Spectral X-ray imaging with the Widepix 1×5 MPX3 CdTe detector
This article describes the applicability of high resolution and sensitivity detectors to material resolving X-ray imaging. Regular X-ray images are black and white, i.e. show only variations in density and thickness of materials. Thin, high density, heavy material can appear at the same grey level as a thick layer of light material. Therefore, some features of samples may remain undetected.
The energy-sensitive (spectral) X-ray imaging can on the other hand provide information on the elemental composition of the inspected object.
Photon counting X-ray detectors
Hawkeye Spectral Imaging provides Advacam’s so-called direct conversion hybrid photon counting (HPC) detectors. X-ray radiation is converted in a semiconductor or semi-insulator sensor directly into a measurable electric signal. Signals are processed and digitally stored in each pixel which is 55 µm size. This approach overcomes several limitations of common scintillator-based imaging flat panel detectors (FPDs). The photon-counting detectors eliminate most of the noise sources present in common FPDs achieving a nearly arbitrary signal-to-noise ratio.
The scintillators in FPDs with a small pixel size (less than 100 µm) have to be deposited in thin layers to minimize the light spread and match the pixel size. The small scintillator thickness limits the sensitivity and increases measurement times. The HPC imagers can use thick sensors while not compromising the resolution. Therefore, they offer considerably higher X-ray sensitivity compared to FPDs of the same pixel size.
The pixel electronics in the new detectors are designed radiation-hard extending further the lifetime of imaging detectors.
Also, the direct conversion photon-counting detectors are capable of X-ray energy discrimination, i.e. only photons above certain energy or within a certain energy window are detected allowing the spectral imaging.
Applications of HPC detectors were limited for a long time due to the availability of Silicon sensors only. The silicon crystal does not provide a sufficient detection efficiency for photons over 20 keV. Today, so-called high-Z CdTe or CZT sensors are available offering a significantly higher sensitivity.
Hawkeye Spectral Imaging offers Advacam’s photon-counting detectors with 1 mm thick CdTe sensors and prepares devices with 2 mm thick CZT sensors. The sensitivity of these sensors is considerably higher not only compared to Si, but also to the scintillators commonly used in high-resolution flat panels (pixels of less than 100 µm).
At the same time, the high detection efficiency of CdTe does not mean compromising the spatial resolution. The pixel pitch is still 55 µm. The signal created by photons in the sensor can split between neighboring pixels. This effect is called charge sharing. It is similar to the light spread in scintillators. However, thanks to the electric field applied to the sensor, this spread is significantly smaller than the light spread in scintillators. The charge sharing affects the spatial resolution of the detector only negligibly. This is demonstrated in the image above where 50 and 63 µm wires pairs are resolved.
However, charge sharing can have a significant impact on the energy response of the detector. This is true especially for thick sensors such as the 1 mm thick CdTe. The charge sharing was limiting the use of CdTe sensors for spectral imaging with pixel size below 100 or 200 µm.
Advacam’s X-ray imaging detectors are based on Medipix3 ASIC  pixellated chips that integrate circuitry for charge sharing correction. The electronics sums signals that leaked to neighboring pixels and compare the summed value with the energy level threshold. The photon hit is then recorded only in the pixel with the highest signal. This effectively suppresses the extra counts caused by the charge sharing and tremendously improves the sensor energy response. In consequence, it allows spectral imaging with CdTe sensor even at the high resolution of 55 µm.
This functionality is available in the detector Widepix 1×5 or 2×5 CdTe MPX3 . Selected basic detector parameters are listed in the table:
Spectral imaging with Medipix3 CdTe
The spectral X-ray imaging was evaluated using Radalytica robotic X-ray scanner  that integrates Widepix 1×5 MPX3 CdTe detector. The scanner essentially extends the Widepix sensitive area from 70×14 mm2 to an area of up to 600×1200 mm2. The robotic scanner repositions the detector and X-ray tube, measures individual image tiles, and automatically stitches them into the large area scan. The spectral imaging was tested on a sample consisting of various pure metal foils stacked to different thicknesses.
The test sample was scanned using an X-ray tube operated at 50 kVp and 1 mA at a distance of 300 mm. The charge sharing correction (CSM) was activated in the detector. Each tile consisted of 13 energy levels from 10.2 keV to 40 keV, i.e. each image contained photons with energies above this level. The exposure time per energy was 1 s. The overall scan area was 165 x 360 mm2. The B&W images were corrected using the signal-to-thickness correction . The measured spectral data were differentiated, i.e. images at subsequent energy levels were subtracted. An X-ray attenuation was calculated:
Im(E) is the measured differentiated signal at energy E, I0(E) is the signal without any sample. µ(E)x data then served to “colorize” the X-ray images as it could contain sudden steps that correspond to k and l edges in X-ray absorption as a function of energy. These edges are characteristic of different chemical elements.
The measured energy steps were relatively coarse to maintain the overall measurement time acceptable. Therefore, similar elements are not fully resolved. Moreover, lighter elements such as Fe, Co, Ni, and Cu have their k-edges between 7 and 9 keV. However, the minimum energy discrimination threshold used was 10 keV. The most distinguishable feature, the k-edge, was therefore out of the energy range. Similarly Al, Ti, and the plastic frames. Here the shape of the measured spectrum was distinctive enough to distinguish these materials to some extend from Fe, Co, Ni, and Cu.
Contrary to light elements, Mo has the k-edge at 20 keV and was recognized very well. Similarly, Ag with the k-edge of 25.5 keV and Sn with the k-edge of 29.2 keV were resolved well. Ta, W, Pt, and Pb have their k-edges outside of the X-ray tube range (50 kVp). Therefore, these elements were “colorized” based on spectrum distortion caused by their l-edges that is however less pronounced. The l-edges are between 11 and 16 keV. The sensitivity of the CdTe sensor would allow using higher X-ray energies and therefore provide more reliable detection of these materials with k-edges. However, an X-ray tube with a higher acceleration voltage was not available for this experiment. Selected measured curves are shown in the plots below.
An example of spectral X-ray imaging is the differentiation of materials in electronics. A laptop was scanned using the Widepix MPX3 CdTe detector on the robotic scanner. The resulting scan has 7852×9880 pixels, i.e. 77.6M pixels. Both, the black-and-white and “color” X-ray images are shown below.
High-Z CdTe sensors combined with the functionality of Medipix3 chips offer a mixture of high resolution (55 µm pixels), sensitivity (100% at up to 60 keV and 30% at 140 keV), and energy measurement, i.e. the spectral X-ray imaging.
The spectral imaging with CdTe at this pixel size is enabled by the charge summing functionality implemented in the device. It overcomes limitations of the CdTe sensor material which always used to be the increased charge sharing, i.e. cross-talk between pixels.
The spectral information can be used to distinguish materials based on their elemental composition and thus bring a new level of information into general X-ray imaging.
This technology finds for example applications in X-ray inspection of electronics, biological and medical samples, material science, art imaging , composite materials imaging, and many others. The photon-counting X-ray imaging technology has progressed in recent years from a “toy” for scientists to a capable tool for a wide variety of industrial X-ray imaging applications.
 R.Ballabriga, M.Campbell, E.Heijne, X.Llopart, L.Tlustos, W.Wong: Medipix3: “A 64 k pixel detector readout chip working in single photon counting mode with improved spectrometric performance,” Nuc. Instr. and Meth. in Phys. A, Vol. 633, Supplement 1, May 2011.
 J. Jakubek: ”Data processing and image reconstruction methods for pixel detectors,” Nuclear Instruments and Methods in Physics Research A 576 (2007) 223–234.