Medipix and Timepix type hybrid-pixel photon counting detectors were originally developed and intended as particle trackers at the Large Hadron Collider at CERN. Nevertheless, the applicability of Medipix technology is much broader and exceeds the field of high energy physics.
Radiation Measurements 137 (2020) 106409 Contents lists available at ScienceDirect Radiation Measurements journal homepage: http://www.elsevier.com/locate/radmeas High-resolution X-ray imaging applications of hybrid-pixel photon counting detectors Timepix Jan Dudak a, b, * a b Institute of Experimental and Applied Physics, Czech Technical University in Prague, Husova 240/5, 110 00 Prague, Czech Republic Faculty of Biomedical Engineering, Czech Technical University in Prague, Namesti Sitna 3105, 272 01, Kladno, Czech Republic A R T I C L E I N F O A B S T R A C T Keywords: Photon counting detectors X-ray imaging X-ray radiography Computed tomography Medipix and Timepix type hybrid-pixel photon counting detectors were originally developed and intended as particle trackers at the Large Hadron Collider at CERN Nevertheless, the applicability of Medipix technology is much broader and exceeds the field of high energy physics The unique features of Medipix devices – namely the dark-current-free quantum-counting, energy-determination and steep point-spread function – make them a powerful tool for imaging using ionizing radiation This work provides insight into the applied results of Medipix technology from fields of transmission X-ray imaging and X-ray fluorescence (XRF) imaging The history of Medipix technology is briefly described The detectors are then characterized by means of key parameters connected with imaging techniques Medipix detectors are compared with conventional X-ray imaging cameras and their advantages and disadvantages are discussed Finally, the methodology principles of high X-ray transmission radiography and XRF imaging are explained and a number of applications from different fields of science are demonstrated Introduction The history of X-ray imaging started in 1895, when X-radiation was discovered and described Very soon after that X-ray radiography became an essential tool of medical diagnostic care (Bushberg, 2002) The photographic plate was used as a standard detection technology for decades The rapid development of imaging X-ray detectors came much later with the introduction of the first digital detection technologies The Medipix detectors are a relatively new family of radiation-sensitive de vices utilizing the particle/photon counting approach Despite Medipix being originally developed for high energy physics, its advantages for other fields of science were very quickly recognized Nowadays, Medipix type detectors are used for radiation imaging, digital dosimetry, educational purposes and many other applications This paper is focused on summarizing the use of Medipix detectors in the field of X-ray imaging The first generation of the Medipix chip was introduced in the 1990s at CERN to serve for tracking of high-energy particles at the Large Hadron Collider It provided a pixelated array of 64 � 64 pixels with a 170 μm pixel pitch (Bisogni, 1998) Several new generations of the Medipix chip have been successfully developed by the established Medipix Collaboration since that time With further development and new generations of the chip coming, the Medipix technology has found a number of applications outside the field of high energy physics The aim of the Medipix collaboration was clear from the very beginning – to create a highly versatile radiation imaging detector of superior quality This original goal can still be seen from the former logo of the collaboration, since the very first radio graphic image (a metal wire formed into the shape of the letter “M”) acquired with a Medipix chip was used The successors of Medipix1 have provided additional functionality, a smaller pixel pitch and a larger sensor area compared to the first gen eration Medipix2 provided a significantly smaller pixel pitch (55 μm) and increased sensor area (14 � 14 mm2 with an array of 256 � 256 pixels) (Llopart, 2002) The Timepix chip enabled the analysis of the time-of-arrival of each detected particle or position-sensitive spectro scopic measurement as the energy of detected particles can be directly estimated (Llopart et al., 2007) The ability to perform fully spectro scopic measurements was a break-through in the field of X-ray radio graphic imaging and computed tomography (CT) The evaluation of the incident beam spectrum provides additional information on the elemental composition of the imaged sample compared to conventional * Institute of Experimental and Applied Physics, Czech Technical University in Prague, Husova 240/5, 110 00 Prague, Czech Republic E-mail address: jan.dudak@cvut.cz https://doi.org/10.1016/j.radmeas.2020.106409 Received 19 February 2019; Received in revised form 20 September 2019; Accepted June 2020 Available online June 2020 1350-4487/© 2021 The Author Published by Elsevier Ltd This is an open (http://creativecommons.org/licenses/by-nc-nd/4.0/) access article under the CC BY-NC-ND license J Dudak Radiation Measurements 137 (2020) 106409 radiography Medipix3 was designed with two user-adjustable energy thresholds per pixel and with on-board processing addressing the issue of charge sharing between neighboring pixels (Gimenez et al., 2011) For purposes of spectroscopic imaging, the Medipix3 chip can be used in the so-called superpixel mode Superpixels are clusters of four read-out pixels behaving like a single detection unit Such an approach obviously sacrifices the spatial resolution of the detector, but on the other hand each superpixel provides eight adjustable energy thresholds Timepix3 ensures the option to simultaneously measure the time of arrival and the energy of the detected particles in a data-driven read-out mode (Frojdh et al., 2015) Timepix4, currently under development, promises to introduce pixels smaller than 55 μm thanks to the utilization of TSMC 65 nm technology for the chip design and with more pixels situated on each chip Furthermore, it will be four-side buttable, therefore, a straight forward assembly of large-area detector arrays with continuous sensi tivity will become possible (Campbell et al., 2016) The photon counting detectors (PCD) utilize two construction ap proaches – monolithic or hybrid The detectors of Medipix type use the hybrid-pixel construction characterized with separated sensor and readout chips interconnected using bump-bonding technology The hybridpixel construction is more versatile as the sensor and read-out chips are individual units, thus it is possible to use various sensor materials The sensor chips have mostly been manufactured from silicon, never theless, a number of alternative sensor materials have appeared Considering the application of Medipix technology for X-ray imaging purposes, the major limitation of silicon is in its quantum efficiency Semi-insulating materials containing high-Z elements (CdTe, CdZnTe, GaAs) seem to be a promising solution for this limitation in future The weakness of Medipix technology, that limited its wider appli cability in X-ray imaging, used to be the size of the sensor Two square centimeters of sensor area were not convenient for real-life X-ray im aging There is a possibility to scan larger samples into a set of tiles using a high-precision remote-control positioning system Nevertheless, such an approach is only suitable for 2D radiography Performing a CT scan with the necessity of moving the detector to several positions during each projection would result in an unbearable scan time prolongation For this reason, development has continued not only in chip archi tecture but an effort has also been placed into increasing the sensitive area of the detectors The first success in this field was introduced by Medipix2 collaboration as the Quad detector – four Timepix chips bumpbonded to a common semiconductor sensor The Quad uses the layout of by chips providing a sensitive area of 28 � 28 mm2 (see Fig 1) Similar attempts of multiple read-out chips connected to a common sensor layer were introduced as Hexa (2 by chips, 512 � 768 pixels) and LAMBDA – Large Area Medipix Based Detector Array (2 by chips) (Zuber et al., 2014; Pennicard et al, 2011) Building larger detectors this way turned out to be impractical due to manufacturing complications and the low yield from the wafer Furthermore, since Medipix chips have been three-sides tileable up until the present, while the fourth side has been kept for chip peripheries, it has not been possible to bond more than two chip rows to a common sensor Therefore, to achieve a larger detector area, the assembling of several detector modules to an array has been necessary The RELAXd project developed by Nikhef and PAN alytical provided a read-out board for Quad detectors perpendicular to the sensor plane (Vykydal et al., 2008) The perpendicular position of the sensor and the read-out board enabled putting several Quad as semblies into a 2D array Similarly, it is possible to assemble several LAMBDA modules into an array too In the case of Hexa modules, only the creation of a row of several devices is possible All mentioned de tector arrays have been used mostly for X-ray diffraction measurements Since it is not possible to assemble them without insensitive gaps be tween adjacent modules, their applicability for X-ray radiography is questionable Canas et al aimed to build a Medipix-based detector with the field of view reaching an area of 24 by 30 cm2 for use in mammography (Canas et al., 2011) The developed device was built of an array of 11 by Medipix2 chips The individual assemblies were positioned with gaps proportional to the chip size Each scan, therefore, consisted of four sub-acquisitions in different positions of the detector array to cover the whole area that were automatically stitched together A significant step forward came with introducing of edgeless sensors Omitting the guard ring around the sensor perimeter has enabled assembling individual detectors to rows with virtually no insensitive gaps The edgeless sensors have then been utilized by WidePIX tech nology developed at the Institute of Experimental and Applied Physics, Czech Technical University in Prague (IEAP CTU) WidePIX detectors solve the problem of chip peripheries preventing the assembly of 2D chip arrays The WidePIX concept is based on building of chip rows and their later arrangement into a 2D array The adjacent rows are slightly overlapping, so the peripheries of the first row are covered by sensors of the second row and so on The largest detector built this way is WidePIX10�10, shown in Fig 2, created out of 100 individual Timepix chips to produce a continuously sensitive area of approximately 14 by 14 cm2 (2560 � 2560 pixels) (Jakubek et al., 2014) Until now, just one such detector has been built and it is operated at the Centre of Excellence Telc (CET), Czech Republic The current necessity of roof-like rows tiling should be omitted once the Timepix4, currently under development, is released Timepix4 will be 4-side buttable as it utilizes the TSV (through-silicon-via) technology Fig Timepix Quad detector with a FITPix readout interface The detector consists of four Timepix chips bump-bonded to a common silicon sensor layer The detector provides an array of 512 by 512 pixels and a total sensitive area of 28 by 28 mm2 (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) J Dudak Radiation Measurements 137 (2020) 106409 Fig WidePIX10�10 – a view inside The detector consists of an array of 10 by 10 Timepix chips precisely aligned to provide a continuously sensitive area of approximately 14 by 14 cm2 (2560 � 2560 pixels) Courtesy of Jan Jakubek, IEAP CTU, ADVACAM s.r.o allowing the signal to be read out using copper-filled holes passing through the chip already commercially available on the market including one system utilizing Medipix technology 1.1 High resolution X-ray imaging with a laboratory set-up 1.2 High resolution X-ray imaging in cone-beam geometry Intensive research and development in the field of laboratory X-ray sources and digital detector technology constantly pushes the limits of achievable spatial resolution of transmission X-ray imaging approaches Spatial resolution at the level of several microns used to be the domain of synchrotron facilities for a long time Nowadays, such a resolution is routinely achievable using laboratory X-ray imaging systems Using the state-of-the-art laboratory X-ray sources with suitable detector tech nology, it is possible to reach a resolution even deeply below μm Besides the utilization of nano-focus X-ray tubes, that are probably the most popular sources for high resolution X-ray transmission radi ography and CT, there are also other attempts focused on sub-micron precision X-ray imaging An electron gun from a scanning electron mi croscope can be modified and used for X-ray projection imaging Spatial resolution better than 60 nm achieved this way was demonstrated (Mayo et al., 2005) Another well-known option is the employment of suitable optics like Fresnel zone plates (FZP) X-ray microscopes equipped with FZP have been successfully used for the imaging of in dividual cells with spatial resolution of 30 nm (Jacobsen, 1999) Although both of the mentioned approaches offer superior spatial res olution, the application field is relatively narrow and the set-up is rather complicated Especially in the case of FZP – their use is restricted for energies lower than 10 keV and a monochromatic beam is demanded It is clear that such a beam can be used for transmission imaging of extremely small objects only The SEM-based source can work with energies up to approximately 30 keV, but the beam intensity is very low and thus long acquisition times are inevitable As it has already been stated, nano-focus X-ray tubes are the most popular sources for high resolution X-ray transmission imaging The detail detectability of an X-ray imaging system with the latest state-ofthe-art nano-focus tube can reach 150 nm (Excillum, 2018) That is an order of magnitude worse than SEM-based sources or FZP, but on the other hand, a polychromatic beam with 60 kVp can be used Further more, an X-ray tube is much more versatile and easy-to-use as it pro duces a divergent beam that enables the scanning of samples in a wide range of sizes Several X-ray nano-CT systems with nano-focus tubes are Laboratory X-ray imaging systems dedicated for high resolution imaging usually utilize a nano-focus X-ray tube as the radiation source An X-ray tube emits X-ray photons in a divergent beam with point-like origin The divergent geometry, frequently called cone-beam, allows the projection of the imaged object to be magnified as the beam spreads The magnified projection then covers more pixels of the detector and, therefore, the sampling density of the obtained image is increased (see Fig A and B) Thanks to the cone-beam geometry, it is possible to achieve much higher resolution compared to the native resolution of the used detector unit The magnification factor (M) is given as a ratio be tween the source-to-detector distance (SDD) and the source-to-object distance (SOD) The actual sampling density is usually denoted as an effective pixel size (EPS) and is given by the physical dimensions of the detector pixel divided by the used magnification factor The increase of M inevitably sacrifices the field of view (FOV) Therefore, large-area and finely pixelated detectors are needed The major limitation of the maximum achievable spatial resolution of an imaging system is the size of the focal spot of the source Once the EPS becomes smaller than the focal spot size, the spatial resolution cannot be further improved as the image will become blurred due to the penumbra effect (see Fig C) The focal spot size of the X-ray source is, therefore, of key importance in the case of high-resolution X-ray imag ing X-ray tubes with a highly focused electron beam also have, unfor tunately, a drawback The electron beam power density at the target must be controlled otherwise a thin anode of the X-ray tube would be quickly destroyed by dissipating heat The X-ray beam intensity is much lower compared to widely used micro- and mini-focus sources Since the photon flux is limited, the acquisition time of each pro jection has to be inevitably prolonged to obtain projections with suffi cient statistics Therefore, the exposure time can vary from several seconds to tens of seconds Photon counting detectors have proven themselves to be an excellent choice in this case As PCDs acquire data without dark current, the shutter can be open for an arbitrary time period and the noise of the data is still given by Poisson distribution alone J Dudak Radiation Measurements 137 (2020) 106409 Fig X-ray imaging in cone-beam geometry The diverging beam enables changing the magnification factor by changing the distance from the source to the sample (SOD) The magnification factor M, given as the ratio of the source-to-detector distance (SDD) and SOD, affects the sampling density (effective pixel size – EPS) of the final image and, therefore, also the achieved spatial resolution as can be seen from the comparison of A) and B) The maximal useful magnification of an X-ray imaging system is limited by the size of the focal spot Once the EPS becomes smaller than the focal spot size the resolution does not further improve, but the obtained image suffers from penumbral blur (C) An example of a CT scan carried out with sub-micrometer resolution can be seen in Fig The figure shows volume rendering of a fossil foraminifera sample (foraminifera are single-celled marine organisms which have inhabited the Earth for more than 500 million years) scan ned with an effective pixel size of 830 nm using the WidePIX4�5 detector and the FeinFocus FXE-160.51 multifocus X-ray tube at IEAP CTU The left part of the figure shows the volume rendering of the whole sample while the right part reveals the inner parts of the sample filled with pyrite crystals The tube was operated in the nano-focus mode with 50 kVp and 20 μA The dataset consisted of 720 projections (10 s acquisition time each), with a 0.5� angular step experimental imaging systems and has not been widely used in commercially available devices yet Nowadays, considering X-ray im aging with a resolution at the level of μm or less, the most frequently used detector technology is a CCD (charge-coupled device) chip with a thin scintillation sensor Such cameras provide extremely high pixel granularity – very often pixels smaller than 10 μm – and the total number of pixels as well (10 megapixels or more) Therefore, it is easy to perform CT scans with an extremely small EPS simultaneously with a wide field of view using these detectors The key disadvantage of scin tillation sensors is that the emitted light spreads evenly in all directions within the sensor and, therefore, the spatial resolution of such a camera depends more on sensor thickness than on the pixel size Since the thickness of a scintillation sensor is typically larger than pixel di mensions, the point spread function (PSF) of these devices is usually much wider than pixel size On the contrary, Medipix technology fea tures 55 μm pixels but the PSF has a box-like shape and its width is proportional to the pixel size due to the high bias voltage applied to the sensor It was previously demonstrated that a Medipix2 detector offers a 1.3 Advantages of Medipix technology in the field of high-resolution Xray imaging Medipix and Timepix detectors offer a set of unique features that can be extremely useful when high resolution X-ray radiography and CT are discussed However, the Medipix technology has mostly been utilized in Fig Example of a high-resolution CT scan of a sample of foraminifera performed with a WidePIX4�5 detector and the FeinFocus FXE160.51 multifocus X-ray tube at the Institute of Experimental and Applied Physics CTU in Prague The left part of the figure shows volume rendering of the whole sample while the right part enables seeing the inner parts of the sample filled with pyrite crystals The sample was scanned with 830 nm EPS, the tube was operated in the nano-focus mode with 50 kVp and 20 μA The dataset con sisted of 720 projections with a 0.5� angular step The sample was kindly provided by Katarina Holcova (Department of Micro Palaentology, Faculty of Science, Charles University) J Dudak Radiation Measurements 137 (2020) 106409 comparable resolution as the X-ray CCD camera CRYCAM (Tous et al., 2011) The experimental comparison of the contrast and modulation transfer function (PSF) showed that the Medipix2 device with its 55 μm pixels provided a spatial resolution of 10.69 lp/mm (line pairs per millimeter) while the CRYCAM camera with μm pixels offered spatial resolution of 13.44 lp/mm A similar experiment was later repeated comparing Timepix with an 11 Mpix CCD X-ray camera with μm pixel pitch and a 22 μm thick Gadox scintillation sensor installed in a Bruker 1172 scanner (Dudak et al., 2017) The CCD detector was designed to be operated in three different binning modes – full resolution (9 μm pixels), � pixels (18 μm pixels) and � pixels (36 μm pixels) Analysis of a slanted edge captured by both detectors showed that the PSF of the Timepix chip has approximately the same width of PSF at FWHM as the tested CCD camera operated in the � pixels binning mode (18 μm pixels) It was also demonstrated that the steep PSF of Timepix detectors enables improving detail detectability within the reconstructed CT data by voxel space oversampling (Dudak et al., 2017) Conventionally, the CT data are reconstructed with a voxel size proportional to EPS Nevertheless in the case of sharp projection data, it was shown to be profitable to make voxels smaller Experimental testing showed that the oversampling of the voxel space by the factor 2–3 suppresses partial volume effects and improves the detail detectability in the data The experimental comparison also demonstrated that the Timepix detector, due to noiseless counting, provides a higher contrast-to-noise ratio (CNR) and thus better detectability of fine details with the same image statistics in the projection (Dudak et al., 2016) Such a compari son is demonstrated in Fig A sample of ex-vivo mouse liver was scanned with a large area Timepix detector and an X-ray CCD camera with μm pixels and a 22 μm thick Gadox (Gd2O2S:Tb) scintillator Both images were measured with 4.3 μm EPS The Timepix detector (Fig left) revealed details down to 15 μm while the detail detectability of the CCD camera was approximately 60 μm due to significantly lower CNR (Fig right) resolution scanner capable of performing a CT scan with EPS within the range from 50 nm to 5.5 μm (Nachtrab et al., 2015) Furthermore, the utilization of a Medipix detector enables performing a K-edge absorp tionmetry simultaneously A high-end nano-focus tube with a spot size of 100 nm is used as the radiation source in this case A large area de tector was needed to ensure reasonable FOV considering the high magnification factors needed to achieve the demanded resolution A dedicated detector unit based on four Hexa assemblies arranged into a row producing the total sensor area of approximately 170 � 28 mm2 (3072 � 512 pixels) was designed for this purpose The FOV can be from 0.15 to 16 mm depending on the actually set magnification factor using such a detector A typical scan time with NanoXCT is approximately 10 h since the detection rate is approximately 2360 events per pixel in Nevertheless, the projection image quality does not suffer from increased image noise, since Medipix detectors work in the dark-current-free mode After a successful demonstration of the system viability, it was used as a prototype for the development and production of a commercially available product – RayScan Nano (2019) 1.5 Analysis of cultural heritage artifacts using X-rays Cultural heritage is a field where a lot of inspection methods of other fields of science find their applications Historical artifacts are analyzed to verify their authenticity, estimate the age, evaluate the current con dition of the artifacts or to find hidden damage etc X-rays are not the only radiation being utilized in this field by far One can use almost all wavelengths of the electromagnetic spectrum to analyze historic arti facts Frequently, fine art researchers utilize illumination using the IR or UV spectrum to obtain information that remains hidden in visible light However, considering the energy and penetrability of IR or UV photons, it is clear that these assessments can only provide information on the surface layers On the contrary, X-rays penetrates the matter easily and can deliver information from structures situated deeply below the surface Beside X-ray radiography or CT, other options for the utilization of Xrays in cultural heritage artifacts have been introduced An alternative option is the analysis of artworks using X-ray fluorescence (XRF) pho tons Since XRF photons have a discrete energy spectrum, the radiation carries information on the elemental composition of the material emit ting XRF photons In this section, the analysis of cultural heritage artifacts like historic painted artworks or sculptures will be discussed 1.4 X-ray microscope and the NanoXCT system An extreme approach considering the spatial resolution has been introduced by Fraunhofer IIS Medipix technology was used in the development of an X-ray microscope and a laboratory scale nano-CT system providing spatial resolution deeply below μm The X-ray mi croscope was introduced in 2011 and it was based on the use of an electron gun originally designed for an electron microscope focused on a very thin transmission tungsten target (Nachtrab et al., 2011) The achieved focal spot size of the source was 50 nm as the thickness of the target was only 0.1 μm The detector unit consisted of a Medipix2 MXR device in quad configuration The smallest achievable EPS was 55 nm as the construction of the device allowed data acquisition with a magnifi cation factor up to 1000 The second approach called NanoXCT is a high 1.6 Large area scanning The routine use of Medipix technology for the imaging of painted artworks has become practically possible with the introducing of largearea detectors as most historical pieces of art have a rather large size Fig X-ray projection of an ex-vivo liver lobe of a laboratory mouse scanned with a large area Timepix detector (left) and a 11megapixel CCD X-ray camera with a 22 μm thick Gadox scintillation sensor (right) The EPS was set to 4.3 μm in both cases and the exposure times were adjusted to get the same image statistics While the Timepix detector revealed fine veins down to 15 μm in diameter, the smallest features captured by the CCD camera were approximately 60 μm in size (Dudak et al., 2016) J Dudak Radiation Measurements 137 (2020) 106409 However, even with the employment of the largest available detectors, it is usually not possible to scan most of the artworks in a single acquisi tion The sample has to be, in this case, scanned in a set of slightly overlapping sub-acquisitions that are later merged together An example of such a scan can be seen in Fig A painting from 19th century � “Cernohorka” (dimensions of 22 by 28 cm) was scanned in 12 tiles using Widepix10�5 and the final radiography provides an image resolution of 4450 � 5600 pixels (Zemlicka, 2016) The final radiography can easily consist of hundreds of megapixels as the dataset is formed typically of several tens or even hundreds of subacquisitions Depending on the size of the scanned painting and the construction of the scanner, there are two options on how the scan can be performed The painting can be either scanned using the simulta neous movement of both the source and detector or with the fixed po sition of the X-ray source, while the detector is moving The first approach requires the precise synchronization of the detector and source movement Since the irradiation of one tile at a time is sufficient, the source and detector can be very close to each other, thus high beam intensity leading to shorter exposure time is achieved On the other hand, this approach can induce geometry artifacts in the final assembled radiographies since overlapping regions were captured with different parallax This fact limits the applicability of the mentioned approach especially in the case of thicker samples like paintings on wooden boards or sculptures The other scanning approach avoids the parallax problem, on the other hand it is not suitable for large paintings, since the field of view is limited by the X-ray beam cone angle and scanning speed is also slowed due to a lower photon flux Fig shows how surprising results can, in particular cases, be pro vided by an X-ray inspection of painted artworks The oil-on-canvas painting “Holy Family” was borrowed for the X-ray radiographic in spection from a private collection The dimensions of the painting were 84 � 65.7 cm and after the X-ray scan, a radiographic image consisting of approximately 23000 � 18600 pixels – an equivalent of ca 433 megapixels – was obtained The presented image was created as an overlay of the partially transparent X-ray image (grayscale) and optical photography (color) The X-ray inspection revealed that the landscapeoriented surface motive originating from the beginning of the 20th century hides another painted motive with portrait-orientation dated to the baroque period The Institute of Experimental and Applied Physics of Czech Technical University in Prague in cooperation with the Academic Material Research Laboratory of Painted Artworks of the Academy of Fine Arts in Prague designed and constructed an X-ray imaging system utilizing Medipix technology for the scanning of painted artworks (see Fig 8) The system consists of a shielded cabinet enclosing two identical frames with long-range linear motorized stages holding a micro-focus X-ray tube and a large-area Timepix detector Both frames move simulta neously in a mirror-like manner and provide a field of view up to one square meter A scanned painting is mounted on a dedicated sample stage between the scanning frames The samples can be scanned with a resolution within the range from 50 to 15 μm due to the fact that both the SDD and SOD of the system are adjustable 1.7 Energy-sensitive transmission radiography Since photon counting detectors in general are operated with one or more user-adjustable energy thresholds, or in the case of Timepix, can provide a fully spectroscopic response, these devices can be used for Fig An example of a painting scanned as a set of 12 tiles before (left) and after the final assembly (right) Partially overlapping tiles are merged together using � image registration techniques Painting from the 19th century “Cernohorka”, private collection Prague Measured at the IEAP CTU micro-CT laboratory equipped with a Hamamatsu L8601-01 X-ray tube operated at 80 kVp and a large area detector Widepix10�5 (Zemlicka, 2016) J Dudak Radiation Measurements 137 (2020) 106409 Fig Holy Family, oil on canvas: An overlay of X-ray radiography (grayscale) and optical photography (color) images The X-ray scan revealed a baroque motive that was over-painted at the beginning of 20th century The dimensions of the painting were 84 � 65.7 cm The final radiography consists of an array of approx imately 23000 � 18600 pixels – equivalent of almost 433 megapixels The presented image was kindly provided by Jan Zemlicka (Institute of Experimental and Applied Physics, Czech Technical University in Prague) and Janka Hradilova (Academic Material Research Laboratory of Painted Artworks) The painting was borrowed from a private collection energy-sensitive or so-called spectral X-ray imaging In the case of normal radiography the detector senses changes in the incident beam intensity Energy-sensitive radiography, on top of that, resolves changes in the beam spectrum The detection and proper interpretation of these changes can be used for the analysis of the material composition of a sample as linear attenuation coefficients are energy-dependent and characteristic for each element Spectroscopic imaging is feasible with all detector types belonging to the Medipix family Timepix devices can provide a fully spectroscopic response using the Time-over-threshold mode and proper cluster anal ysis However, the ToT mode also brings disadvantages concerning application in X-ray imaging A successful cluster analysis requires sparsely occupied frames without pile-ups Thousands of such frames are needed to create an X-ray projection with reasonable statistics Considering that the read-out speed of a large area detector is not faster than 10 frames per second, the exposure time for an X-ray projection of considerable quality becomes extremely long Therefore, sequential scanning of several acquisitions in the Medipix mode with different J Dudak Radiation Measurements 137 (2020) 106409 Fig X-ray scanning system for the imaging of painted artworks developed by the Institute of Experimental and Applied Physics CTU in Prague and the Academic Material Research Laboratory of Painted Artworks of the Academy of Fine Arts in Prague The system consists of two identical motorized frames designed to cover a field of view of m2 and an adjustable sample holder The whole set-up is situated in a walk-in shielded cabinet (Zemlicka, 2016) thresholds can shorten the exposure time significantly The detected spectrum can then be divided into several narrow energy bins using the subtraction of frames with different thresholds Fine pixel segmentation of a Timepix chip, which degrades its energy resolution, can on the other hand be easily used for topology mapping using fluorescence photons Once suitable optics is selected and moun ted in front of the chip, a 2D image based on XRF photons can be ob tained The simplest solution can be offered by a pinhole collimator A pinhole camera enables projecting a 2D area of the investigated object with high spatial resolution The obvious disadvantage of a pinhole collimator is its low geometric efficiency as just a small fraction of the emitted photons is accepted Furthermore, the probability of production of an XRF photon must be considered Therefore, the use of intensive sources and long exposure times are inevitable 1.8 X-ray fluorescence imaging The applicability of Timepix detectors in the analysis of cultural heritage artifacts is not only limited to transmission radiography The non-destructive spectroscopic evaluation of the elemental composition of artwork using an X-ray fluorescence (XRF) is a highly demanded task Hand-held devices with a pencil beam and a silicon drift detector for spectroscopic XRF analysis are available on the market These devices provide great spectral sensitivity, hence the ability to clearly identify various elements (i.e energy resolution 122 eV FWHM at 5.9 keV in the case of a spectroscopic camera by AMPTEK (Amptek, 2018)), on the other hand they are not suitable for making XRF-based topology map s/images since these devices have been designed just for local analyses On the contrary, a Timepix detector, despite the fact that it provides a fully spectroscopic response, cannot compete with these hand-held spectrometers by means of energy resolution The energy resolution of a Timepix detector with a 300 μm silicon sensor is approximately keV and, therefore, it is not possible to use it for the direct identification of the characteristic lines, considering that the energy difference between Z and Zỵ1 elements is ca 500 eV The energy resolution and thus the capability to resolve the elemental composition of the analyzed object can, however, be improved under certain circumstances The per-pixel response of the whole detector can be calibrated using pure character istic lines of targeted elements if the a-priori information of the global elemental composition of the object to be analyzed is available The calibration data are then used as base vectors for spectral decomposition of the later XRF scan It was demonstrated that elements heavier than potassium can be identified this way (Tichy et al., 2008; Zemlicka et al., 2009) 1.8.1 XRF pinhole camera with Timepix A prototype of a pinhole-based XRF camera was constructed to evaluate the applicability of Timepix technology for XRF imaging at IEAP CTU (Zemlicka, 2016) The prototype utilized a RasPIX camera equipped with a custom-made pinhole collimator with a 100 μm aper ture The collimator position was adjustable, so it was possible to align it with the sensor and also to adjust the distance between the sensor and thus change the field of view of the obtained image The XRF camera was mounted on a shared base plate with a Mini-X X-ray tube and revolver with a set of aluminum filters dedicated for the modulation of the mean energy of the X-ray tube spectrum (see Fig 9) See an example of the applied use of the camera for the XRF mapping of technological copies of historic paintings in Fig 10 The figure shows a photograph of a ROI of a technological copy of the Gothic period painting “Epitaph of Margaret” and an XRF map created from photons in the energy range 10–12 keV where a characteristic L-line of lead is ex pected Since lead used to be used as a component of white pigments in history, the brightest areas of the XRF image are expected to match with the white areas of the ROI J Dudak Radiation Measurements 137 (2020) 106409 typically metal based (Fe, Co, Pb, Hg, Zn …) For that reason, XRF analysis of the surface can provide information of its elemental composition or reveal previous restorations of the investigated artwork as organic pigments are usually used nowadays Imaging applications of Medipix detectors in industry and material engineering X-ray CT is usually connected with an imaging and analysis of pro duction parts in the industrial field Since those are frequently manu factured from various types of metal, silicon as a sensor material is unfortunately inconvenient as its quantum efficiency is very low, above approximately 30 keV Novel sensor materials (CdTe, CdZnTe, GaAs) become extremely useful here as they provide reasonable detection ef ficiency up to 100 keV For example a mm thick CdTe sensor provides detection efficiency higher than 50% for 120 keV photons (Greiffenberg et al., 2011) The other extensively developing material engineering field utilizing X-ray CT is the development of composite parts Composites are extremely lightweight and simultaneously durable Such materials are highly sought-after i.e in the aeronautics industry Micro-CT techniques can easily reveal an inner imperfection like a delamination of composite layers etc Medipix detectors can be very useful in this field due to energy-resolving capabilities Energy sensitive X-ray radiography and CT can visualize the variations in the density of the investigated object but also provide valuable information on its elemental composition That is possible due to the characteristic behavior of the linear attenu ation coefficient with respect to the energy for all elements or materials To distinguish and quantify different materials, dual-energy CT is widely used in both the industrial and medical field There are several ap proaches based on using a pair of sources with a different kVp value and a pair of detectors, fast kVp switching of a single source or a single source with a detector consisting of two layers (McCollough et al., 2015) Photon counting detectors enable further extend these ap proaches – either by selecting a detection threshold i.e precisely matching a searched absorption edge or by a fully spectroscopic response opening access to so-called spectral imaging It was already demonstrated that CT using Timepix technology is suitable for the ma terial decomposition of metallic-organic composite materials (Pichotka et al., 2015) Although the industrial field usually relies on highly attenuating materials, the ability of Medipix technology to acquire data with extremely high CNR is profitable in this area as well The Micro-CT laboratory at CET, Czech Republic is equipped with an in-house built high resolution CT system that operates the WidePIX10�10 detector CET has published impressive works on the border between medicine and Fig Prototype of a pinhole-based XRF camera based on the utilizing of a RasPIX device, Mini-X tube and a revolver holding beam-modulating filters mounted on a common plate 1.8.2 XRF mapping in combination with micro-CT Recently, an interesting approach putting together X-ray micro-CT techniques with XRF-based mapping was presented and successfully demonstrated on a Baroque wooden sculpture (Vavrik et al, 2016, Vavrik, 2019) The approach is based on measurements of a CT scan together with X-ray fluorescence photons emitted from the object sur face The measurement was carried out at CET using a patented modular CT scanner called TORATOM equipped with two orthogonally mounted X-ray tubes (Fila et al., 2015) While the first one was used for the CT scan, the other one was used for the excitation of XRF photons (see Fig 11) The XRF signal was detected using a Timepix-based gamma camera containing a 300 μm Si sensor and a 1000 μm CdTe sensor ar ranged into a telescope The XRF images obtained during the scan were then mapped onto the surface of the 3D model obtained after the CT data reconstruction While the CT-based voxel model provides information on the inner structures and overall shape of the objects (Fig 12 left), the XRF mapping helps to identify the elemental composition of surface layers (Fig 12 right) The XRF image targets the elements expected in the pigments – Fe (Kα 6.4 keV), Zn (Kα 8.64 keV) and Ag (Kα 22.16 keV) – coded in red, green and blue, respectively This approach might be extremely useful i.e for the analysis of his toric wooden statues as the one used in the presented proof-of-concept measurement Wooden statues used to be typically decorated by a thin polychrome layer or were sometimes covered by precious metals like silver or gold Even the historical pigments used in polychrome are Fig 10 XRF mapping of the region of interest (approximately � cm2) of a technological copy of the Gothic period painting “Epitaph of Margaret” Photo of the scanned area (left) and an XRF map created from photons with the energy of 10–12 keV where an L-line of lead is expected (right) (Zemlicka, 2016) J Dudak Radiation Measurements 137 (2020) 106409 walls within the GG-BAG sample visualized using the WidePIX10�10 detector in Fig 13 2.1 4D CT The fourth dimension of a CT scan usually means the employment of time It allows the analysis of changes within the investigated object during a certain time period The same object is scanned repeatedly in a series of CT scans and then the differences between individual datasets are observed Since each of the datasets consists typically from hundreds of projections, 4D CT analysis usually places strict requirements on the scanning time The basic assumption of a CT scan is that the object is stable during scanning otherwise motion artifacts are induced The temporal sampling should be, therefore, faster than the fastest expected changes in the object The Shannon sampling theorem should be obeyed in the ideal case Depending on the dynamics of the observed process each of the CT scans has to be captured within a few seconds or even faster A CT scanner suitable for such an approach must fulfill several requirements A powerful radiation source is a necessity as a huge amount of photons has to be delivered in a short time The detector readout has to be very fast to be able to acquire tens or hundreds of frames per second The quantum efficiency of the detector should be as high as possible And finally, the speed of the sample rotation stage has to be considered as well Kumpova et al carried out a 4D CT analysis of the crack development in samples of concrete at the CT laboratory of CET (Kumpova et al., 2016b) The modular CT system TORATOM and large-area Time pix-based detector arranged in a row of chips with a mm thick CdTe sensor capable of reading out 42 frames per second was used for the study The work analyzed the formation and propagation of cracks in imaged samples during a three point bending test with a continuously increasing load Eleven subsequent CT scans with a resolution of ca 36 μm were acquired Each scan consisted of 400 angular projections with the total time of 50 s per scan The tube was operated at 60 kVp with 96W of output power After all the datasets were processed and recon structed, a visualization of differences between subsequent datasets clearly showed the development of cracks and enabled quantitatively analyzing properties of the studied material Kytyr et al used the same set-up for on-fly tomography for the analysis of deformations of GG-BAG bone scaffolds (Kytyr et al., 2017) In this particular work, a modular Timepix-based detector system with a Fig 11 The TORATOM CT system at the Centre of Excellence Telc, Czech Republic, prepared for the scanning of a polychrome wooden baroque sculpture using micro-CT and an XRF mapping (courtesy of Daniel Vavrik, Centre of Excellence Telc) material engineering focused on the micro-CT analysis of biocompatible bone scaffolds fabricated from nanoparticulate bioactive glass rein forced gelan-gum (GG-BAG) and fibrin fibers (Kumpova et al., 2016a; Vavrik et al., 2018) Since both GG-BAG and fibrin exert very low X-ray absorption contrast, it is complicated to visualize them using conven tional X-ray cameras The work concludes that the used Timepix de tector provided three times better detail detectability than a flat-panel detector used for comparison and a spatial resolution of μm was reached in the case of the fibrin sample See the microstructure of cell Fig 12 Example of an obtained micro-CT slice of the scanned sculpture (left) and volume rendering of the reconstructed dataset with an XRF image mapped on its surface (right) While the CT data provides information on the overall shape and inner structures of the object, the XRF mapping helps to analyze the elemental composition of the thin polychrome layer at the surface of the sculpture The XRF image visualizes the presence of the elements expected in the pigments – Fe, Zn and Ag – coded in red, green and blue, respectively (courtesy of Daniel Vavrik, Centre of Excellence Telc) (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) 10 J Dudak Radiation Measurements 137 (2020) 106409 Fig 13 High resolution CT scan of a GG-BAG bone scaffold sample scanned with use of WidePIX10�10 detector The left part shows volume rendering of the whole sample while the right part presents a transversal slice revealing the fine microstructure of scaffold cell walls in great detail and contrast (courtesy of Ivana Kumpova, CET Telc, Czech Republic) fast parallel read-out was used to reach the demanded scanning speed (Vavrik et al., 2014) Four modular devices were assembled together to provide a sensor area of 512 by 512 pixels and simultaneously operated The samples were mounted in a dedicated loading stage and total of 34 datasets was scanned under a gradually increasing load The constructed volumes were then used for digital volume correlation analysis revealing the dislocations of the inner microstructure See Fig 14 illustrating the amount of the deformation of the sample at four different loading states The fastest CT scan performed with a Timepix-based detector was carried out and published by Vavrik et al (2017) The paper presents results of a 4D CT scan of a vitamin C pill being dissolved in water The experimental scan was performed again at the CET X-ray laboratory using a modular Timepix-based detector with parallel read-out capa bility The tube was operated at 60 kVp with 50W output power The source-to-detector distance was set to 150 mm only to achieve maximal possible beam intensity The whole scan consisted of 30000 projections (3.4 ms exposure time) captured within while the sample stage performed 120 revolutions The data income rate was so high that it was not possible to write the data directly to HDD It had to be kept in RAM and was written to HDD offline after the measurement finished The data were then divided into 240 individual CT datasets covering a 180� range each Despite the fact that motion blur caused by the continuous movement of the stage was observed, the dissolving process was suc cessfully captured The authors conclude that even much faster CT scans would be feasible The scanning rate was limited by the rotation table speed and beam intensity while the detector itself could have been capable of capturing data significantly faster 2.2 X-ray imaging in the time delay integration mode The time delay integration (TDI) is a newly available operation mode of Timepix devices potentially useful for scanning of large objects The TDI is a read-out principle widely used by CCD scanners dedicated for the imaging of moving objects Typically a 1D detector (single row of pixels) is used In the case of a Timepix device the TDI mode is suitable for any configuration of by N chips Each point of the TDI-obtained Fig 14 Visualization of bone scaffold deformation at four different loading states The displacement field is color-coded using the scale from blue to red pro portional to the interval 0–1.5 mm dislocation (courtesy of Daniel Kytyr, Institute of Theoretical and Applied Mechanics of the Czech Academy of Sciences) (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) 11 J Dudak Radiation Measurements 137 (2020) 106409 image is created as an integration of 256 sub-acquisitions acquired while the studied object was moving along the detector columns The continuous image is created at detector level by shifting the image matrix by one row after each sub-acquisition is taken The TDI successfully suppresses image defects caused by dead or noisy pixels as each point of the final image is obtained as a sum of 256 sub-acquisitions acquired by different pixels On the other hand, the detector has to be perfectly aligned with the object movement direction and the speed of the object has to be synchronized with acquisition and read-out time to avoid image blurring Considering that the first 256 rows of the final image are produced by a various number of integrated sub-acquisitions, the photon statistic in individual image rows is grad ually changing Therefore, the TDI is effective if the scan length is much larger than the detector height only Fig 15 demonstrates the results of the TDI scan of a laser-cut pattern in a thin steel plate (Zemlicka, 2016) So far, the TDI mode has been successfully tested and is awaiting further application use An obvious field of use can be found in the online inspection of objects on a running belt (i.e at airports, or in factories) mentioned fields obtained using Medipix2 and Timepix devices were demonstrated in this paper Timepix-based detectors provide, thanks to the dark-current-free quantum counting, X-ray images with exception ally high contrast-to-noise ratio The steep point-spread function makes these detectors an ideal choice for high-resolution applications And energy-resolving capabilities open the way to material recognition based on advanced imaging techniques like dual-energy or spectral radiography and CT or XRF-based imaging Such a set of features simultaneously provided by a single device is very unique and, there fore, Timepix detectors are very versatile tools capable of covering a wide range of imaging applications Despite this, Timepix technology is still mostly used in experimental science and is gradually finding its way into the commercial field There are several companies on the market cooperating with the Medipix collaboration on further development of the technology and selling devices in a variety of different versions Medipix and Timepix detectors have also already been utilized in commercially available X-ray imaging systems in several cases Beside RayScan Nano, mentioned in section 2.3 of this paper, a spectral CT scanner MARS (Medipix All Resolution System) for small animal imag ing was introduced by the University of Canterbury, New Zealand (MARS, 2019), and Fraunhofer IIS developed a compact and lightweight micro-CT system called CT-Portable (2018) Performance of the latest available generation – Timepix3 – was not demonstrated in this paper, since wider experience and the applied re sults of using Timepix3 for X-ray imaging has not been published yet The device is used mostly for particle tracking taking advantage of a new mode providing simultaneous information on the position and time stamp of an event X-ray imaging can profit from faster data read-out, lowering the minimal detection threshold and improved spectroscopic capabilities provided by Timepix3 The Timepix4, currently still under development, promises to bring a smaller pixel pitch and 4-sides buttable chips Especially the second feature will be extremely important for X-ray imaging, since the assembling of large-area gap-less detector arrays will become much easier The development of sensor materials proceeds simultaneously with chip electronics Silicon is still very popular, nevertheless, the quality of sensors manufactured from semi-insulating materials con taining high-Z elements like CdTe, CdZnTe, GaAs, etc is better and better Such sensors should overcome the main limitation of silicon – its low quantum efficiency for higher energies Photon counting detectors are on the rise The application range is becoming constantly wider thanks to its versatility and imaging per formance As a relatively new detection technology, it is still under intensive development and the detectors have gradually changed from Conclusions and outlook The aim of the previous chapters was to briefly introduce the prin ciples of high resolution X-ray imaging and present recent applications of hybrid pixel photon-counting detectors of the Medipix family in this field The range of applications is constantly increasing along with the progress of the technology itself, which is still under intensive devel opment, and significantly exceeds the scope of this paper X-ray imaging is just one field of science where Medipix technology has been employed The Medipix device has grown from a chip carrying 4096 pixels (64 � 64 pixels) to large-area detectors with more than 6.5 megapixels (2560 � 2560 pixels) Apart from the increase of the detector area, development has also been focused on building different advanced detection geometries like multiple-layered particle tracking telescopes or Compton cameras New application areas have appeared as these upgrades have been coming Timepix devices are still used at the Large Hadron Collider at CERN and for other high-energy physics experiments But due to its versatility, Timepix technology has also found its way into satellites for the monitoring of radiation belts in outer space, to lowbackground underground laboratories searching for double beta decay and other rare events or even to schools for educational use Concerning X-ray imaging, Medipix and Timepix technology have been proven to be a highly useful tool in life sciences, material engi neering and cultural heritage Interesting applied results from the Fig 15 Laser-cut stainless steel pattern used for the testing of the TDI mode (left) and its X-ray radiography obtained this way (right) The intensity gradually increases at the beginning of the scan as the number of rows used for integration raises A slight blur can be observed at the end of the scan as the motorized stage decelerated Both effects can be avoided by continuous scanning (courtesy of Jan Zemlicka, Institute of Experimental and Applied Physics, Czech Technical University in Prague) 12 J Dudak Radiation Measurements 137 (2020) 106409 experimental devices to easy-to-use and reliable cameras for highresolution X-ray imaging And as energy-sensitive computed tomogra phy becomes a highly demanded technique with potential use in in dustry, life sciences and also in medicine it can be expected that photon counting detectors will gain an important and stable position in the field of X-ray imaging Kytyr, D., et al., 2017 Deformation analysis of gellan-gum based bone scaffold using onthe-fly tomography Mater Des 134 (-), 400–417 https://doi.org/10.1016/j matdes.2017.08.036 Llopart, X., 2002 Medipix2: a 64-k pixel readout chip with 55-μm square elements working in single photon counting mode IEEE Trans Nucl Sci 49 (5), 2279–2283 https://doi.org/10.1109/TNS.2002.803788 Llopart, X., et al., 2007 Timepix, a 65K programmable pixel readout chip for arrival time, energy and/or photon counting measurements Nucl Instrum Methods A 581 (1–2), 485–494 https://doi.org/10.1016/j.nima.2007.08.079 Mars https://www.marsbioimaging.com/mars/, 2019 Mayo, S.C., et al., 2005 Attainment of