CZT Photon Counting Computed Tomography: Technology and Applications

Document Objectives

The primary objective of this document is to provide information about the advantages of Redlen’s CZT Photon Counting Detectors (PCDs) technology over traditional Energy Integrated Detectors (EIDs).

The second objective of this document is to provide recommendations for the PCCT Module user on how to configure and use Redlen products. The user is free to explore operation outside the range of these recommendation with the understanding that performance characteristics might differ from the ones tested at Redlen. The document clarifies what is tested at the Redlen factory prior to the product shipment. For any questions related to this document please contact your Redlen program manager or sales representative.

Photon Counting Technology

Introduction

X-ray detectors are the major, and arguably, the most important component of a clinical CT system and have substantial impact on image quality as well as patient radiation dose, which consequently affects a wide range of clinical applications. Compared to the conventional scintillator-based CT detectors using an indirect conversion technology that integrate photon energy deposited by all received x-ray photons (EIDs), photon-counting detectors (PCDs) possess many inherent advantages, which can be largely attributed to its direct conversion technology and energy discrimination capability. Although PCD technology has been widely utilized in nuclear medicine in the last several years, its adoption in X-ray CT system has long been hampered due to the high photon flux effects encountered in clinical CT imaging. The recent dramatic advances in CZT (Cadmium Zinc Tellurium) technology have overcome these problems and CZT based Photon Counting Computed Tomography (PCCT) is now a clinical reality!

In this white paper, we will discuss the potential clinical benefits of PCD-CT that have been demonstrated in a series of recent studies. These include:

• high spatial resolution

• increased iodine contrast-to-noise ratio (CNR)

• increased dose efficiency

• reduced electronic noise

• reduced beam-hardening and metal artifacts

• K-edge contrast imaging

• simultaneous multi-contrast agents

In the following sections we will then discuss how these benefits can impact CT applications in each clinical area, such as cardio, vascular, neuro and musculoskeletal imaging.

High Spatial Resolution

Spatial resolution is much better in PCD detectors than in EID detectors due to much smaller pixels used in PCD. In conventional EID detectors, a reflective layer (septa) of a finite width thickness is incorporated between each scintillating detector pixel, along the detector row and channel directions. Since scintillating detectors create visible light in the process of X-ray detection, the septa is designed to prevent leakage of visible light between adjacent detector pixels. While it is possible to create smaller EID pixels, there is a physical limit to the width of the septa. Smaller detector pixels with a finite-width septa lead to reduced geometric dose efficiency, a consequence of the reduced detector fill factor. Today’s EID scanners typically utilize 1 mm pixel pitch which translates into roughly 580 um spatial resolution at the ISO center.

Further improvement in EID scanner spatial resolution can be obtained using comb filters that are placed near the detector to reduce the pixel aperture, resulting in smaller pixel size for ultra-high-resolution imaging. Though this approach is routinely used for specific imaging tasks such as the CT of the inner ear, it is a highly dose-inefficient approach as the comb filter is placed in front of the detector array, X-rays blocked by the filter have already passed through the patient, contributing to patient dose.

This spatial resolution limitation could be overcome in the direct conversion PCD technology, where no wall between detector pixels is necessary, thereby allowing the design of smaller pixel sizes without losing fill factor and dose efficiency. Typical detector pixel sizes ranging from 300 μm to 500 μm (pixel pitch) have been reported in the literature. Redlen offers both 330 um and 500 um PCCT models.

Reduced Electronic Noise

The first obvious advantage of Redlen’s PCCT detector is lower electronic noise. Image noise in CT mainly originates from two sources: quantum noise and electronic noise. The quantum noise, which is determined by the number of detected photons, can be traced back to the random nature of X-ray photon interactions. The electronic noise comes from the detector electronic circuits within the X-ray detection system and is therefore to a first degree independent of the number of photons reaching detector. The relative significance of quantum noise and electronic noise is determined by the incident photon flux and clinical parameters. For example, in obese patients the number of photons reaching the detector can be very small.

The difference in electronic noise between PCD and EID detectors is illustrated in Fig. 1. In the PCD case one sets an electronic counting threshold (typically at 20 keV) just above the level of electronic noise making PCD almost “noise-free”. However, in the EID case the low energy electronic noise is integrated together with the useful signal reducing signal to noise ratio (SNR).

Fig. 1 Schematic illustration of photon electron pulses processed by PCD detector.

Iodine Contrast

To increase contrast to noise ratio (CNR) CT tests frequently utilize contrast agents. The most common one is Iodine that has a K-edge position at 33 keV. Iodinated contrast is a form of water-soluble, intravenous radiocontrast agent containing iodine, which enhances the visibility of vascular structures and organs during radiographic procedures. Many pathologies, such as cancer, have particularly improved visibility with iodinated contrast. The radiodensity of iodinated contrast is 25–30 Hounsfield units (HU) per milligram of iodine per milliliter at a tube voltage of 100–120 kVp.

PCDs show great advantages over EIDs in iodine detection for the following reasons. In the diagnostic X-ray energy range, the X-ray photons primarily interact with scanned objects through two physical mechanisms: the photoelectric effect and Compton scattering. Theoretically, the attenuation due to photoelectric effect is proportional to the effective atomic number of a material and is inversely proportional to the energy of incident X-ray photon. On the other hand, the energy dependence of Compton effect is relatively flat within diagnostic energy range. Consequently, photoelectric effect contributes more to the attenuation of low-energy photons, while Compton scattering is the dominant effect accounting for high-energy photons. As a result, high-Z materials, such as iodine, have higher photon attenuation in the low-energy range due to the photoelectric effect, and therefore demonstrate higher signal or contrast in CT images.

In conventional EID detectors, the detector signal is proportional to the total energy of all detected X-ray photons. Due to the energy integrating nature of EID, lower-energy photons contribute less to the detector signal than the higher-energy photons. However, higher-energy photons have less information content from high-Z materials such as iodine, as the improved contrast of iodine is mainly explained by photoelectric interaction which manifests its effect in the low energy range. Consequently, the underweighting of signal produced by lower energy photons reduces the CNR of iodine signal in a CT system using EID. On the other hand, PCD counts each individual photon equally regardless of the photon energy, without the energy weighting in EID. Hence, the low-energy photons have a greater contribution to the image contrast for PCD-CT compared to EID-CT, which improves the iodine contrast and CNR

Beam Hardening and Metal Artifacts

Beam hardening is a well-known effect in X-ray imaging. As photons pass through the scanned object, low-energy photons are preferentially attenuated compared to high-energy photons. Since poly-energetic beams are used in CT, this causes the effective photon energy to be shifted towards the higher end of the spectrum. This introduces artifacts, which are typically dark areas adjacent to highly attenuating objects such as cortical bone, affecting the image appearance and CT number accuracy for nearby soft tissues. In the presence of foreign objects such as metal implants, severe image streaking artifacts together with dark/bright regions can be observed across the image.

These metal artifacts have a characteristic appearance, and are caused by multiple physical mechanisms, including X-ray scattering, photon starvation, as well as beam hardening effect. The severity of metal artifacts can vary according to the atomic number, density, and shape of the metal, as well as the surrounding anatomy and scan acquisition/reconstruction parameters. In many cases, the resulting metal artifacts can dramatically obscure critical structures, and result in a significant reduction in diagnostic confidence in various clinical tasks, such as distinguishing pathologic findings from normal structures.

Since each individual photon is sorted according to its energy in a PCD-CT acquisition, an energy-bin image can be reconstructed using only higher-energy photons. Compared to the conventional EID-CT image or the low-energy threshold image in a PCD acquisition, the high-energy bin image is more immune to beam hardening effect in areas around dense bones.

K-edge Imaging and New Contrast Agents

PCCT detectors provide an opportunity to simultaneously discriminate multiple tissue types or contrast materials. Their spectral signatures provides an opportunity for the development of new contrast agents and imaging techniques. Nanoparticle-based contrast agents have been a subject of active research and it is expected that some of these applications will hit the clinical market applications in several years. Several reports using heavy-metal based nanoparticles like Gold, Ytterbium, and Gadolinium as potential CT contrast agents have been published. Individualized nanoparticles that are functionalized and antibody-conjugated to target a specific tissue type or region of interest (e.g., cancerous cells, fibrotic collagenous tissue, macrophages) have been used in combination with PCCT systems to facilitate molecular imaging.

Ability to detect multiple contrast agents relies on so called K-edge imaging. PCDs detectors, due to their ability to discriminate photon energy can measure the position of K-edges of those high-Z materials that fall within the diagnostic CT energy range. The user-defined energy thresholds in PCDs can be placed close to the K-edge energy of high-Z contrast materials in order to capture the discontinuity in attenuation profile. This helps distinguish the K-edge of one material from other materials (e.g., bone, soft tissue, second contrast agent) and more importantly, quantify the concentration of the contrast materials in a given target site using material decomposition techniques. This approach has been mainly tested on rodents imaged using small-animal PCD-CT scanners. References

Despite the promising outcomes of this approach, translation to large animals and subsequently to human trials remains challenging due to the pending evaluation of in vivo biocompatibility of these nanoparticles, high cost of raw material (e.g., gold) and manufacturing, and manufacturing variability for producing large volumes of nanoparticles for large animal or human imaging.

Due to the ability of PCD-CT to capture the spectral signature of multiple K-edge materials in the same acquisition, new reports have emerged demonstrating the potential for multi-phase imaging using single scan and multi-contrast materials. The traditional approach to biphasic imaging involves one contrast injection and two CT scans, occurring at different delays after injection for arterial and venous phase imaging. PCD-CT aims to accomplish multi-phase imaging using a single scan and two contrast agents, each administered using a separate injection at a different time prior to the scan. A single PCD-CT scan occurring tens of seconds after the injection of the initial contrast agent (e.g., iodine) and shortly after the injection of the 2nd contrast agent (e.g., gadolinium) would capture both contrast agents but in different phases (e.g., iodine would have reached the venous phase, while gadolinium would still be in the arterial phase). Individual material maps obtained using material decomposition could then be used to highlight the different enhancement phases (e.g., arterial and venous phases).

Clinical Applications

Musculoskeletal (MSK) Imaging

Vascular Imaging

Thoracic imaging

Neuro Imaging

Conclusions

In this white paper, we have demonstrated the potential benefits of photon counting detectors (PCDs) and its impact in clinical CT imaging. Given the unique capability of counting individual photon and discriminating its energy, photon counting detectors have unique benefits over conventional energy integrating detectors (EIDs), such as reduced electronic noise, increased contrast and contrast to noise ratio, reduced beam hardening and metal artifacts, and simultaneous high-resolution and multi-energy imaging. These benefits have a broad impact on clinical CT exams in terms of image quality and radiation dose. Certain benefits, such as the high resolution, will benefit specific exams and clinical areas more than the others. In addition to the improvement to the current clinical practice, there are also substantial opportunities for PCD-CT to enable new clinical applications, such as nanoparticles and multi-contrast imaging.

Currently, widespread use of PCD-CT in clinical imaging is restricted since there is only a limited number of research PCD-CT systems capable of patient scans, with no commercial scanners available. Applications of this technology will likely be expanded once commercial scanners are available with large scale manufacturing techniques. In addition, extensive ongoing research activities and technical developments to improve PCD performance will also broaden the horizon of clinical applications.

Redlen PCCT Detectors

PCCT Products at Glance

This family of Photon Counting CT products from Redlen is built using a modular approach and share the same core technology. The main product characteristics can be summarized as base unit called “mini-module” (MM) and detector module (DM) blade and logic system as follows:

Base unit called “Mini-Module” (MM) sensor array based on a modular architecture

• Base unit called “Mini-Module” (MM) – 4 way buttable;

• Two-pixel patterns: 330um and 500um pitch designs;

• Supports 16 to 180keV energy range, 6 Energy bins per pixel, <10 keV energy resolution;

• Power at < 2mW per pixel at maximum count rate; and

• Includes logic for charge sharing detection and correction.

Detector module (DM) blade and logic system

• Supports up to 16 PCCT ASICs, aggregates, decimates, and frames data;

• User specified system data interface – multilane LVDS or SFP+ (Twin-ax or Fiber);

• Manages detector configuration and calibration data from Flash;

• Generates low voltage local DC supplies from single system DC supply; and

• Distributes HV from system supply to sensor cathode.

Due to different pixel patterns available, Redlen currently offers two Photon Counting CT Module products: PCCT Model 330 and PCCT Model 500. Other products, using different pixel configuration and sensor thicknesses, might be offered in the future. The basic specifications at the mini-module (MM) and detector module (DM) levels are listed in Figs.1 and 2.

Fig.1 Basic MM PCCT physical specifications: Model 330 and 500.

Fig.2 Basic DM PCCT performance specifications: Model 330 and 500.

PCCT Operation

PCCT operation is recommended to follow the following configuration. The user is free to explore operation beyond factory guaranteed boundaries, but performance stated in this document is in that case not guaranteed. Basic operational regime is shown in Fig.3. The PCCT has 6 energy bins, CC0-CC5, that can be set-up anywhere between 16 and 180 keV. However, for proper operation and characterization PCCT should be constrained as follows.

Fig.3 Operating boundaries of PCCT Model 330

Noise Levels

Schematic representation of the CZT readout system is shown in Figure 4. After converting the X-ray generated current pulse into a voltage pulse via the charge-sensitive amplifier (CSA), and subsequent filtering, the signal is ready for digitization. Redlen uses 6 user selectable threshold voltages operating simultanously to produce a 1-bit trigger signal indicating the detection of the pulse. In parallel, the value of the shaped signal is sent to the ADC converter with 16-bit accuracy. The corresponding digital counters can be read out simultaneously effectively creating multi-energy photon counting system.

Fig.4 Schematic representation of the signal processing chain present in each pixel of the Redlen PCCT module.

A critical property of every electronic system is the noise level, and PCCT electronics is not an exception. In fact, one might argue that the noise level is the critical parameter of the readout system as it must read out individual photons representing electronic charge of femto Coulombs induced onto the CSA. The best representation of the overall system noise in energy resolving detector is the Energy Resolution (ER). For monoenergetic radiation sources, such as Am241 or Co57, the ER is defined as a Full-Width Half Max (FWHM) metric. Typical value of ER characteristics of PCCT is in the 8-10keV range.

There are many consequences of that 8-10keV value:

• PCCT will not be able to resolve X-ray characteristics peaks which would require ER being better than 5keV. Measured K-edge filter slopes will be gradual, not abrupt, and somewhat dependent on the ER value. Measured X-ray tube spectra will not terminate abruptly at kVpp end point under non-pile up conditions, instead they will gradually diminish to zero.

• Minimum detectable photon energy will be 2-3 times higher than ER, which gives 16-30keV range, depending on how false count specification is being formulated (false count being the noise event that triggers the readout system in absence of the real photon being detected by the sensor).

• Recommended minimum width of the energy bin will be twice ER plus uncertainty in the threshold voltage position (which is 1-2keV as shown later) resulting in minimum bin width required to be 18-24keV depending on how aggressive uniformity specifications are set.

Energy bins

Although PCCT has 6 energy bins, 2 of those are used for diagnostic purposes. Only 4 bins are being qualified in terms of stability and uniformity and are subject to factory acceptance tests (FAT). One can argue that 4 energy bins is more than enough for spectral CT applications.

The first bin, Bin0, is used to monitor noise levels. Bin0 monitors the level of noise below 30keV and is set-up in the 16-30keV range. The reason for the particular value of 16keV selected for the lower bound is explained in Appendix 3 where charge-sharing correction is also being discussed.

The last bin, Bin5, is used to monitor X-ray tube ripple effects. Many X-ray tube controllers do not maintain constant kVpp while focusing on keeping the X-ray flux constant and produce voltage ripple. As a result, these kVpp movements lead to spectral distortions that are degrading spectral CT performance. In order to enable monitoring of these fluctuation affecting measured results, Bin5 is being set-up to be equal to kVpp - typically 120 keV..

With a 90keV energy range between 30keV (noise) and 120keV (kVpp), and each bin having a min width of 20keV, there is a limited number of ways to configure 4 unused bins. A configuration which attempts to equalize photon counts across the bins, is proposed as follows:

• Bin 1 – 30 to 50keV, Bin 2 – 50 to 70keV, Bin 3 – 70 to 90keV, Bin 4 – 90 to 120keV

Fig.5 shows simulated X-ray spectrumin the scenarios of no filter, 1mm Cu filter and 2mm Cu filter test cases. As expected, the spectrum gets more hardened as Cu thickness increases. As indicated by Fig.6 when the “no filter” is used Bin1 (35%) and Bin2 (32%) have the highest count, for the “1mm filter” case Bin2 (45%) and Bin3 (31%) are the highest counting, and finally for the “2mm Cu” case Bin 3 has the highest count (38%) followed by Bin4 (32%) and Bin2 (27%).

Fig.5 Simulated 120kVpp transmitted X-ray spectra for cases of i) no filter, ii) 1mm Cu filter, and iii) 2mm Cu filter. Counts normalized to represent the same number of photons in each case (left) and not normalized (right).

Fig.6 Energy bin counts (%) corresponding to the plots of Fig.5

Fig.7 Simulated 120kVpp received X-ray spectra for cases of i) no filter, ii) 1mm Cu filter, and iii) 2mm Cu filter. ER=10keV and 40% charge-sharing factor assumed. Not- normalized counts (left) and normalized (right).

Fig.8 Energy bin counts (%) corresponding to the plots of Fig.7.

The results presented in Figs. 5 and 6 are for ideal transmitted spectra generated using on-line tool (https://www.oem-xray-components.siemens.com/x-ray-spectra-simulation) assuming zero kVpp ripple and tungsten anode material. More realistic received spectra detected by the PCCT are shown in Figs. 7 and 8. These results have been obtained using Redlen custom Matlab code that deals with finite energy resolution and charge-shared events.

Comparing X-ray spectra of Figs. 5 and 7 it is clear, and expected, that the received spectra do not show any characteristic X-ray peaks due to finite energy resoluton of the PCCT detectors. Comparing Fig. 8 to 9 it is clear that energy bins counts shift to lower energies, again as expected, due to charge-sharing effects. The choice of the filter used in testing is difficult as one tries to mimic various attenuation scenarios during CT scan.

Flux Rates

PCCT models 330 and 500 operate at high flux incoming rates typical for Computed Tomography (CT). Like every other photon counting system they both have limitation on how well they count individual photons at very high rates due to pile-up effects. PCCT 330 can “count” until 600 Mcps/mm2 and PCCT 500 until 300 Mcps/mm2, as was indicated in Fig3. Both models follow non-paralyzable dead time model. More information is provided in Appendix D.

The incoming flux or count rate, typically refered as ICR, is limited by the strength of the X-ray sources and distance to the detector. As Redlen operates 25mA COMET X-ray tubes and need to maintain 30cm distance to the detector its maximum ICR value is 245 Mcps/mm2. This is the limit of testability in production as indicated in Fig.3.

In addition to the limitation on the high side there is a limitation on the low side of the flux rate. While PCCT will operate fine during CT scanning even with few photons in each energy bin due to power of software iterative reconstructive algorithms the FAT testing at Redlen is not benefitting from that software. In order to provide decent statistics for stability and uniformity measurements there is a limit on how short the exposure period can be. Assuming standard 1ms view time, the count rate of 0.1 Mcps/mm2 implies a count of 100 photons distributed across 6 energy bins, giving an average photon count of 15. Needless to say this is very low statistics. For the aggressive view time of 0.1ms, which might be typical for heart CT scans, these photon counts would get 10x smaller resulting in non-acceptable test condition. Therefore, 1ms view time and count rates higher 0.1 Mcps/mm2 are typically used in Factory Acceptance Testing. In addition to the consideration related to the photon statistics, instability of the X-ray tube for tube currents lower then 1 mA is known and any measurments performed under these conditions should be handeled with sceptisism.

Conclusions

With anenergy resolution (ER) of the system of 8-10keV, the minimum energy detection bin can be established at the 25-30keV range, and a minimum energy bin width of 20keV. These limitations are schematically shown in Fig. 9.

<<<The PCCT detector configuration setup allows for the detection of ….>>>

• Bin 1 – 30 to 50keV, Bin 2 – 50 to 70keV, Bin 3 – 70 to 90keV, Bin 4 – 90 to 120keV

Fig. 9 Schematic illustration of the energy bin limitations. Each threshold is typically faced with 2keV uncertainty due to non-ideal detector behavior and finite resolution of the PCCT calibration. Each energy bin faces 9keV uncertainty transition on each of its side due to typical energy resolution of the readout system being 8-10keV.