current of 0.50 mA, and an X-ray exposure time of 5.0 s. A 1.0-mm-thick aluminum filter was used for absorbing low-photon-energy bremsstrahlung X-rays. Embossed radiography was achieved with cohesion imaging by use of the FPD with pixel sizes of 48 × 48 μm, and the shifting dimension of an object in the horizontal direction ranged from 100 to 200 μm. At a shifting distance of 100 μm, the spatial resolutions in the horizontal and vertical directions measured with a lead test chart were both 83 μm. In embossed radiography of non-living animals, we obtained high-contrast embossed images of fine bones, gadolinium oxide particles in the kidney, and coronary arteries approximately 100 μm in diameter.


Embossed radiographyDigital subtractionEnergy subtractionContrast resolutionPolychromatic X-rays


Extremely clean monochromatic parallel X-ray beams have been formed by use of synchrotrons and single silicon crystals, and these beams have been applied to carrying out enhanced iodine K-edge angiography [13], phase-contrast radiography [46], and topography [7, 8]. For performing K-edge angiography, monochromatic X-rays with energies just beyond the iodine K-edge of 33.2 keV have been used, because these rays are absorbed effectively by iodine-based contrast media. Currently, phase-contrast radiography is based primarily on X-ray refraction in objects, and soft tissues, such as breast cancers [6], can be imaged with high contrast. However, it is difficult to carry out phase-contrast imaging of contrast media for medical angiography, delivered nano-particles, and hard tissues. Two-dimensional X-ray topography also is a method for imaging defects in crystals by utilizing X-ray diffraction.

Without the use of synchrotrons, several different monochromatic X-ray sources [912] have been developed, and K-edge angiography, phase radiography, and topography have been carried out. In particular, a cerium X-ray generator [13, 14] has been developed and has been applied to carrying out cone-beam K-edge angiography because cerium Kα rays (34.6 keV) are absorbed effectively by iodine media. Next, phase radiography for edge enhancement of objects has been performed by magnification radiography with a microfocus X-ray generator and polychromatic X-rays [15]. However, this conventional phase radiography is usable for imaging only of soft tissues, and a microfocus X-ray tube is necessary. Therefore, we are very interested in the development of a novel radiography system instead of the phase imaging by use of a generalized digital X-ray image sensor and a conventional large-focus tungsten tube that produces polychromatic X-rays.

Energy subtraction radiography [16, 17] is an important technique for imaging target regions in vivo by removing muscle and bone regions from radiograms. Currently, two different energy radiograms are obtained with two different tube voltages, and dual-energy subtraction radiography is carried out by digital subtraction between the high- and low-energy images. Furthermore, embossed radiography (ER) [18] is a novel technique for constructing concavoconvex radiography, such as phase-differential imaging, and is realizable with digital image subtraction after image shifting with an optimal dimension between two images. The image shifting is carried out by moving an X-ray source [18], by moving an object precisely, and by shifting image pixels in a flat panel detector (FPD) with use of a computer program before subtraction. By use of this radiography, the target region in various objects can be imaged with embossing, and the maximum contrast resolution is achieved without a drop in spatial resolution. In addition, our ultimate goal of embossing is the development of a novel computed tomography (CT) system with a high contrast resolution for cancer diagnosis that utilizes a drug delivery system.

In our research, major objectives are as follows: construction of edge enhancement radiography such as phase-differential imaging utilizing embossing, an increase in the contrast resolution of the target region, and energy subtractions of contrast media and nano-particles. Therefore, we carried out preliminary experiments for ER utilizing two-shot dual-energy subtraction by shifting objects.

Experimental methods

Experimental setup for ER

Figure 1 shows an experimental setup for performing ER utilizing an FPD (1024EV, Rad-icon Imaging, Santa Clara, CA). This experiment is carried out for a fundamental study on single-dimensional ER, although two-dimensional shifting is realizable with a pixel-shifting computer program. ER was performed by digital subtraction, cohesion radiography, and shifting of an object with a dimension ranging from 100 to 200 μm and with a traveling microscope (NRM–2XZ, Shimadzu Rika, Tokyo, Japan) that had a length resolution of 10 μm. The computer program names for observing raw-file images and for carrying out subtraction radiography were ShadoCam (Rad-icon Imaging, Santa Clara, CA) and EnergyCalc Composition 1.0 (AD Science, Funabashi, Japan), respectively.
Fig. 1

Experimental setup for performing embossed radiography (ER) utilizing energy subtraction, where ξ is the image-shifting dimension in the FPD, df is the distance from the FPD to the position in the object, d0 is the distance between the FPD face and the pixel plane, and d is the distance between the source and the FPD face

For carrying out ER such as phase-differential imaging, the minimum shifting dimension is two times the pixel size of 48 μm, and the maximum shifting dimension is almost equal to an allowable spatial resolution. Absorption contrast of unnecessary regions is decreased by energy subtraction. An object is exposed by an X-ray generator with a 100-μm-focus X-ray tube, and the first radiogram is obtained at position 1 in Fig. 1. The second radiogram is taken at position 2 in Fig. 1 after shifting of the object. Two radiograms are taken using two different photon energy spectra with maximum photon energies of 45 and 65 keV, and are recorded as raw files. In this setup, the image shifting dimension ξ in the FPD is given by
$$ \xi = \sigma \left( {d + d_{0} } \right)/\left( {d - d_{f} } \right) \cong \sigma , $$
where σ is the shifting dimension of the object, df is the distance from the FPD to the position in the object, d0 (=10 mm) is the distance between the FPD face and the pixel plane, and d is the distance between the source and the FPD face. In our research, σ and d were determined to be 100–200 μm and 1.0 m, respectively. In dual-energy subtraction for embossing, an embossed raw-file Re(x, y) as a function of horizontal x and vertical y distances in the FPD is calculated by
$$ R_{e} \left( {x,y} \right) = R_{1} \left( {x,y} \right) - KR_{2} \left( {x - \xi ,y} \right), $$
where R1(x, y) is the first raw-file image at position 1, R2(x − ξ, y) is the second raw-file image with different energy spectra at position 2 after shifting of the object, K (0 < ≤ 1.0) is a constant for subtraction, and K varies by 0.1 increments.
Typical one-dimensional projection curves for ER are shown in Fig. 2. R1(x) is the first projection curve from the FPD, KR2(x − ξ) is the shifted curve for subtraction, and Re(x) is the embossed projection curve in the raw file system. In this experiment, an embossed projection curve T(x) and its inverted curve Ti(x) in the TIFF file system are written as
$$ T\left( x \right) = \left\{ {R_{1} \left( x \right) - KR_{2} \left( {x - \xi } \right) - R_{ \min } } \right\} \times D_{\text{m}} /\left( {R_{ \max } - R_{ \min } } \right), $$
$$ T_{i} \left( x \right) = D_{\text{m}} - T\left( x \right), $$
where Dm (= 255) is the maximum density, and Rmax and Rmin are the maximum and minimum densities of Re(x), respectively. In the FPD, the densities of 0 and Dm correspond to black and white, respectively. Therefore, the contrast resolution increases to 1 after subtraction by a linear-contrast system. The matrix size is about 1 M (1,048 × 1,048) pixels, and the pixel sizes are 48 μm × 48 μm. The gray scale ranges from 0 to 255, and the relation between the X-ray intensity and the gray level is linear.
Fig. 2

One-dimensional projection curves for performing ER utilizing subtraction, where ξ the image shifting dimension in the FPD, Re(x) is the embossed projection curve in the raw file system, R1(x) is the first projection curve from the FPD, KR2(x − ξ) is the shifted curve for subtraction, K is a constant, T(x) is an embossed projection curve in the TIFF file system, Ti(x) is its inverted curve, and Rmax and Rmin are the maximum and minimum densities of Re(x), respectively

X-ray generator

Figure 3 shows the circuit diagram of an X-ray generator that was designed for regulating the focal size of an X-ray tube (Toshiba, 1–311) by controlling the negative bias voltage of the focusing electrode. In this experiment, the size was regulated as 100 μm at a bias voltage of −20 V because the size decreased with increasing negative bias voltage, and tube voltage, current, and exposure time could be controlled by a main controller. The high-voltage line employs the Cockcroft-Walton circuit, and positive and negative high voltages are applied to the anode and cathode electrodes, respectively. The filament heating current is supplied by an AC power supply with an insulation transformer that is used for isolation from the high voltage from the Cockcroft–Walton circuit. For measurement of X-ray intensity, the tube voltage ranged from 40 to 70 kV, and the tube current was regulated to be 0.50 mA. Because the maximum exposure time of the FPD was 6.0 s, the time was set as 5.0 s, which increased the X-ray intensity for radiography.
Fig. 3

Circuit diagram of a 100-μm-focus X-ray generator

Measurement of X-ray intensity

The measurement of X-ray intensity for ER is important because the intensity is decreased by the subtraction. The X-ray intensity was measured with an ionization chamber (660, Victoreen, Solon, OH) with a 400-cm3 volume probe (660-5, Victoreen, Solon, OH) with a tube current of 0.50 mA at 1.0 m from the X-ray source. For performing ER utilizing dual-energy subtraction, the total X-ray intensity of the two shots is equal to the exposed dose Iex for patients. Thus, Iex and the X-ray intensity after the subtraction Is are given by
$$ I_{\text{ex}} = I_{ 1} + I_{ 2} , $$
$$ I_{s} = I_{1} - KI_{2} , $$
where I1 and I2 are the X-ray intensities at positions 1 and 2, respectively. Roughly speaking, the absorption contrast decreases when Is decreases. Moreover, the image contrast reverses in a case where Is is a negative value.

Measurement of X-ray spectra

The image contrast varies corresponding to the X-ray spectra for radiography, and we have to measure the spectral distributions before energy subtraction. In order to measure X-ray spectra, we employed a cadmium telluride (CdTe) detector (XR-100T, Amptek, Bedford, MA) and measured spectra at 1.0 m from the X-ray source. The cooled detector unit with a charge amplifier was set at 1.0 m from the X-ray source, and event signals from the detector unit were amplified again by a shaping amplifier unit. A 0.1-mm-diameter lead pinhole was set in front of the detector facing the X-ray source to decrease the photon count rate. The photon energy was discriminated by a multichannel analyzer (MCA-8000A, Amptek, Bedford, MA), and the X-ray spectra were observed on a personal computer monitor. Measurement results for the X-ray spectra are important for carrying out dual-energy subtraction because the spectral distribution is the X-ray intensity as a function of the photon energy. In particular, when we perform K-edge subtraction, we have to confirm the spectra with energies below and beyond the K-edge energy, and the spectra should be controlled to optimal distributions by selection of the tube voltage, the element of the filter, and its filter thickness.

Method for ER utilizing single-energy subtraction

Although unnecessary regions are not deleted by single-energy subtraction, we carried out single-energy ER to observe the effect of embossing. Two radiograms were taken at positions 1 and 2 at a tube voltage of 45 kV by use of the filter, and the embossing was performed by image shifting. The embossed raw-file radiogram Re(x, y) was calculated by
$$ R_{e} \left( {x,y} \right) = R_{ 4 5} \left( {x,y} \right) - KR_{ 4 5} \left( {x - \xi ,y} \right), $$
where R45(x, y) is a radiogram obtained with a tube voltage of 45 kV and R45(x – ξ, y) is a radiogram at a voltage of 45 kV after shifting of the object.

The X-ray exposure time was 5.0 s, and two embossed radiograms were obtained with K values of 1.0 and 0.7. At a K of 1.0, the absorption contrast is removed, and the embossed effect is confirmed easily. When K is decreased, the absorption contrast increases, and the embossed effect decreases. Because the absorption contrast seldom varied in a K range from 0.5 to 0.7, K was determined as 0.7 for performance of single-energy ER with absorption contrast.

Method for ER utilizing dual-energy subtraction

As compared with single-energy ER, the target regions are imaged effectively by dual-energy ER. Because the K edges of iodine and gadolinium are 33.2 and 50.3 keV, respectively, a high-energy radiogram should be taken at an energy region beyond 50.3 keV. The minimum difference between the high and low tube voltages was approximately 20 kV. Therefore, two different energy radiograms were obtained with tube voltages of 45 and 65 kV with use of the filter, and dual-energy ER was performed by image shifting. Utilizing energy subtraction, Re(x,y) was written as
$$ R_{e} \left( {x,y} \right) = R_{ 4 5} \left( {x,y} \right) - KR_{ 6 5} \left( {x - \xi ,y} \right), $$
where R65(x − ξ, y) is a radiogram with a voltage of 65 kV after shifting of the object. For an increase in the embossed effect with high contrast, K is determined to be within a range from 0.7 to 1.0.

Measurement of spatial resolution

The spatial resolution of the horizontal and vertical directions was measured with a cuneal lead test chart. The resolution was measured for line pairs L per 1.0 mm. Therefore, the spatial resolution S (μm) is given by
$$ S = 1,000/2L. $$
Thus, when five line pairs (LP) are seen, the spatial resolution is determined as 100 μm.

Making of gadolinium oxide particles

A gadolinium oxide suspension for injection was made in a high-pressure machine (Starburst Mini, Sugino Machine, Namerikawa, Japan) for dispersing micro-particles. The high-pressure dispersing was carried out five times with use of physical saline, and the average particle diameter and the density of gadolinium oxide were 700 nm and 20%, respectively.

Objects for ER

For performing ER, we used five objects as follows: a nonliving nude mouse with a weight of 15 g, a lead test chart (PTW, L659035), an extracted rabbit kidney (8 g), an extracted rabbit heart (11 g), and a dried pig vertebra (22 g). The mouse, kidney, and heart were preserved in 10% formalin solution. The kidney and heart were extracted from two different rabbits with weights of approximately 2 kg after anesthetic (5% Nembutal) injection of 1.8 ml in an ear. A renal pelvis in the kidney was filled with gadolinium oxide suspension of 0.3 ml as described above by injection, and coronary arteries were filled with iodine-based microspheres. When a 20%-microsphere suspension by use of physical saline was injected into the arteries, the saline penetrated capillaries with diameters below 10 μm and reached the veins. Thus, the 37%-iodine microspheres 15 μm in diameter remained in the arteries.


X-ray intensity

When the tube voltage was increased, the X-ray intensity after penetrating a 1.0-mm-thick aluminum filter increased (Fig. 4). At a tube voltage 65 kV, the X-ray intensity was 69.0 μGy/s, with errors of less than 0.2%. For carrying out dual-energy ER, the exposure time was 5.0 s, and the low and high tube voltages were 45 and 65 kV, respectively. Thus, the total exposed dose for patients was approximately 0.5 mGy. In our experimental results for chest radiography with a computed radiography (CR) system, because the exposure dose ranged from 0.2 to 1.0 mGy, these values were almost equal to the dose in dual-energy ER.
Fig. 4

X-ray intensities with a 1.0-mm-thick aluminum filter at 1.0 m from the X-ray source and a tube current of 0.5 mA

X-ray spectra

When the tube voltage was increased, both the maximum photon energy and the spectral peak energy increased (Fig. 5). When the filter was used for absorbing soft bremsstrahlung rays, the peak energies at tube voltages of 45 and 65 kV were 23 and 27 keV, respectively, where the K-edge absorption of the CdTe detector was considered.
Fig. 5

X-ray spectra measured with a cadmium telluride detector and the filter at tube voltages of 45 and 65 kV

ER utilizing single-energy subtraction

Figure 6 shows ER of a nude mouse according to changes in K. As compared with normal radiography (Fig. 6a), the absorption contrast was removed at a K of 1.0 (Fig. 6b), and the contrast increased with decreasing K (Fig. 6c). With embossing, edge enhancement images of a mouse skull were obtained. In this single-energy ER, because the object was shifted horizontally, the vertical edges were effectively enhanced.
Fig. 6

Radiography of a nude-mouse head at a tube voltage of 45 kV. a Normal radiography, b embossed radiography (ER) at a K of 1.0, and c ER at a K of 0.7. Edge enhancement images of a skull were obtained in ER with use of single-energy subtraction

ER utilizing dual-energy subtraction

The spatial resolution of the horizontal and vertical directions was measured with a lead test chart (Fig. 7). In cohesion radiography, the spatial resolution of the horizontal and vertical directions was approximately 80 μm. ER was performed at a K of 0.7, and the spatial resolution fell with increasing shifting dimension in the horizontal direction. The horizontal spatial resolutions were 83 and 143 μm at the shifting dimensions of 100 and 200 μm, respectively (Fig. 7a). However, the resolution seldom varied with increasing shifting dimension, and 83-μm lines were visible in the vertical direction (Fig. 7b). Therefore, the shifting dimension should be controlled corresponding to the allowable spatial resolution, and two-dimensional image shifting is useful for enhancing of the target region.
Fig. 7

Radiography and ER of a lead test chart according to changes in the shifting distance in the horizontal direction. a Horizontal spatial resolutions and b vertical spatial resolution

Figure 8 shows a radiogram and an embossed radiogram of a region of the mouse abdomen. The embossed radiogram was calculated by a formula (8) with changing K. In radiography with a tube voltage of 45 kV, a mouse dung was barely visible. When K was increased in dual-energy ER, the absorption contrast of the muscle decreased slightly, and both the dung and the dorsal were observed clearly with embossing.
Fig. 8

Radiography of a mouse abdomen. Dung is clearly seen on ER

Radiography of a rabbit kidney is shown in Fig. 9, and gadolinium oxide particles 700 nm in average diameter were used for observing a renal pelvis in the kidney. As compared with a normal radiogram, the image contrast of the renal pelvis was improved by use of embossing. With increasing K, the absorption contrast of the muscle decreased too, and the contrast of the renal pelvis was improved. For imaging of the pelvis, an iodine-based contrast medium used in medical angiography and a cerium oxide suspension are also usable.
Fig. 9

Radiograms of a rabbit kidney. In ER, a renal pelvis is visible with gadolinium oxide nano-particles

Angiograms of an extracted rabbit heart with iodine-based microspheres 15 μm in diameter are shown in Fig. 10. According to increases in K with embossing, the contrast of heart muscles slightly decreased, and fine coronary arteries were clearly observed at K values ranging from 0.7 to 1.0. To observe fine blood vessels below 100 μm in diameter, the shifting dimension should be minimized to 100 μm.
Fig. 10

Radiography of a rabbit heart using iodine-based microspheres. Coronary arteries are clearly observed in ER

In radiography of a vertebra, the contrast resolution increased with embossing, and bone structures were visible (Fig. 11). In dual-energy ER for observing dry bones, because the image contrast decreased with increasing K, the single-energy ER was usable.
Fig. 11

Radiograms of a vertebra. With embossing, contrast resolution in the vertebra improves

Conclusion and outlook

We performed ER by utilizing energy subtraction and image shifting. In this method of radiography, concavoconvex images were obtained by shifting of the second image, and absorption contrasts of unnecessary regions were decreased by use of energy subtraction. As compared with conventional radiography, the contrast resolution of target regions increased to approximately 1 when we used a linear-contrast system. Therefore, it was easy to observe target regions as embossed images without decreasing the spatial resolution. Currently, although spatial resolution decreases with increasing shifting distance ξ, the minimum shifting distance ξmin should be kept at approximately 100 μm to enhance embossing and is written as
$$ \xi_{ \min } \cong 2s_{p} \le \xi \le \xi_{ \max } , $$
where sp is the pixel size of 48 μm and ξmax is the maximum shifting distance corresponding to the allowable spatial resolution.

To image a target region in vivo, we have to confine the X-ray spectra to optimal distributions for energy subtraction, and unnecessary regions for diagnosis should be deleted from radiograms. However, in cases where we carry out only embossment, single energy subtraction is usable. The tube voltage should be selected corresponding to radiographic objectives, and we have to use an optimal-element filter and to determine its effective thickness.

With this FPD, the exposure time was 5.0 s because the exposed dose rate from the 100-μm-focus tube was low owing to the maximum tube current of 0.5 mA. In addition, focal spot diameters below 1.0 mm are usable, and the exposure time decreases substantially with increasing spot diameter. Recently, a program for two-dimensional pixel shifting has been developed, and one-shot ER is realizable with single-energy subtraction.

Magnification radiography [19, 20] is useful for improving the spatial resolution in digital radiography. The spatial resolution improves with increasing magnification ratio, and the maximum ratio increases with decreasing focus diameter. Therefore, magnification ER utilizing energy subtraction can be performed, and a program for digital subtraction including a pixel-shifting function is useful for composing embossed images.

As compared with conventional radiography, the advantages of dual-energy ER are as follows: (1) contrast resolution can be increased up to approximately 1.0. (2) Edge-enhancement radiography such as phase-differential imaging is realizable. (3) Hard tissues are imaged effectively as compared with phase imaging. (4) Nano-particles and liquid contrast media in medical angiography can be observed as concavoconvex images with high contrast.

In human imaging, a computed radiography system is usable, and various applications will become possible as follows: angiography with iodine and gadolinium media, cancer diagnosis with nano-particles, hard-tissue (bone) imaging, and various radiographies with high contrast resolutions. In addition, an energy discriminating FPD system that can get spectral data is very useful for carrying out single-shot dual-energy subtraction with short exposure times.

This ER utilizing dual-energy subtraction is usable in various digital radiography systems, including a CR system. Therefore, we developed an ER program for CR utilizing a two-dimensional image-shifting function. However, it was difficult to carry out dual-energy ER precisely by using two imaging plates. We are currently developing a multi-slice mini-focus X-ray CT system utilizing a 100-μm-focus tube, and embossed tomography with use of single- and dual-energy subtractions would be employed for increasing the contrast resolution of the target region. In our research, energy subtraction has been effective for imaging iodine contrast media and gadolinium oxide nano-particles, and ER is useful for edge enhancement radiography like phase imaging and for increasing the contrast resolution without a decrease in the spatial resolution.


We would like to express our thanks to the reviewers for giving very helpful advice concerning our paper. This work was supported by Grants-in-Aid for Scientific Research and Advanced Medical Scientific Research from MECSST, Health and Labor Sciences Research Grants, Grants from the Keiryo Research Foundation, Promotion and Mutual Aid Corporation for Private Schools of Japan, the Japan Science and Technology Agency (JST), and the New Energy and Industrial Technology Development Organization (NEDO).

Copyright information

© Japanese Society of Radiological Technology and Japan Society of Medical Physics 2008