Ehsan Abadi

The Effect of Contrast Material on Radiation Dose at CT: Part II. A Systematic Evaluation across 58 Patient Models

Purpose To estimate the radiation dose as a result of contrast medium administration in a typical abdominal computed tomographic (CT) examination across a library of contrast material-enhanced computational patient models. Materials and Methods In part II of this study, first, the technique described in part I of this study was applied to enhance the extended cardiac-torso models with patient-specific iodine-time profiles reflecting the administration of contrast material. Second, the patient models were deployed to assess the patient-specific organ dose as a function of time in a typical abdominal CT examination using Monte Carlo simulation. In this hypothesis-generating study, organ dose refers to the total energy deposited in the unit mass of the tissue inclusive of iodine. Third, a study was performed as a strategy to anticipate the biologically relevant dose (absorbed dose to tissue) in highly perfused organs such as the liver and kidney. The time-varying organ-dose increment values relative to those for unenhanced CT examinations were reported. Results The results from the patient models subjected to the injection protocol indicated up to a total 53%, 30%, 35%, 54%, 27%, 18%, 17%, and 24% increase in radiation dose delivered to the heart, spleen, liver, kidneys, stomach, colon, small intestine, and pancreas, respectively. The biologically relevant dose increase with respect to the dose at an unenhanced CT examination was in the range of 0%-18% increase for the liver and 27% for the kidney across 58 patient models. Conclusion The administration of contrast medium increases the total radiation dose. However, radiation dose, while relevant to be included in estimating the risk associated with contrast-enhanced CT, may still not fully characterize the total biologic effects. Therefore, given the fact that many CT diagnostic decisions would be impossible without the use of iodine, this study suggests the need to consider the effect of iodinated contrast material on the organ doses to patients undergoing CT studies when designing CT protocols.

Patient‐specific quantification of image quality: An automated technique for measuring the distribution of organ Hounsfield units in clinical chest CT images

Purpose To develop and validate an automated technique for measuring organ Hounsfield units (HUs) in clinical chest CT images. Materials and methods An automated computer algorithm was developed to measure the distribution of HUs inside four major organs: the lungs, liver, aorta, and spine. These organs were first identified using image processing techniques. Each organ was segmented into multiple regions of interest (ROIs) and characterized in terms of HU values. The medians of the ROI histograms were computed for each dataset. The automated results were validated by assessing their correlation with manual measurements in fifteen contrast‐enhanced and fifteen non‐contrast‐enhanced clinical chest CT datasets. The robustness of the measurements with respect to dependency on image noise and CTDIvol was ascertained. One utility of the approach was further demonstrated in assessing the variability in aorta HUs across 732 patients undergoing noncontrast and contrast‐enhanced examinations. Results The algorithm successfully measured the histograms of the four organs in both contrast and non‐contrast‐enhanced chest CT exams. The automated measurements were in agreement with manual measurements with a near unity slope of the relationship between automated and manual measurements with high coefficient of determination (slope = 0.931–1.003, R2 = 0.89–0.99). Organ median HU measurements were found to be largely independent of both image noise and CTDIvol (P > 0.05), as expected. Across patient cases, the program ran successfully across 95% (697/732) of cases. Aorta median HUs demonstrated five times more variability in contrast‐enhanced exams compared to that in non‐contrast‐enhanced exams. Conclusions Patient‐specific organ HUs can be measured from clinical datasets. The algorithm that was developed can be run on both contrast‐enhanced and non‐contrast‐enhanced clinical datasets. The method can be applied to automatically extract image HU‐contrast characteristics of clinical CT images, not captured in phantom data, whereby enabling quantification and optimization of image quality and contrast administration.

Development of a fast, voxel-based, and scanner-specific CT simulator for image-quality-based virtual clinical trials

This study aimed to develop a simulation framework to synthesize accurate and scanner-specific Computed Tomography (CT) images of voxel-based computational phantoms. Two phantoms were used in the simulations, a geometry-based Mercury phantom and a “textured” anthropomorphic XCAT phantom, both with an isotropic voxel size of 0.25 mm. The simulator geometry and physics were based on a clinical scanner. The projection images were calculated by computing each detector’s signal using the Beer-Lambert law. To avoid aliasing artifacts, the focal spot and detectors were subsampled four and nine times, respectively. The simulator was designed to function both axially and helically, and account for “Z” and in-plane flying focal spots and various bowtie filters. Quantum and electronic noise were added to the detector signals as a function of the tube current using experimental measurements. The resulting projection images were calibrated to suppress the beam hardening artifact using a 4th-order polynomial water correction. The simulation procedure was accelerated using multi-threading and graphics processing unit (GPU) computing. The projection images were reconstructed using clinical reconstruction software. To evaluate the accuracy of the simulator, the reconstructed images of the computational Mercury phantom were compared against experimental CT scans of its physical counterpart in terms of resolution, noise, and HU values. Results showed that our proposed simulator can generate CT images with image quality attributes close to real clinical data. The new CT simulator, combined with anthropomorphic “textured” phantoms, provides a new way to generate clinically realistic CT data and has the potential to enable virtual clinical studies in advance or in lieu of costly clinical trials.

A rapid GPU-based Monte-Carlo simulation tool for individualized dose estimations in CT

The rising awareness towards the risks associated with CT radiation has pushed forward the case for patient- specific dose estimation, one of the prerequisites for individualized monitoring and management of radiation exposure. The established technique of using Monte Carlo simulations to provide such dose estimates is computationally intensive, thus limiting their utility towards timely assessment of clinically relevant questions. To overcome this impediment, we have developed a rapid Monte Carlo simulation tool based on the MC-GPU frame- work for individualized dose estimation in CT. This tool utilizes the multi-threaded x-ray transport capability of MC-GPU, scanner-specific geometry and voxelized patient-specific models to produce realistic estimates of radiation dose. To demonstrate its utility, we utilized this tool to provide scanner-specific (LightSpeed VCT, GE Healthcare) organ dose estimates in abdominopelvic CT for a virtual population of 58 adult XCAT patient models. To gauge the accuracy of these estimates, the organ dose values from this new tool were compared against those from a previously published tool based on PENELOPE framework. The comparisons demonstrated the capability of our new simulation tool to produce dose estimates that agree with the published data within 5% for organs within primary field while simultaneously providing speedups as high as 70x over a CPU cluster-based execution model. This high accuracy of dose estimates coupled with the demonstrated speedup provides a viable model for rapid and personalized dose estimation.

How reliable are texture measurements

The purpose of this study was to assess the bias (objectivity) and variability (robustness) of computed tomography (CT) texture features (internal heterogeneities) across a series of image acquisition settings and reconstruction algorithms. We simulated a series of CT images using a computational phantom with anatomically-informed texture. 288 clinically-relevant simulation conditions were generated representing three slice thicknesses (0.625, 1.25, 2.5 mm), four in-plane pixel sizes (0.4, 0.5, 0.7, 0.9 mm), three dose levels (CTDIvol = 1.90, 3.75, 7.50 mGy), and 8 reconstruction kernels. Each texture feature was sampled with 4 unique volumes of interest (VOIs) (244, 1953, 15625, 125000 mm3). Twenty-one statistical texture features were calculated and compared between the ground truth phantom (i.e., pre-imaging) and its corresponding post-imaging simulations. Metrics of comparison included (1) the percent relative difference (PRD) between the post-imaging simulation and the ground truth, and (2) the coefficient of variation (%COV) across simulated instances of texture features. The PRD and %COV ranged from -100% to 4500%, and 0.8% to 49%, respectively. PRD decreased with increased slice thickness, in-plane pixel size, and dose. The dynamic range of results indicate that image acquisition and reconstruction conditions (i.e., slice thicknesses, in-plane pixel sizes, dose levels, and reconstruction kernels) can lead to significant bias and variability in texture feature measurements.

From patient-informed to patient-specific organ dose estimation in clinical computed tomography

Many hospitals keep a record of dose after each patients CT scan to monitor and manage radiation risks. To facilitate risk management, it is essential to use the most relevant metric, which is the patient-specific organ dose. The purpose of this study was to develop and validate a patient-specific and automated organ dose estimation framework. This framework includes both patient and radiation exposure modeling. From patient CT images, major organs were automatically segmented using Convolutional Neural Networks (CNNs). Smaller organs and structures that were not otherwise segmented were automatically filled in by deforming a matched XCAT phantom from an existing library of models. The organ doses were then estimated using a validated Monte Carlo (PENELOPE) simulation. The segmentation and deformation components of the framework were validated independently. The segmentation methods were trained and validated using 50-patient CT datasets that were manually delineated. The deformation methods were validated using a leave-one-out technique across 50 existing XCAT phantoms that were deformed to create a patient-specific XCAT for each of 50 targets. Both components were evaluated in terms of dice similarity coefficients (DSC) and organ dose. For dose comparisons, a clinical chest-abdomen-pelvis protocol was simulated under fixed tube current (mA). The organ doses were estimated by a validated Monte Carlo package and compared between automated and manual segmentation and between patient-specific XCAT phantoms and their corresponding XCAT targets. Organ dose for phantoms from automated vs. manual segmentation showed a ~2% difference, and organ dose for phantoms deformed by the study vs. their targets showed a variation of ~5% for most organs. These results demonstrate the great potential to assess organ doses in a highly patient-specific manner.

Realistic lesion simulation: application of hyperelastic deformation to lesion-local environment in lung CT

Lesion simulation programs can be used to insert realistic, generated lesions into anatomical images for further study. Most lesion simulation programs rely on insertion of a mask within an otherwise unchangeable, i.e., static surrounding—in reality, lesions deform their immediate surroundings. The goal of the current study was to develop a lesion model based on realistic morphology, but with additional hyperelastic modification of the lesion-local environment in accordance with lesion morphology and location. Physical displacement of the existing tissue was modeled by finite element application of hyperelastic theory to a lung tissue segmentation, incorporating the material properties for both parenchymal and stromal tissue. An observer study was conducted with the data generated from this model to ascertain the realism of hyperelastic and static lesion insertions compared to real lesions. The comparisons were characterized in terms of the area under the ROC curve, AUC. The results indicate that observers are less able to distinguish between hyperelastically-inserted lesions and real ones (AUC=0.62) compared to statically-inserted lesions (AUC=0.75). The findings indicate that hyperelastic deformation offers an improvement in the realism of simulated lesions in CT imaging.

Virtual clinical trial in action: textured XCAT phantoms and scanner-specific CT simulator to characterize noise across CT reconstruction algorithms

Although non-linear CT systems offer improved image quality over conventional linear systems, they disrupt certain assumptions of the dependency of noise and resolution on radiation dose that are true of linear systems. As such, simplistic phantoms do not fully represent the actual performance of current systems in the clinic. Assessing image quality from clinical images address this limitation, but full realization of image quality attributes, particularly noise, requires the knowledge of the exact heterogeneous anatomy of the patient (not knowable) and/or repeated imaging (ethically unattainable). This limitation can be overcome through realistic simulations enabled by virtual clinical trials (VCTs). This study aimed to characterize the noise properties of CT images reconstructed with filtered back-projection (FBP) and non-linear iterative reconstruction (IR) algorithms through a VCT. The study deployed a new generation version of the Extended Cardio-Torso (XCAT) phantom enhanced with anatomically-based intra-organ heterogeneities. The phantom was virtually “imaged” using a scanner-specific simulator, with fifty repeats, and reconstructed using clinical FBP and IR algorithms. The FBP and IR noise magnitude maps and the relative noise reduction maps were calculated to quantify the amount of noise reduction achieved by IR. Moreover, the 2D noise power spectra were measured for both FBP and IR images. The noise reduction maps showed that IR images have lower noise magnitude in uniform regions but higher noise magnitude at edge voxels, thus the noise reduction attributed to IR is less than what could be expected from uniform phantoms (29% versus 60%). This work demonstrates the utility of our CT simulator and “textured” XCAT phantoms in performing VCT that would be otherwise infeasible.

Modeling Lung Architecture in the XCAT Series of Phantoms: Physiologically Based Airways, Arteries and Veins

The purpose of this paper was to extend the extended cardiac-torso (XCAT) series of computational phantoms to include a detailed lung architecture including airways and pulmonary vasculature. Eleven XCAT phantoms of varying anatomy were used in this paper. The lung lobes and initial branches of the airways, pulmonary arteries, and veins were previously defined in each XCAT model. These models were extended from the initial branches of the airways and vessels to the level of terminal branches using an anatomically-based volume-filling branching algorithm. This algorithm grew the airway and vasculature branches separately and iteratively without intersecting each other using cylindrical models with diameters estimated by order-based anatomical measurements. Geometrical features of the extended branches were compared with the literature anatomy values to quantitatively evaluate the models. These features include branching angle, length to diameter ratio, daughter to parent diameter ratio, asymmetrical branching pattern, diameter, and length ratios. The XCAT phantoms were then used to simulate CT images to qualitatively compare them with the original phantom images. The proposed growth model produced 46369 ± 12521 airways, 44737 ± 11773 arteries, and 39819 ± 9988 veins to the XCAT phantoms. Furthermore, the growth model was shown to produce asymmetrical airway, artery, and vein networks with geometrical attributes close to morphometry and model based studies. The simulated CT images of the phantoms were judged to be more realistic, including more airways and pulmonary vessels compared with the original phantoms. Future work will seek to add a heterogeneous parenchymal background into the XCAT lungs to make the phantoms even more representative of human anatomy, paving the way towards the use of XCAT models as a tool to virtually evaluate the current and emerging medical imaging technologies.

Effect of Iodine-based Contrast Material on Radiation Dose at CT

From Shivani Pahwa, MD,* Nicholas K. Schiltz, PhD,† Lee E. Ponsky, MD,‡ Ziang Lu, BA,§ Mark A. Griswold, PhD,*|| Vikas Gulani, MD, PhD*‡|| Departments of Radiology* and Urology,‡ University Hospitals Cleveland Medical Center, Cleveland, Ohio Departments of Population & Quantitative Health Sciences† and Biomedical Engineering,|| Case Western Reserve University, 10900 Euclid Ave, Bolwell B120, Cleveland, OH 44106-0500 e-mail: vikas@case.edu Case Western Reserve University School of Medicine, Cleveland, Ohio§