Digital Radiography: Overview of Common X-ray Detectors

Fabian Bauer, winter semester 2016/17

Digital radiography is the generic term for radiographic techniques that detect X-rays with digital sensors instead of photoplates or photographic films (analog radiography). One of the greatest advantages of a digital radiography system compared to an analog one is the possibility to store captured pictures and data digitally. As a consequence, digital picture archiving and distribution are far easier than before and subsequent image processing is made possible. Other benefits include a greater dynamic range and – when speaking of medical usage – a higher patient throughput, increased dose efficiency and even smaller X-ray exposure^[1][2]. Figure 1 shows the subclassification of related systems.

Computed radiography

Computed radiography (CR) systems are the ones closest to conventional radiography. But instead of a radiographic film, an image plate is used within the cassette of the X-ray machine. This plate is coated by a layer of so called storage phosphor – a photostimulable crystal that is made up of different halogenides (like bromide, chlorine or iodine compounds). During exposure the energy of incident photons is temporarily stored by the phosphor since electrons are lifted to higher energy levels. Subsequently the plate is removed from the cassette and a laser beam is pixelwise applied to the surface (flying spot scanner). During this process, the stored energy is released by emitting photons by photostimulated luminescence (PSL) of a wavelength different to the one of the laser, which is detected by photodiodes. Although the energy can be stored for several hours, the read-out process is usually accomplished directly after irradiation in order to prevent energy loses within the phosphor by spontaneous phosphorescence over time. Before its next use, the image plate is exposed to high-intensity white light in order to erase remaining higher state electrons. The complete process takes around 30-40 seconds^[1][3]. Since computed radiography devices are still based on cassettes, they can easily be integrated into existing X-ray systems without any need for profound modifications. Therefore they are also highly mobile and can be replaced uncomplicatedly. On the other hand, despite having also a wide dynamic range, CR systems are inferior to conventional screen-film combinations. Also, the image quality is worse in comparison with more modern digital detectors, such as flat-panel-based systems^[1].

Direct radiography

Direct radiography systems can be further subdivided into direct and indirect conversion systems. Although indirect systems are still used more frequently, the scope of ongoing research shifts to direct systems. Direct conversion systems are usually based on X-ray photoconductors which are capable of converting the inciding X-rays directly into an electric signal without the need for any further intermediate steps. Contrary, indirect conversion systems are based on two separate steps. First, a scintillator turns X-ray radiation into visible light which can then be detected by a photodetector in the following. In both designs the resulting information is sensed by an electronic readout mechanism that passes it to further processing^[4][5].

Indirect conversion

This class of sensors operates with two separate stages. First, the incident X-ray radiation is converted into visible light within a first layer, the so called scintillator. This produced amount of scintillator-photons is related to the original energy of the detected radiation. In a second step this visible light is translated into an electric charge by a sensor chip, which is read out subsequently. Both steps are linked by a scintillator-detector coupling^[4].

Scintillators

Incident X-rays that strike the scintillator interact with it and produce visible light (390 – 700 nm) due to photoluminescence. The number of generated light photons is proportional to the incident X-ray radiation. Therefore, measuring the height of photon pulses allows to draw conclusions about the spectral composition of the original radiation. Popular materials for scintillators are for instance NaI(Tl), CsI or gadulinium compounds^[4][6].

Some (e.g. CsI) scintillators exist in structured and unstructured implementations. Structured versions use a needle-shaped phosphor crystals (typically ca. 5-10 micrometers in diameter in case of CsI) in order to reduce lateral propagation of photons. This way the number of interactions is increased because thicker scintillators can be used and a better spacial resolution can be obtained^[1][3][4].

Scintillator-photodetector coupling

Several different approaches exist in order to couple the scintillating crystal to the photosensor. The losses within the coupling should be as low as possible in order to minimize signal noise, especially in combination with insufficient conversion gain of the scintillator. Conventional photodetectors are often limited to fixed sizes by reason of the fabrication process. In order to use the whole area of the sensor, coupling is supposed to (de)magnify the image emitted by the phosphor to cover the whole area of the sensor chip^[7].

Optical lens coupling

This easy coupling consists of one or several common optical lenses that magnify the image to the needed size. Disadvantages include a low signal-to-noise ratio and a low quantum efficiency due to photon losses within the coupling^[1]. For a single-lense system the coupling efficiency $//<![CDATA[ \begin{array}{l}ξ\end{array} //]]>$ can be calculated according to the following equation^[8]:

$//<![CDATA[ \begin{array}{l}\xi=\frac{\tau}{4F^2 (m+1)^2}\end{array} //]]>$

• $//<![CDATA[ \begin{array}{l}\tau\end{array} //]]>$: optical transmission factor

• $//<![CDATA[ \begin{array}{l}F\end{array} //]]>$: the f-number of the lens

• $//<![CDATA[ \begin{array}{l}m\end{array} //]]>$: demagnification factor

Fiber-optic coupling

Another possibility is the fiber-optic coupling. Here, a bundle of optical fibers is used as a light guide. In order to match the difference between the pixelsize on the screen and the geometry of the sensor chip, the diameter of one fiber is not constant but tapers into one direction. Furthermore, it is necessary to prevent crosstalk, i.e. light escaping out of one fiber can influence another one and lead to different signals in both pathways. Therefore an optically attenuating material, the extramural absorber (EMA) is placed between the fibers in order to minimize this effect. The coupling efficiency $//<![CDATA[ \begin{array}{l}ξ\end{array} //]]>$ is usually higher than for optical lense systems and can be calculated by:

$//<![CDATA[ \begin{array}{l}\xi=\alpha\tau(\Theta)\frac{NA^2}{m^2}\end{array} //]]>$

• $//<![CDATA[ \begin{array}{l}\tau\end{array} //]]>$: transmission factor of the core glass (dependent on the angle of incidence $//<![CDATA[ \begin{array}{l}Θ\end{array} //]]>$)

• $//<![CDATA[ \begin{array}{l}NA\end{array} //]]>$: numerical aperture of the untempered fiber

• $//<![CDATA[ \begin{array}{l}m\end{array} //]]>$: demagnification factor of the tapered fiber

• $//<![CDATA[ \begin{array}{l}α\end{array} //]]>$: fraction of entrance surface that is covered by the core fiber

Electron-optic coupling/image intensifier

A quite sophisticated coupling that also amplifies the incoming signal is the image intensifier. Here, the X-rays first strike an input phosphor layer that converts the incident radiation into visible light. Directly connected to it, a negatively charged photocathode converts these photons again into electrons. These are now accelerated towards a positively charged anode where they strike another phosphor layer and get converted again into visible light, which is subsequently directed to an optical lens coupling system that leads to the detector. During their flight between cathode and anode they are accelerated by a voltage around usually 25 kV and get focused by additional focussing electrodes that concentrate their intensity to the smaller surface of the output phosphor. This way, final amplifications on the order of 5.000 can be achieved. Figure 2 shows the typical composition of such a coupling system. Advantages beside the high signal amplification include an almost noiseless gain and the possibility to adjust the field of view easily. On the other hand, the input phosphor needs to be spherically shaped in order to withstand the strong vacuum within the image intensifier that is necessary for the electron acceleration. As a result of a spherical input and a planar output phosphor, the final image gets warped to some degree (so called pin-cushion distortion). Beside that a whole slew of different additional distortions are possible as well as a reduction of image contrast. Nevertheless image intensifiers have proved quite useful, but need some expertise in order to find optimal working parameters^[9].

Sensors for indirect conversion systems

In order to generate an output signal, the light emitted by the scintillator needs to be finally detected by the sensor chip. This is usually achieved by a photodiode, that is based in most cases on amorphous silicon. The obtained signal is read out subsequently by a suitable circuitry, e.g. thin-film-transistors (TFTs), complementary metal-oxide-semiconductor (CMOS) sensors or charge-coupled-devices (CCDs). When using the latter, no photodiodes are necessary. But due to their form factor, CCDs usually need a demagnifying scintillator-photodetector coupling, where TFTs are generally used without such^[4][10].

Direct conversion

Direct conversion radiography relies on photoconductor sensors that convert incident X-rays directly into an electric charge without any need for further components. They consist of a sandwich structure of two electrodes with an intermediate photoconductor layer and collection electrodes at the bottom (see figure 3). These electrodes form a capacitor and apply a strong electric field through the photoconductor layer. Amorphous selenium is used usually as photoconductor material. Alternative approaches are based on lead iodide and oxide, thallium bromide, gadolinium compounds and cadmium telluride (CdTe). When X-rays strike this layer, they are attenuated and produce electron-hole pairs proportional to the deposited energy. Due to the electric field, the charge carriers are drawn to electrodes at the bottom, where they are finally detected. Direct conversion detectors allow photon counting with energy discrimination. The applied field needs to be strong because of various reasons. First, the produced charges need to be hindered to recombine again. Next, the active charge collections allow to implement relatively thick photodiodes for increased efficiency. Third, a high voltage and thus a resulting strong electric field is able to collect the produced charges with a low lateral spread. Due to the high intrinsic spatial resolution of the used photoconductor materials, the pixel size as well as the matrix and spatial resolutions are only limited by the recording and readout devices. If amorphous selenium is used, the needed plates can be fabricated by evaporation, which is comparably easy to handle and inexpensive. Compared to indirect detectors a simpler structure is possible since the photodiode is replaced by a far less complex charge collecting electrode^[1][4][9][7]. As photoconductor preferably materials with high absorption are used, i.e. CdTe. This material shows a good quantum efficiency, even for higher energies (several hundred keV). The absorption efficiency is significantly higher than for silicon photoconductors for the range up to 140 keV, which is especially relevant for medical applications. A small band gap (1,52 eV) enables to use the detector at room temperature and offers an excellent energy resolution^[5].

For direct conversion detectors different effects can occur, that prove problematic or at least cumbersome. Some are enumerated here^[11]. First, the signals of different photons can sum up if they hit the detector within a short period of time. This is dependent on the count rate and dead time of the sensor. This way several captured pulses are observed as a single one that has a higher energy than the real input. This phenomenon is called pulse pileup. Next, charge sharing can occur. Within the photoconductor a hole-charge pair is created and both charge carriers drift opposite electrodes. On their way there due to diffusive effects and the Coloumb force they create a charge cloud that has an increasing spatial spread within the conductor. This spread can be large enough to hit more than one pixel finally – the cloud is divided and at two or more adjacent pixels a signal is detected, that shows a smaller energy than the photon was caring in reality. The created charge carrier can also be captured by a trapping center and released some time later due to thermal exhaustion. This phenomenon is called charge trapping. As a consequence, the spectral response is degraded due to reduced pulse amplitudes and by generating low energy tailing effects in the spectral characteristics. Incident X-rays can also interact with the sensor material due to Compton scattering or the photoelectric effect. These effects result in scattered photons that can be subsequently detected at a different pixel, leave the detector or result in secondary effects like the emission of characteristic wavelength photons or Auger electrons. Due to their statistic nature, these effects can lead to higher or lower detected pulses as well as spatial inaccuracies. Selemium-based detectors are usually constructed of either a so called selenium drum or a flat panel detector:

Selenium drum detector

The design of this detector consists of a selenium-coated aluminum drum that spins slowly during irradiation. Before each use the drum is positively charged. When X-ray photons hit the surface of the drum they cause a change of this surface charge proportionally dependent on their local radiation intensity. During use, this shift is detected and processed by an analog-digital converter. Afterwards the detector’s surface needs to get recharged before it is ready to use again. By reason of their mechanical composition selenium drum detectors are usually not mobile^[1][12].

Selenium-based flat panel detector

The flat panel detector design is newer than the drum-based systems. The principle of operation equals the one of drum detectors, but instead of a spinning drum a flat plate is used. The generated shift of charge is detected by a second, underlying layer of TFTs. At the very top of this sandwich structure a bias electrode is placed in order to easily apply an initial electric field to cause the separation of charges^[1][4].

Image quality parameters

A compilation of the most important properties of detectors that determine the image quality are enumerated in the following.

Quantum efficiency

The probability for the interaction of an incident X-ray with the detector material is called the quantum efficiency $//<![CDATA[ \begin{array}{l}η\end{array} //]]>$. It can be calculated by use of the following equation:

$//<![CDATA[ \begin{array}{l}\eta = 1- e^{-\mu(E)T}\end{array} //]]>$

• $//<![CDATA[ \begin{array}{l}\eta\end{array} //]]>$: efficiency

• $//<![CDATA[ \begin{array}{l}μ\end{array} //]]>$: linear attenuation coefficient of detector material (depends on particle energy $//<![CDATA[ \begin{array}{l}E\end{array} //]]>$)

• $//<![CDATA[ \begin{array}{l}T\end{array} //]]>$: active detector thickness

Detector size

The geometrical dimensions of the X-ray detector. The detector needs to be large enough to be able to capture the whole area of interest but also appropriate small for the desired application. Also, with increasing detector size the price for the system increases. Common prices for a detector are around 50.000-100.000 Euros^[4].

Image acquisition time and frame rate

The readout process of the detector involves a flow of current for each pixel. This process takes a certain amount of time and incomplete flows can result in artifacts within the final image. Therefore, the image readout time is related to the obtainable precision for each pixel. This is particularly important, since electronic noise contributes as main source to the overall noise of the detector. Taking these considerations into account, different frame rates can be achieved for several detectors. For example, the XRD a-Si family from Perkin Elmer, typical TFT sensor are able to operate at maximum 25 fps with full resolution. Contrary to that, the Dexela CMOs family from the same producer can achieve 191 fps under the same conditions^[4][13].

Spatial resolution

Spatial resolution is the minimal resolvable separation between two adjacent structures. It is defined and limited by the pixel size on the sensor. Low spatial resolution result in a blurred image. Spatial resolution is often expressed by the so called modulation transfer function (MTF), which has proven to be a useful measure for the effective resolution^[1][2][4].

Dynamic range

Dynamic range is the quotient between the largest and the smallest input intensity between which a meaningful picture can be obtained. It is limited by the noise level (smallest value) and the detector saturation level (highest value). Compared to AR, digital radiography systems have a wider and linear dynamic range^[1][2].

Sensitivity

Sensitivity can be defined as the charge produced at the sensor for an incident X-ray photon of a prescribed energy. It can also be expressed as the minimum energy that is necessary to produce an output (i.e. release a photon in the phosphor layer or an electron-hole pair in the detector)^[7].

Noise

These are fluctuations within the final image that do not correspond to the actual differences in X-ray absorption. Sources of noise are quantum fluctuations in the X-ray, the data digitalization (AD-converter), electronic noise and several more. A measurement of noise is the noise power spectrum (NPS), which measures the correlation of noise to spatial frequency^[2]. A possibility to enhance the signal-noise-ratio is to average between several pictures of the same object (see figure 4).

Literature

1. Körner, Markus, Weber, Christof, Wirth, Stefan, Pfeifer, Klaus-Jürgen, Reiser, Maximilian, Treitl, Marcus: "Advances in Digital Radiography: Physical Principles and System Overview". RadioGraphics. (2007) 27, p. 675–686
2. Oborska-Kumaszyńska, Dominika, Wiśniewska-Kubka, Sylwia: "Analog and digital systems of imaging in roentgenodiagnostics". Pol J Radiol. (2010) 75:2, p. 73-81
3. Lanca, Luis, Silva, Augusto: "Digital Imaging Systems for Plain Radiography". First Edition, Springer, New York(2013), p. 9-19
4. Chotas, Harrell, Dobbins, James, Ravin, Carl: "Principles of Digital Radiography with Large-Area, Electronically Readable Detectors: A Review of the Basics". Radiology (1999) 210, p. 595–599
5. Guni, Ewald: "Untersuchung von CdTe als Sensormaterial fur die spektroskopische Rontgenbildgebung". Dissertation, Friedrich-Alexander-Universität Erlangen-Nürnberg, 2012, p. 7, 39
6. Shultis, Kenneth, Faw, Richard: "Fundamentals of Nuclear Science and Engineering". First Edition, Marcel Dekker, New York(2002), p. 214-234
7. Yaffe, M. J., Rowlands, J. A.: "X-ray detectors for digital radiography". Phys Med Biol. (1997) 42, p. 1-39
8. Biberman, Lucien: "Photoelectronic Imaging Devices. Volume 1: Physical Processes and Methods of Analysis. First Edition, Plenum Press, New York (1971), p.289
9. Seibert, Anthony: "Flat-panel detectors: how much better are they?". Pediatr Radiol. (2006) 36:2, p. 173–181
10. Farrier, Michael, Achterkirchen, Thorsten, Weckler, Gene, Mrozack, Alex:"Very Large Area CMOS Active-Pixel Sensor for Digital Radiography". IEEE Transactions on electron devices. (2009) 56:11, p. 2623-2631
11. Taguchi, Katsuyuki, Iwanczykb, Jan: "Vision 20/20: Single photon counting x-ray detectors in medical imaging". Med Phys. (2013) 40:10, p. 100901-1 - 100901-19
12. van Heesewijk, J. P. M., de Valois, J. C.: "Clinical evaluation of a new digital chest imaging system". Eur Radiol. (1995) 5, p. 83-87
13. Official homepage of PerkinElmer Inc., [1], 21.12.2016
14. Normenausschuss Materialprüfung (NMP) im DIN: Deutsche Norm DIN EN ISO 19232-5 in der Fassung vom Dezember 2013