Spectral Angiography Material Decomposition Using an Empirical Forward Model and a Dictionary-Based Regularization

Monte-Carlo study of contrast-enhanced spectral mammography with cadmium telluride photon-counting x-ray detectors

BACKGROUND

Contrast-enhanced spectral mammography (CESM) with photon-counting x-ray detectors (PCDs) can be used to improve the classification of breast cancers as benign or malignant. Commercially-available PCD-based mammography systems use silicon-based PCDs. Cadmium-telluride (CdTe) PCDs may provide a practical advantage over silicon-based PCDs because they can be implemented as large-area detectors that are more easily adaptable to existing mammography systems.

PURPOSE

The purpose of this work is to optimize CESM implemented with CdTe PCDs and to investigate the influence of the number of energy bins, electronic noise level, pixel size, and anode material on image quality.

METHODS

We developed a Monte Carlo model of the energy-bin-dependent modulation transfer functions (MTFs) and noise power spectra, including spatioenergetic noise correlations. We validated model predictions using a CdTe PCD with analog charge summing for charge-sharing suppression. Using the ideal-observer detectability, we optimized CESM for the task of detecting a 7-mm-diameter iodine nodule embedded in a breast with 50% glandularity. We optimized the tube voltage, beam filtration, and the location of energy thresholds for 50 and 100- μ $\mu$ m pixels, tungsten and molybdenum anodes, and two electronic noise levels. One of the electronic noise levels was that of the experimental system; the other was half that of the experimental system. Optimization was performed for CdTe PCDs with two or three energy bins. We also estimated the impact of anatomic noise due to background parenchymal enhancement and computed the minimum detectable iodine area density in the presence of quantum and anatomic noise.

RESULTS

Model predictions of the MTFs and noise power spectra agreed well with experiment. For optimized systems, adding a third energy bin increased quantum noise levels and reduced detectability by ∼55% compared to two-bin approaches that simply suppress contrast between fibroglandular and adipose tissue. Decreasing the electronic noise standard deviation from 3.4 to 1.7 keV increased iodine detectability by ∼5% and ∼30% for two-bin imaging and three-bin imaging, respectively. After optimizing for tube voltage, beam filtration, and the location of energy thresholds, there was ∼a 3% difference in iodine detectability between molybdenum and tungsten anodes for two-bin imaging, but for three-bin imaging, molybdenum anodes provided up to 14% increase in detectability relative to tungsten anodes. Anatomic noise decreased iodine detectability by 15% to 40%, with greater impact for lower electronic noise settings and larger pixel sizes.

CONCLUSIONS

For CESM implemented with CdTe PCDs, (1) quantitatively-accurate three-material decompositions using three energy bins are associated with substantial increases in quantum noise relative to two-energy-bin approaches that simply suppress contrast between fibroglandular and adipose tissues; (2) tungsten and molybdenum anodes can provide nearly equal iodine detectability for two-bin imaging, but molybdenum provides a modest detectability advantage for three-bin imaging provided that all other technique parameters are optimized; (3) reducing pixel sizes from 100 to 50  μ $\mu$ m can reduce detectability by up to 20% due to charge sharing; (4) anatomic noise due to background parenchymal enhancement is estimated to have a substantial impact on lesion visibility, reducing detectability by approximately 30%.

Systematic Review on Learning-based Spectral CT

Spectral computed tomography (CT) has recently emerged as an advanced version of medical CT and significantly improves conventional (single-energy) CT. Spectral CT has two main forms: dual-energy computed tomography (DECT) and photon-counting computed tomography (PCCT), which offer image improvement, material decomposition, and feature quantification relative to conventional CT. However, the inherent challenges of spectral CT, evidenced by data and image artifacts, remain a bottleneck for clinical applications. To address these problems, machine learning techniques have been widely applied to spectral CT. In this review, we present the state-of-the-art data-driven techniques for spectral CT.

Signal-difference-to-noise comparison of temporal subtraction, kV-switching dual-energy and photon-counting dual-energy x-ray angiography

BACKGROUND

Dual-energy (DE) x-ray angiography with photon-counting detectors (PCDs) may enable single-exposure DE imaging of coronary vasculature.

PURPOSE

To compare the iodine signal-difference-to-noise ratio (SDNR) of single-exposure DE angiography with digital subtraction angiography (DSA) and kV-switching DE angiography for matched patient x-ray exposure.

METHODS

In a phantom study, we determined the technique parameters that maximized the iodine SDNR per root entrance air kerma for DSA, kV-switching DE angiography and single-exposure DE angiography. We measured SDNR from images of a phantom consisting of an iodine step-wedge immersed in a water tank of either 20  or 30 cm in thickness. We also imaged a phantom with simulated vessels embedded in background clutter and measured vessel SDNR. For this second phantom, we also applied anti-correlated noise reduction (ACNR) and calculated the resulting iodine SDNR. All images were acquired using a cadmium telluride PCD with two energy bins and analog charge summing for charge sharing suppression. The energy-discrimination capabilities were only used for the single-exposure DE approach. Optimized techniques were compared in terms of SDNR per root air kerma for two levels of x-ray scatter.

RESULTS

For the same patient x-ray exposure, the SDNR of single-exposure DE imaging without ACNR was 75% to 85% of that of kV-switching DE imaging (also without ACNR) and DSA, the latter two of which had nearly equal SDNR. The single-exposure DE approach required ∼50% of the tube load of the kV-switching approach to achieve the same SDNR. For matched patient air kermas, the single exposure approach required only ∼25% of the tube load of the kV-switching approach. ACNR increased SDNR by 2.4 and 3.0 for kV-switching and single-exposure DE imaging, respectively.

CONCLUSIONS

Photon-counting, single-exposure DE angiography can suppress soft tissues and provide iodine SDNR levels comparable to DSA and kV-switching DE angiography for matched patient radiation exposures. When ACNR is used to reduce DE image noise, the SDNR of single-exposure DE imaging and kV-switching DE imaging exceed that of DSA by more than a factor of two. Compared to kV-switching DE imaging, single-exposure DE imaging requires substantially lower tube loading to achieve the same SDNR.

Spectral photon counting for panoramic dental imaging

Panoramic x-ray imaging is a versatile, low-dose imaging tool, which is routinely used for dental applications. In this work, we explore a further improvement of the concept by introducing recently developed spectral photon-counting detector technology into a conventional panoramic imaging unit. In addition we adapt spectral material decomposition algorithms to panoramic imaging needs. Finally, we provide first experimental results, demonstrating decomposition of an anthropomorphic head phantom into soft tissue and dentin basis material panoramic images, while keeping the noise level acceptable using regularization approaches. The obtained results reveal a potential benefit of spectral photon-counting technology also for dental imaging applications.

Experimental optimization of single-exposure dual-energy angiography with photon-counting x-ray detectors

BACKGROUND

Photon-counting x-ray detectors may enable single-exposure dual-energy (DE) x-ray angiography.

PURPOSE

The purpose of this paper is to experimentally optimize the energy thresholds and tube voltage for single-exposure DE x-ray angiography.

METHODS

We optimized single-exposure DE x-ray angiography using the iodine signal-difference-to-noise ratio (SDNR) per root patient air kerma (κ) as a figure of merit. We measured the iodine SDNR by imaging an iodine stepwedge immersed in a water tank with a depth of 30 cm in the direction of x-ray propagation. The stepwedge was imaged using tube voltages ranging from 90 to 150 kV and a cadmium telluride (CdTe) x-ray detector with two energy bins and analog charge summing for charge sharing suppression. The energy threshold that separates the two energy bins was varied from approximately 35 keV to approximately 75% of the maximum energy of the x-ray beam. Curve fitting was used to determine the threshold that maximized SDNR / κ $\mathrm{SDNR}/\sqrt {\kappa }$ . The effect of scatter was determined from measurements of the scatter-to-primary ratios (SPRs) of the low-energy and high-energy images and a semi-empirical model of the relationship between SDNR and SPR. Using the optimal parameters, we imaged a phantom with vessel-simulating structures and background clutter.

RESULTS

The optimal energy thresholds increased monotonically from ∼50 to ∼85 keV over the range of tube voltages considered. For tube voltages greater than 90 kV, the optimal energy thresholds consistently allocated approximately two thirds of all detected primary photons to the low energy bin; this ratio was preserved without scatter. Consistent with prior modeling studies, SDNR / κ $\mathrm{SDNR}/\sqrt {\kappa }$ increased monotonically with tube voltage from 90 to 150 kV; SDNR / κ $\mathrm{SDNR}/\sqrt {\kappa }$ at 150 kV was approximately 38% higher than that at 90 kV for an iodine area density of ∼50 mg/cm² . Scatter reduced SDNR by approximately 25% for SPRs of ∼1 and 0.4 in low-energy and high-energy images, respectively.

CONCLUSIONS

Achieving optimal image quality in single-exposure DE angiography with photon-counting x-ray detectors will require high tube voltages (i.e., >130 kV) and, for thick patients, energy thresholds that allocate approximately two thirds of all primary photons to the low-energy image. Future work will compare the image quality of singe-exposure photon-counting and kV-switching approaches to DE x-ray angiography.

Theoretical comparison of energy-resolved and digital-subtraction angiography

BACKGROUND

X-ray coronary angiography is a sub-optimal vascular imaging technique because it cannot suppress un-enhanced anatomy that may obscure the visualization of coronary artery disease.

PURPOSE

The purpose of this paper is to evaluate the theoretical image quality of energy-resolved x-ray angiography (ERA) implemented with spectroscopic x-ray detectors (SXDs), which may overcome limitations of x-ray coronary angiography.

METHODS

We modeled the large-area signal-difference-to-noise (SDNR) of contrast-enhanced vasculature in ERA images and compared with that of digital-subtraction angiography (DSA), which served as a gold standard vascular imaging technique. To this end, we used calibrated numerical models of the response of cadmium telluride SXDs including the effects of charge sharing, electronic noise, and energy thresholding. Our models assumed zero scatter, no pulse pile up and small signals such that image contrast is approximately linear in the area density of contrast agents. For DSA, we similarly modeled x-ray detection by cesium iodide energy-integrating detectors using validated numerical models. For ERA, we investigated iodine and gadolinium (Gd) contrast agents, two-material and three-material decompositions, analog charge summing for charge sharing correction, and optimized image quality with respect to the tube voltage and energy thresholds assuming cadmium telluride SXDs with three energy bins.

RESULTS

Our analysis reveals that a three-material decomposition using iodine contrast agents will require x-ray exposures that are approximately 400 times those of DSA to achieve the same SDNR as DSA in coronary applications, and is therefore not feasible in a clinical setting. However, three-material decompositions with Gd contrast agents have the potential to provide SDNR that is ∼45% of that of DSA for matched patient air kerma. For two-material decompositions that suppress soft-tissue, ERA has the potential to produce images with SDNR that is 50%-75% of that of DSA for matched patient air kermas but lower levels of tube loading. Achieving these SDNR levels will require the use of analog charge summing for charge sharing correction, which increased SDNR by up to a factor of 1.7 depending on the contrast agent and whether or not a two-material or three-material decomposition was assumed.

CONCLUSIONS

We conclude that three-material ERA implemented with Gd contrast agents and two-material ERA implemented with either iodine or Gd contrast agents, should be investigated as alternatives to DSA in situations where motion artifacts preclude the use of DSA, such as in coronary imaging.

Spectral Photon Counting CT: Imaging Algorithms and Performance Assessment

Photon counting x-ray detectors (PCDs) with spectral capabilities have the potential to revolutionize computed tomography (CT) for medical imaging. The ideal PCD provides accurate energy information for each incident x-ray, and at high spatial resolution. This information enables material-specific imaging, enhanced radiation dose efficiency, and improved spatial resolution in CT images. In practice, PCDs are affected by non-idealities, including limited energy resolution, pulse pileup, and cross talk due to charge sharing, K-fluorescence, and Compton scattering. In order to maximize their performance, PCDs must be carefully designed to reduce these effects and then later account for them during correction and post-acquisition steps. This review article examines algorithms for using PCDs in spectral CT applications, including how non-idealities impact image quality. Performance assessment metrics that account for spatial resolution and noise such as the detective quantum efficiency (DQE) can be used to compare different PCD designs, as well as compare PCDs with conventional energy integrating detectors (EIDs). These methods play an important role in enhancing spectral CT images and assessing the overall performance of PCDs.

Dual-Energy X-Ray Dark-Field Material Decomposition

Dual-energy imaging is a clinically well-established technique that offers several advantages over conventional X-ray imaging. By performing measurements with two distinct X-ray spectra, differences in energy-dependent attenuation are exploited to obtain material-specific information. This information is used in various imaging applications to improve clinical diagnosis. In recent years, grating-based X-ray dark-field imaging has received increasing attention in the imaging community. The X-ray dark-field signal originates from ultra small-angle scattering within an object and thus provides information about the microstructure far below the spatial resolution of the imaging system. This property has led to a number of promising future imaging applications that are currently being investigated. However, different microstructures can hardly be distinguished with current X-ray dark-field imaging techniques, since the detected dark-field signal only represents the total amount of ultra small-angle scattering. To overcome these limitations, we present a novel concept called dual-energy X-ray dark-field material decomposition, which transfers the basic material decomposition approach from attenuation-based dual-energy imaging to the dark-field imaging modality. We develop a physical model and algorithms for dual-energy dark-field material decomposition and evaluate the proposed concept in experimental measurements. Our results suggest that by sampling the energy-dependent dark-field signal with two different X-ray spectra, a decomposition into two different microstructured materials is possible. Similar to dual-energy imaging, the additional microstructure-specific information could be useful for clinical diagnosis.

Single spectrum three-material decomposition with grating-based x-ray phase-contrast CT

Grating-based x-ray phase-contrast imaging provides three simultaneous image channels originating from a single image acquisition. While the phase signal provides direct access to the electron density in tomography, there is additional information on sub-resolutional structural information which is called dark-field signal in analogy to optical microscopy. The additional availability of the conventional attenuation image qualifies the method for implementation into existing diagnostic routines. The simultaneous access to the attenuation coefficient and the electron density allows for quantitative two-material discrimination as demonstrated lately for measurements at a quasi-monochromatic compact synchrotron source. Here, we investigate the transfer of the method to conventional polychromatic x-ray sources and the additional inclusion of the dark-field signal for three-material decomposition. We evaluate the future potential of grating-based x-ray phase-contrast CT for quantitative three-material discrimination for the specific case of early stroke diagnosis at conventional polychromatic x-ray sources. Compared to conventional CT, the method has the potential to discriminate coagulated blood directly from contrast agent extravasation within a single CT acquisition. Additionally, the dark-field information allows for the clear identification of hydroxyapatite clusters due to their micro-structure despite a similar attenuation as the applied contrast agent. This information on materials with sub-resolutional microstructures is considered to comprise advantages relevant for various pathologies.

Radon Inversion via Deep Learning

The Radon transform is widely used in physical and life sciences, and one of its major applications is in medical X-ray computed tomography (CT), which is significantly important in disease screening and diagnosis. In this paper, we propose a novel reconstruction framework for Radon inversion with deep learning (DL) techniques. For simplicity, the proposed framework is denoted as iRadonMAP, i.e., inverse Radon transform approximation. Specifically, we construct an interpretable neural network that contains three dedicated components. The first component is a fully connected filtering (FCF) layer along the rotation angle direction in the sinogram domain, and the second one is a sinusoidal back-projection (SBP) layer, which back-projects the filtered sinogram data into the spatial domain. Next, a common network structure is added to further improve the overall performance. iRadonMAP is first pretrained on a large number of generic images from the ImageNet database and then fine-tuned with clinical patient data. The experimental results demonstrate the feasibility of the proposed iRadonMAP framework for Radon inversion.

A theoretical framework for comparing noise characteristics of spectral, differential phase-contrast and spectral differential phase-contrast x-ray imaging

Spectral and grating-based differential phase-contrast (DPC) x-ray imaging are two emerging technologies that offer additional information compared with conventional attenuation-based x-ray imaging. In the case of spectral imaging, energy-resolved measurements allow the generation of material-specific images by exploiting differences in the energy-dependent attenuation. DPC imaging uses the phase shift that an x-ray wave exhibits when traversing an object as contrast generation mechanism. Recently, we have investigated the combination of these two imaging techniques (spectral DPC imaging) and demonstrated potential advantages compared with spectral imaging. In this work, we present a noise analysis framework that allows the prediction of (co-) variances and noise power spectra for all three imaging methods. Moreover, the optimum acquisition parameters for a particular imaging task can be determined. We use this framework for a performance comparison of all three imaging methods. The comparison is focused on (projected) electron density images since they can be calculated with all three imaging methods. Our study shows that spectral DPC imaging enables the calculation of electron density images with strongly reduced noise levels compared with the other two imaging methods for a large range of clinically relevant pixel sizes. In contrast to conventional DPC imaging, there are no long-range noise correlations for spectral DPC imaging. This means that excessive low frequency noise can be avoided. We confirm the analytical predictions by numerical simulations.

Spectral Differential Phase Contrast X-Ray Radiography

We investigate the combination of two emerging X-ray imaging technologies, namely spectral imaging and differential phase contrast imaging. By acquiring spatially and temporally registered images with several different X-ray spectra, spectral imaging can exploit differences in the energy-dependent attenuation to generate material selective images. Differential phase contrast imaging uses an entirely different contrast generation mechanism: The phase shift that an X-ray wave exhibits when traversing an object. As both methods can determine the (projected) electron density, we propose a novel material decomposition algorithm that uses the spectral and the phase contrast information simultaneously. Numerical experiments show that the combination of these two imaging techniques benefits from the strengths of the individual methods while the weaknesses are mitigated: Quantitatively accurate basis material images are obtained and the noise level is strongly reduced, compared to conventional spectral X-ray imaging.

Contrast-enhanced spectral mammography with a compact synchrotron source

For early breast cancer detection, mammography is nowadays the commonly used standard imaging approach, offering a valuable clinical tool for visualization of suspicious findings like microcalcifications and tumors within the breast. However, due to the superposition of anatomical structures, the sensitivity of mammography screening is limited. Within the last couple of years, the implementation of contrast-enhanced spectral mammography (CESM) based on K-edge subtraction (KES) imaging helped to improve the identification and classification of uncertain findings. In this study, we introduce another approach for CESM based on a two-material decomposition, with which we expect fundamental improvements compared to the clinical procedure. We demonstrate the potential of our proposed method using the quasi-monochromatic radiation of a compact synchrotron source-the Munich Compact Light Source (MuCLS)-and a modified mammographic accreditation phantom. For direct comparison with the clinical CESM approach, we also performed a standard dual-energy KES at the MuCLS, which outperformed the clinical CESM images in terms of contrast-to-noise ratio (CNR) and spatial resolution. However, the dual-energy-based two-material decomposition approach achieved even higher CNR values. Our experimental results with quasi-monochromatic radiation show a significant improvement of the image quality at lower mean glandular dose (MGD) than the clinical CESM. At the same time, our study indicates the great potential for the material-decomposition instead of clinically used KES to improve the quantitative outcome of CESM.

Metric-guided regularisation parameter selection for statistical iterative reconstruction in computed tomography

As iterative reconstruction in Computed Tomography (CT) is an ill-posed problem, additional prior information has to be used to get a physically meaningful result (close to ground truth if available). However, the amount of influence of the regularisation prior is crucial to the outcome of the reconstruction. Therefore, we propose a scheme for tuning the strength of the prior via a certain image metric. In this work, the parameter is tuned for minimal histogram entropy in selected regions of the reconstruction as histogram entropy is a very basic approach to characterise the information content of data. We performed a sweep over different regularisation parameters showing that the histogram entropy is a suitable metric as it is well behaved over a wide range of parameters. The parameter determination is a feedback loop approach we applied to numerically simulated FORBILD phantom data and verified with an experimental measurement of a micro-CT device. The outcome is evaluated visually and quantitatively by means of root mean squared error (RMSE) and structural similarity (SSIM) for the simulation and visually for the measured sample (no ground truth available). The final reconstructed images exhibit noise-suppressed iterative reconstruction. For both datasets, the optimisation is robust where its initial value is concerned. The parameter tuning approach shows that the proposed metric-driven feedback loop is a promising tool for finding a suitable regularisation parameter in statistical iterative reconstruction.