Bayesian regularization restoration algorithm for photon counting images

CMOS photon detectors (CPDs) are recently proposed photon sensing devices utilizing the latest CMOS image sensor (CIS) technologies [1]. CPDs are non-electron-multiplying devices, whose pixels have a fully depleted photo diode and have a readout noise of sub electron RMS even at room temperature. Using a 15μm pixel CPD test device coupled to a CsI(Tl) scintillator, we successfully obtained photon-count X-ray images. A Hamamatsu Photonics scintillator with 150μm CsI(Tl) layer coupled to fiber optic plate (FOP) of 3mm thickness is diced in dry condition and directly glued to the sensor surface. X-ray photons are injected from an X-ray tube with accelerating voltage of 30kV and 45kV using W target. Each X-ray photon creates a scintillation light spot in the captured images, where the injected position and photon energy are determined by integrating multiple pixel outputs at that spot. X-ray energy distributions were obtained at both 30kV and 45kV with reasonable differences. Modulated transfer function (MTF) of over 0.7 at 10LP/mm was achieved by mapping injected positions at 30kV. Photon-count images for slanted-edge MTF measurements as well as 10LP/mm of X-ray test chart were achieved. Those photon-count images were compared with conventional energy integrating images obtained with the same sensing device. Both image types confirmed superior resolution with photon-counting. This indirect X-ray photon counting technique using CPD has a potential of getting critical X-ray information for medical applications by achieving accurate injected positions of X-ray photons and their energies simultaneously.

Technological innovations over the past several decades have led to pixelated photon-counting x-ray detectors suitable for medical imaging applications. One component of this research effort is the development of mathematical models of detector performance. Such models provide physical insight and understanding of relationships between the basic physics of photon-counting x-ray detectors and image quality. Models also enable identifying upper limits of detector performance, establishing performance benchmarks, and optimizing detector design. Over the past several decades, the image science community developed a Fourier-based approach to the objective assessment of detector performance. Within this framework, image spatial resolution and image noise are quantified in terms of the modulation transfer function and noise power spectrum, respectively. These metrics are combined in the detective quantum efficiency, which provides an objective measure of the signal-to-noise performance of an x-ray imaging system as a function of spatial frequency. In this chapter, we summarize recent efforts to develop a framework for modelling and understanding of Fourier-based metrics of the performance of photon-counting x-ray detectors. We present a brief introduction to Fourier-based metrics of image quality, present a generic framework for modelling of Fourier-based metrics for photon-counting systems, and then present a simple model of a photon-counting detector that enables understanding relationships between detector design parameters, spatial resolution, and image noise. The development of models of the performance of photon-counting x-ray imaging systems is an ongoing effort and future research is expected to transcend the material presented in this chapter.

The photon counting lidar system with a single photon detector as the core device can reconstruct the target image under low light or high background noise environment. The traditional imaging method requires a long integration time to acquire the target point cloud data and the target depth is extracted by the photon counting distribution histogram. We propose a reconstruction algorithm for photon-counting depth images based on background noise censoring to overcome the shortcomings of traditional methods. According to the different distribution characteristics of background photons and signal photons in time domain, we set a window to examine the photon flight data in this window and choose appropriate thresholds to find signal photon units, then we use computational imaging method to further smooth the reconstructed image. By calculating the root mean square error (RMSE) of reconstructing images using different algorithms, it is known that the results of reconstructing images using proposed algorithm are better than those using traditional maximum likelihood estimation (MLE) algorithm, the imaging accuracy of our method is increased by over 1.4-fold more than the maximum likelihood estimation and improving imaging performance significantly. The experimental results show that the proposed algorithm can effectively improve the reconstructed image of photon counting lidar, and it has positive significance for expanding the application range of photon counting lidar.

We present a novel time-resolved photon counting imaging technique and its use in multi-dimensional luminescence spectroscopy. By using an ultrafast camera coupled to an image intensifier on a microscope, we demonstrate the potential of wide-field time-correlated single photon counting, with a count rate of up to 5 Mhz. This system has the advantage of allowing the detection of single photons in parallel in every pixel. We measured the luminescence decay of Europium Polyoxometalate (POM), and observed contrast on lifetime images of Eu-POM on silver nanocrystals.

We have used an electron-bombarded CCD for optical photon counting imaging. The photon event pulse height distribution was found to be linearly dependent on the gain voltage. We propose on this basis that a gain voltage sweep during exposure in an electron-bombarded sensor would allow photon arrival time determination with sub-frame exposure time resolution. This effectively uses an electron-bombarded sensor as a parallel-processing photoelectronic time-to-amplitude converter, or a two-dimensional photon counting streak camera. Several applications that require timing of photon arrival, including Fluorescence Lifetime Imaging Microscopy, may benefit from such an approach. A simulation of a voltage sweep performed with experimental data collected with different acceleration voltages validates the principle of this approach. Moreover, photon event centroiding was performed and a hybrid 50% Gaussian/Centre of Gravity + 50% Hyperbolic cosine centroiding algorithm was found to yield the lowest fixed pattern noise. Finally, the camera was mounted on a fluorescence microscope to image F-actin filaments stained with the fluorescent dye Alexa 488 in fixed cells.

We develop a method for obtaining 3D polarimetric integral images from elemental images recorded in low light illumination conditions. Since photon-counting images are very sparse, calculation of the Stokes parameters and the degree of polarization should be handled carefully. In our approach, polarimetric 3D integral images are generated using the Maximum Likelihood Estimation and subsequently reconstructed by means of a Total Variation Denoising filter. In this way, polarimetric results are comparable to those obtained in conventional illumination conditions. We also show that polarimetric information retrieved from photon starved images can be used in 3D object recognition problems. To the best of our knowledge, this is the first report on 3D polarimetric photon counting integral imaging.


Photon-counting detectors (PCDs) for CT imaging use energy thresholds to simultaneously acquire projections at multiple energies, making them suitable for spectral imaging and material decomposition. Unfortunately, setting multiple energy thresholds results in noisy analytical reconstructions due to low photon counts in high-energy bins. Iterative reconstruction provides high quality photon-counting CT (PCCT) images but requires enormous computation time for 5D (3D + energy + time) in vivo cardiac imaging. 

Approach. 
We recently introduced UnetU, a deep learning (DL) approach that accurately denoises axial slices from 4D (3D + energy) PCCT reconstructions at various acquisition settings. In this study, we explore UnetU configurations for 5D cardiac PCCT denoising, focusing on singular value decomposition (SVD) modifications along the energy and time dimensions and alternate network architectures such as 3D U-net, FastDVDNet, and Swin Transformer UNet. We compare our networks to multi-energy non-local means (ME NLM), an established PCCT denoising algorithm. 

Main results. 
Our evaluation, using real mouse data and the digital MOBY phantom, revealed that all DL methods were more than 16 times faster than iterative reconstruction. DL denoising with SVD along the energy dimension was most effective, consistently providing low root mean square error and spatio-temporal reduced reference entropic difference, alongside strong qualitative agreement with iterative reconstruction. This superiority was attributed to lower effective rank along the energy dimension than the time dimension in 5D cardiac PCCT reconstructions. ME NLM sometimes outperformed DL with time SVD or time and energy SVD, but lagged behind iterative reconstruction and DL with energy SVD. Among alternate DL architectures with energy SVD, none consistently outperformed UnetU Energy (2D). 

Significance. 
Our study establishes UnetU Energy as an accurate and efficient method for 5D cardiac PCCT denoising, offering a 32-fold speed increase from iterative reconstruction. This advancement sets a new benchmark for DL applications in cardiovascular imaging.

The feasibility of computed tomography (CT) with energy-resolved x-ray detection has been investigated. A breast CT design with multi slit multi slice (MSMS) data acquisition was used for this study. The MSMS CT includes linear arrays of photon counting detectors separated by gaps. This CT configuration allows for efficient scatter rejection and 3D data acquisition. The energy-resolved CT images were simulated using a digital breast phantom and the design parameters of the proposed MSMS CT. The phantom had 14 cm diameter and 50/50 adipose/glandular composition, and included carcinoma, adipose, blood, iodine and CaCO3 as contrast elements. The x-ray technique was 90 kVp tube voltage with 660 mR skin exposure. Photon counting, charge (energy) integrating and photon energy weighting CT images were generated. The contrast-to-noise (CNR) improvement with photon energy weighting was quantified. The dual energy subtracted images of CaCO3 and iodine were generated using a single CT scan at a fixed x-ray tube voltage. The x-ray spectrum was electronically split into low- and high-energy parts by a photon counting detector. The CNR of the energy weighting CT images of carcinoma, blood, adipose, iodine, and CaCO3 was higher by a factor of 1.16, 1.20, 1.21, 1.36 and 1.35, respectively, as compared to CT with a conventional charge (energy) integrating detector. Photon energy weighting was applied to CT projections prior to dual energy subtraction and reconstruction. Photon energy weighting improved the CNR in dual energy subtracted CT images of CaCO3 and iodine by a factor of 1.35 and 1.33, respectively. The combination of CNR improvements due to scatter rejection and energy weighting was in the range of 1.71–2 depending on the type of the contrast element. The tilted angle CZT detector was considered as the detector of choice. Experiments were performed to test the effect of the tilting angle on the energy spectrum. Using the CZT detector with 20° tilting angle decreased the tailing of the measured x-ray spectrum as compared to a conventional CZT detector. It was concluded that the energy-resolved MSMS CT with tilted angle CZT detector is potentially feasible and could provide a unique combination of photon counting, energy weighting, scatter rejection and single kVp dual energy subtraction CT imaging.

The most heavily investigated detector concept for medical photon-counting X-ray imaging, in particular photon-counting X-ray computed tomography (CT), is based on the semiconductors CdTe and CdZnTe (CZT). Alternatively, a detector design based on Si is under investigation. Both approaches have advantages and disadvantages, which leaves room for the development of other types of detector. In this chapter, we investigate the feasibility of scintillation detectors based on silicon photomultipliers (SiPMs) for this application. We first summarize the potential clinical advantages of energy-resolving photon-counting X-ray detectors, the associated challenges, and the working principles and drawbacks of detectors based on CdTe/CZT and Si. We then introduce the concept of SiPM-based scintillation detectors and discuss the scintillator and SiPM properties that determine two fundamental performance metrics, viz. the rate capability and the energy resolution. We have previously developed a pulse shape model that takes all of these properties into account. We use the model to estimate the expected detector performance as a function of pixel size, for LaBr3:Ce and YAP:Ce scintillators coupled to currently available SiPM technology and to a more advanced SiPM prototype. The model calculations show that the short decay time constant of the scintillators, combined with the reduced recharge time constant of the new SiPM prototype, results in pulses short enough that a pixel size of 400 µm (LaBr3:Ce) or 333 µm (YAP:Ce) yields comparable rate capability as a CdTe/CZT detector with a pixel size of 500 µm. The high light yield and nearly proportional response of the investigated scintillators results in good energy resolution, provided that the SiPM does not suffer from excessive saturation when the pixel size is reduced to such small dimensions. We show that the level of saturation remains limited with the new SiPM prototype and that energy resolutions of 12% (LaBr3:Ce) and 15% (YAP:Ce) at 60 keV may be achieved. Our work indicates that it may be feasible to develop SiPM-based scintillation detectors for medical photon-counting X-ray imaging with comparable rate capability and energy resolution as CdTe/CZT detectors. Remaining challenges include a dose-efficient optical isolation between pixels. However, the fact that the detector is based on light transport rather than charge transport means that issues typically associated with CdTe/CZT detectors, such as charge sharing and polarization, do not play a role.

Tailoring CT scan acquisition parameters to individual patients is a topic of much research in the CT imaging community. It is now common place to find automatically adjusted tube current options for modern CT scanners. In addition, the use of beam shaping filters, commonly called bowtie filters, is available on most CT systems and allows for different body regions to receive different incident x-ray fluence distributions. However, no method currently exists which allows for the form of the incident x-ray fluence distribution to change as a function of the view angle. This study represents the first experimental realization of fluence field modulated CT (FFMCT) for a c-arm geometry CT scan. X-ray fluence modulation is accomplished using a digital beam attenuator (DBA). The device is composed of ten iron wedge pairs that modulate the thickness of iron, the x-rays must traverse before reaching a patient. Using this device, experimental data was taken using a Siemens Zeego c-arm scanner. Scans were performed on a cylindrical polyethylene phantom and on two different sections of an anthropomorphic phantom. The DBA was used to equalize the x-ray fluence striking the detector for each scan. Non DBA, or ‘flat field’ scans were also acquired of the same phantom objects for comparison. In addition, a scan was performed in which the DBA was used to enable volume of interest (VOI) imaging. In VOI, only a small sub-volume within a patient receives full dose and the rest of the patient receives a much lower dose. Data corrections unique to using a piece-wise constant modulator were also developed. The feasibility of FFMCT implemented using a DBA device has been demonstrated. Initial results suggest dose reductions of up to 3.6 times relative to ‘flat field’ CT. In addition to dose reduction, the DBA enables a large improvement in image noise uniformity and the ability to provide regionally enhanced signal to noise using VOI imaging techniques. The results presented in this paper take the field of FFMCT from the theoretical stage to that of possible clinical implementation. FFMCT, as shown in this paper, can reduce the patient dose while maintaining or improving image quality. In addition, the DBA has been experimentally shown to be well suited to implement entirely new imaging methods like photon counting and VOI imaging.

Investigation of the LabPET™ detector and electronics for photon-counting CT imaging

We have developed a compact ultra-high-resolution gamma camera (UGC), based on an electron-multiplying charge-coupled device (EMCCD) coupled to a columnar CsI(Tl) scintillator. The EMCCD is suitable for photon-counting gamma-camera imaging, since it has an extremely low readout noise, even at high frame rates (<1 electron/pixel at 50 images/s or 11 Mpixel/s). The high frame rate prevents overlapping of scintillation events and reduces the accumulation of dark current noise. In this paper, we describe the drive- and readout electronics of the EMCCD, the image processing hardware and software, and the first ultra-high-resolution gamma-ray images obtained with the UGC. A digital signal processor (DSP) facilitates real-time frame analysis, comprising photon counting and energy discrimination, and reduces the data stream from 162 kB per frame to 8 B per detected scintillation event (i.e., for a typical application in multipinhole single photon emission computed tomography (SPECT) a data reduction from 28 GB/h by a factor of 20 000 to 1.4 MB/h). Such a reduced data stream is needed for applications that require the use of a large number of gamma cameras simultaneously. The image processing hardware that we describe allows the images from the EMCCD to be processed in real-time, at a rate of 50 images per second. First gamma-camera images with a spatial resolution as good as 60 /spl mu/m full-width of half-maximum (FWHM) are shown. The prototype UGC allows for photon counting without the need for an image intensifier and has energy discriminating capabilities.

We study the influence of the locations of energy thresholds on the basis image noise in spectral X-ray imaging using photon counting devices. Our simulations assume photon counting detectors equipped with up to four discriminator units allowing the binning of photons into up to four distinct energy windows. We use the binned counting data in order to compute the Cramer-Rao lower bound (CRB) for the basis image noise in a photo-effect -Compton-effect decomposition of the attenuation. By studying the variations of the CRB with respect to the threshold energies, we obtain optimized arrangements for each number of energy thresholds and for different values of external copper filtration. We compare the basis image noise for the optimal arrangement with the noise corresponding to an equidistant threshold arrangement and with the results of a tube dual-energy acquisition. Our simulations were performed on two different X-ray paths corresponding to two perpendicular directions through the human thorax. Our results show that in many cases the basis image noise can be strongly reduced by optimizing the locations of the threshold energies.

Compared with the traditional image intensifier with phosphor screen readout, the photon-counting imaging detector with charge induction readout is more beneficial in several aspects (e.g., good imaging properties and time resolution) to astronomy, reconnaissance, bioluminescence, and materials research. However, the annealing temperature during the tube-making process can affect the properties of the Ge film, and consequently impair the performance of the detector. Therefore, the influence of annealing temperature on Ge film and on the detector is studied in order to determine the crucial parameters. The Ge films are prepared on ceramic and quartz glass by the use of an electron gun. They are analyzed by scanning electron microscope (SEM), high-resistance meter, and X-ray diffraction (XRD). The results show that the optimum substrate and annealing temperature are ceramic plate and 250 ?, respectively.

We developed and evaluated an x-ray photon-counting imaging system using an energy-resolving cadmium zinc telluride (CZT) detector coupled with application specific integrated circuit (ASIC) readouts. This x-ray imaging system can be used to identify different materials inside the object. The CZT detector has a large active area (5×5 array of 25 CZT modules, each with 16×16 pixels, cover a total area of 200 mm × 200 mm), high stopping efficiency for x-ray photons (~ 100 % at 60 keV and 5 mm thickness). We explored the performance of this system by applying different energy windows around the absorption edges of target materials, silver and indium, in order to distinguish one material from another. The photon-counting CZT-based x-ray imaging system was able to distinguish between the materials, demonstrating its capability as a radiation-spectroscopic decomposition system.