Advanced Techniques for Image Quality Assessment of Modern X-ray Computed Tomography Systems

X-ray computed tomography (CT) imaging constitutes one of the most widely used diagnostic tools in radiology today with nearly 85 million CT examinations performed in the U.S in 2011. CT imparts a relatively high amount of radiation dose to the patient compared to other x-ray imaging modalities and as a result of this fact, coupled with its popularity, CT is currently the single largest source of medical radiation exposure to the U.S. population. For this reason, there is a critical need to optimize CT examinations such that the dose is minimized while the quality of the CT images is not degraded. This optimization can be difficult to achieve due to the relationship between dose and image quality. All things being held equal, reducing the dose degrades image quality and can impact the diagnostic value of the CT examination.

A recent push from the medical and scientific community towards using lower doses has spawned new dose reduction technologies such as automatic exposure control (i.e., tube current modulation) and iterative reconstruction algorithms. In theory, these technologies could allow for scanning at reduced doses while maintaining the image quality of the exam at an acceptable level. Therefore, there is a scientific need to establish the dose reduction potential of these new technologies in an objective and rigorous manner. Establishing these dose reduction potentials requires precise and clinically relevant metrics of CT image quality, as well as practical and efficient methodologies to measure such metrics on real CT systems. The currently established methodologies for assessing CT image quality are not appropriate to assess modern CT scanners that have implemented those aforementioned dose reduction technologies.

Thus the purpose of this doctoral project was to develop, assess, and implement new phantoms, image quality metrics, analysis techniques, and modeling tools that are appropriate for image quality assessment of modern clinical CT systems. The project developed image quality assessment methods in the context of three distinct paradigms, (a) uniform phantoms, (b) textured phantoms, and (c) clinical images.

The work in this dissertation used the “task-based” definition of image quality. That is, image quality was broadly defined as the effectiveness by which an image can be used for its intended task. Under this definition, any assessment of image quality requires three components: (1) A well defined imaging task (e.g., detection of subtle lesions), (2) an “observer” to perform the task (e.g., a radiologists or a detection algorithm), and (3) a way to measure the observer’s performance in completing the task at hand (e.g., detection sensitivity/specificity).

First, this task-based image quality paradigm was implemented using a novel multi-sized phantom platform (with uniform background) developed specifically to assess modern CT systems (Mercury Phantom, v3.0, Duke University). A comprehensive evaluation was performed on a state-of-the-art CT system (SOMATOM Definition Force, Siemens Healthcare) in terms of noise, resolution, and detectability as a function of patient size, dose, tube energy (i.e., kVp), automatic exposure control, and reconstruction algorithm (i.e., Filtered Back-Projection– FPB vs Advanced Modeled Iterative Reconstruction– ADMIRE). A mathematical observer model (i.e., computer detection algorithm) was implemented and used as the basis of image quality comparisons. It was found that image quality increased with increasing dose and decreasing phantom size. The CT system exhibited nonlinear noise and resolution properties, especially at very low-doses, large phantom sizes, and for low-contrast objects. Objective image quality metrics generally increased with increasing dose and ADMIRE strength, and with decreasing phantom size. The ADMIRE algorithm could offer comparable image quality at reduced doses or improved image quality at the same dose (increase in detectability index by up to 163% depending on iterative strength). The use of automatic exposure control resulted in more consistent image quality with changing phantom size.

Based on those results, the dose reduction potential of ADMIRE was further assessed specifically for the task of detecting small (<=6 mm) low-contrast (<=20 HU) lesions. A new low-contrast detectability phantom (with uniform background) was designed and fabricated using a multi-material 3D printer. The phantom was imaged at multiple dose levels and images were reconstructed with FBP and ADMIRE. Human perception experiments were performed to measure the detection accuracy from FBP and ADMIRE images. It was found that ADMIRE had equivalent performance to FBP at 56% less dose.

Using the same image data as the previous study, a number of different mathematical observer models were implemented to assess which models would result in image quality metrics that best correlated with human detection performance. The models included naïve simple metrics of image quality such as contrast-to-noise ratio (CNR) and more sophisticated observer models such as the non-prewhitening matched filter observer model family and the channelized Hotelling observer model family. It was found that non-prewhitening matched filter observers and the channelized Hotelling observers both correlated strongly with human performance. Conversely, CNR was found to not correlate strongly with human performance, especially when comparing different reconstruction algorithms.

The uniform background phantoms used in the previous studies provided a good first-order approximation of image quality. However, due to their simplicity and due to the complexity of iterative reconstruction algorithms, it is possible that such phantoms are not fully adequate to assess the clinical impact of iterative algorithms because patient images obviously do not have smooth uniform backgrounds. To test this hypothesis, two textured phantoms (classified as gross texture and fine texture) and a uniform phantom of similar size were built and imaged on a SOMATOM Flash scanner (Siemens Healthcare). Images were reconstructed using FBP and a Sinogram Affirmed Iterative Reconstruction (SAFIRE). Using an image subtraction technique, quantum noise was measured in all images of each phantom. It was found that in FBP, the noise was independent of the background (textured vs uniform). However, for SAFIRE, noise increased by up to 44% in the textured phantoms compared to the uniform phantom. As a result, the noise reduction from SAFIRE was found to be up to 66% in the uniform phantom but as low as 29% in the textured phantoms. Based on this result, it clear that further investigation was needed into to understand the impact that background texture has on image quality when iterative reconstruction algorithms are used.

To further investigate this phenomenon with more realistic textures, two anthropomorphic textured phantoms were designed to mimic lung vasculature and fatty soft tissue texture. The phantoms (along with a corresponding uniform phantom) were fabricated with a multi-material 3D printer and imaged on the SOMATOM Flash scanner. Scans were repeated a total of 50 times in order to get ensemble statistics of the noise. A novel method of estimating the noise power spectrum (NPS) from irregularly shaped ROIs was developed. It was found that SAFIRE images had highly locally non-stationary noise patterns with pixels near edges having higher noise than pixels in more uniform regions. Compared to FBP, SAFIRE images had 60% less noise on average in uniform regions for edge pixels, noise was between 20% higher and 40% lower. The noise texture (i.e., NPS) was also highly dependent on the background texture for SAFIRE. Therefore, it was concluded that quantum noise properties in the uniform phantoms are not representative of those in patients for iterative reconstruction algorithms and texture should be considered when assessing image quality of iterative algorithms.

The move beyond just assessing noise properties in textured phantoms towards assessing detectability, a series of new phantoms were designed specifically to measure low-contrast detectability in the presence of background texture. The textures used were optimized to match the texture in the liver regions actual patient CT images using a genetic algorithm. The so called “Clustured Lumpy Background” texture synthesis framework was used to generate the modeled texture. Three textured phantoms and a corresponding uniform phantom were fabricated with a multi-material 3D printer and imaged on the SOMATOM Flash scanner. Images were reconstructed with FBP and SAFIRE and analyzed using a multi-slice channelized Hotelling observer to measure detectability and the dose reduction potential of SAFIRE based on the uniform and textured phantoms. It was found that at the same dose, the improvement in detectability from SAFIRE (compared to FBP) was higher when measured in a uniform phantom compared to textured phantoms.

The final trajectory of this project aimed at developing methods to mathematically model lesions, as a means to help assess image quality directly from patient images. The mathematical modeling framework is first presented. The models describe a lesion’s morphology in terms of size, shape, contrast, and edge profile as an analytical equation. The models can be voxelized and inserted into patient images to create so-called “hybrid” images. These hybrid images can then be used to assess detectability or estimability with the advantage that the ground truth of the lesion morphology and location is known exactly. Based on this framework, a series of liver lesions, lung nodules, and kidney stones were modeled based on images of real lesions. The lesion models were virtually inserted into patient images to create a database of hybrid images to go along with the original database of real lesion images. ROI images from each database were assessed by radiologists in a blinded fashion to determine the realism of the hybrid images. It was found that the radiologists could not readily distinguish between real and virtual lesion images (area under the ROC curve was 0.55). This study provided evidence that the proposed mathematical lesion modeling framework could produce reasonably realistic lesion images.

Based on that result, two studies were conducted which demonstrated the utility of the lesion models. The first study used the modeling framework as a measurement tool to determine how dose and reconstruction algorithm affected the quantitative analysis of liver lesions, lung nodules, and renal stones in terms of their size, shape, attenuation, edge profile, and texture features. The same database of real lesion images used in the previous study was used for this study. That database contained images of the same patient at 2 dose levels (50% and 100%) along with 3 reconstruction algorithms from a GE 750HD CT system (GE Healthcare). The algorithms in question were FBP, Adaptive Statistical Iterative Reconstruction (ASiR), and Model-Based Iterative Reconstruction (MBIR). A total of 23 quantitative features were extracted from the lesions under each condition. It was found that both dose and reconstruction algorithm had a statistically significant effect on the feature measurements. In particular, radiation dose affected five, three, and four of the 23 features (related to lesion size, conspicuity, and pixel-value distribution) for liver lesions, lung nodules, and renal stones, respectively. MBIR significantly affected 9, 11, and 15 of the 23 features (including size, attenuation, and texture features) for liver lesions, lung nodules, and renal stones, respectively. Lesion texture was not significantly affected by radiation dose.

The second study demonstrating the utility of the lesion modeling framework focused on assessing detectability of very low-contrast liver lesions in abdominal imaging. Specifically, detectability was assessed as a function of dose and reconstruction algorithm. As part of a parallel clinical trial, images from 21 patients were collected at 6 dose levels per patient on a SOMATOM Flash scanner. Subtle liver lesion models (contrast = -15 HU) were inserted into the raw projection data from the patient scans. The projections were then reconstructed with FBP and SAFIRE (strength 5). Also, lesion-less images were reconstructed. Noise, contrast, CNR, and detectability index of an observer model (non-prewhitening matched filter) were assessed. It was found that SAFIRE reduced noise by 52%, reduced contrast by 12%, increased CNR by 87%. and increased detectability index by 65% compared to FBP. Further, a 2AFC human perception experiment was performed to assess the dose reduction potential of SAFIRE, which was found to be 22% compared to the standard of care dose.

In conclusion, this dissertation provides to the scientific community a series of new methodologies, phantoms, analysis techniques, and modeling tools that can be used to rigorously assess image quality from modern CT systems. Specifically, methods to properly evaluate iterative reconstruction have been developed and are expected to aid in the safe clinical implementation of dose reduction technologies.

Advanced Applications of 3D Dosimetry and 3D Printing in Radiation Therapy

As complex radiotherapy techniques become more readily-practiced, comprehensive 3D dosimetry is a growing necessity for advanced quality assurance. However, clinical implementation has been impeded by a wide variety of factors, including the expense of dedicated optical dosimeter readout tools, high operational costs, and the overall difficulty of use. To address these issues, a novel dry-tank optical CT scanner was designed for PRESAGE 3D dosimeter readout, relying on 3D printed components and omitting costly parts from preceding optical scanners. This work details the design, prototyping, and basic commissioning of the Duke Integrated-lens Optical Scanner (DIOS).

The convex scanning geometry was designed in ScanSim, an in-house Monte Carlo optical ray-tracing simulation. ScanSim parameters were used to build a 3D rendering of a convex ‘solid tank’ for optical-CT, which is capable of collimating a point light source into telecentric geometry without significant quantities of refractive-index matched fluid. The model was 3D printed, processed, and converted into a negative mold via rubber casting to produce a transparent polyurethane scanning tank. The DIOS was assembled with the solid tank, a 3W red LED light source, a computer-controlled rotation stage, and a 12-bit CCD camera. Initial optical phantom studies show negligible spatial inaccuracies in 2D projection images and 3D tomographic reconstructions. A PRESAGE 3D dose measurement for a 4-field box treatment plan from Eclipse shows 95% of voxels passing gamma analysis at 3%/3mm criteria. Gamma analysis between tomographic images of the same dosimeter in the DIOS and DLOS systems show 93.1% agreement at 5%/1mm criteria. From this initial study, the DIOS has demonstrated promise as an economically-viable optical-CT scanner. However, further improvements will be necessary to fully develop this system into an accurate and reliable tool for advanced QA.

Pre-clinical animal studies are used as a conventional means of translational research, as a midpoint between in-vitro cell studies and clinical implementation. However, modern small animal radiotherapy platforms are primitive in comparison with conventional linear accelerators. This work also investigates a series of 3D printed tools to expand the treatment capabilities of the X-RAD 225Cx orthovoltage irradiator, and applies them to a feasibility study of hippocampal avoidance in rodent whole-brain radiotherapy.

As an alternative material to lead, a novel 3D-printable tungsten-composite ABS plastic, GMASS, was tested to create precisely-shaped blocks. Film studies show virtually all primary radiation at 225 kVp can be attenuated by GMASS blocks of 0.5cm thickness. A state-of-the-art software, BlockGen, was used to create custom hippocampus-shaped blocks from medical image data, for any possible axial treatment field arrangement. A custom 3D printed bite block was developed to immobilize and position a supine rat for optimal hippocampal conformity. An immobilized rat CT with digitally-inserted blocks was imported into the SmART-Plan Monte-Carlo simulation software to determine the optimal beam arrangement. Protocols with 4 and 7 equally-spaced fields were considered as viable treatment options, featuring improved hippocampal conformity and whole-brain coverage when compared to prior lateral-opposed protocols. Custom rodent-morphic PRESAGE dosimeters were developed to accurately reflect these treatment scenarios, and a 3D dosimetry study was performed to confirm the SmART-Plan simulations. Measured doses indicate significant hippocampal sparing and moderate whole-brain coverage.

A Model of Lung Tumor Angiogenesis in a Biomimetic Poly(ethylene glycol)-based Hydrogel System

Tumor angiogenesis is critical to tumor growth and metastasis, yet much is unknown about the role vascular cells play in the tumor microenvironment. A major outstanding challenge associated with studying tumor angiogenesis is that existing preclinical models are limited in their recapitulation of in vivo cellular organization in 3D. This disparity highlights the need for better approaches to study the dynamic interplay of relevant cells and signaling molecules as they are organized in the tumor microenvironment. In this thesis, we combined 3D culture of lung adenocarcinoma cells with adjacent 3D microvascular cell culture in 2-layer cell-adhesive, proteolytically-degradable poly(ethylene glycol) (PEG)-based hydrogels to study tumor angiogenesis and the impacts of neovascularization on tumor cell behavior.

In initial studies, 344SQ cells, a highly metastatic, murine lung adenocarcinoma cell line, were characterized alone in 3D in PEG hydrogels. 344SQ cells formed spheroids in 3D culture and secreted proangiogenic growth factors into the conditioned media that significantly increased with exposure to transforming growth factor beta 1 (TGF-β1), a potent tumor progression-promoting factor. Vascular cells alone in hydrogels formed tubule networks with localized activated TGF-β1. To study cancer cell-vascular cell interactions, the engineered 2-layer tumor angiogenesis model with 344SQ and vascular cell layers was employed. Large, invasive 344SQ clusters developed at the interface between the layers, and were not evident further from the interface or in control hydrogels without vascular cells. A modified model with spatially restricted 344SQ and vascular cell layers confirmed that observed 344SQ cluster morphological changes required close proximity to vascular cells. Additionally, TGF-β1 inhibition blocked endothelial cell-driven 344SQ migration.

Two other lung adenocarcinoma cell lines were also explored in the tumor angiogenesis model: primary tumor-derived metastasis-incompetent, murine 393P cells and primary tumor-derived metastasis-capable human A549 cells. These lung cancer cells also formed spheroids in 3D culture and secreted proangiogenic growth factors into the conditioned media. Epithelial morphogenesis varied for the primary tumor-derived cell lines compared to 344SQ cells, with far less epithelial organization present in A549 spheroids. Additionally, 344SQ cells secreted the highest concentration of two of the three angiogenic growth factors assessed. This finding correlated to 344SQ exhibiting the most pronounced morphological response in the tumor angiogenesis model compared to the 393P and A549 cell lines.

Overall, this dissertation demonstrates the development of a novel 3D tumor angiogenesis model that was used to study vascular cell-cancer cell interactions in lung adenocarcinoma cell lines with varying metastatic capacities. Findings in this thesis have helped to elucidate the role of vascular cells in tumor progression and have identified differences in cancer cell behavior in vitro that correlate to metastatic capacity, thus highlighting the usefulness of this model platform for future discovery of novel tumor angiogenesis and tumor progression-promoting targets.

A Generic Model of Execution for Synthesizing Domain-Specific Models

Software engineering researchers are challenged to provide increasingly more pow- erful levels of abstractions to address the rising complexity inherent in software solu- tions. One new development paradigm that places models as abstraction at the fore- front of the development process is Model-Driven Software Development (MDSD). MDSD considers models as first class artifacts, extending the capability for engineers to use concepts from the problem domain of discourse to specify apropos solutions. A key component in MDSD is domain-specific modeling languages (DSMLs) which are languages with focused expressiveness, targeting a specific taxonomy of problems. The de facto approach used is to first transform DSML models to an intermediate artifact in a HLL e.g., Java or C++, then execute that resulting code. Our research group has developed a class of DSMLs, referred to as interpreted DSMLs (i-DSMLs), where models are directly interpreted by a specialized execution engine with semantics based on model changes at runtime. This execution engine uses a layered architecture and is referred to as a domain-specific virtual machine (DSVM). As the domain-specific model being executed descends the layers of the DSVM the semantic gap between the user-defined model and the services being provided by the underlying infrastructure is closed. The focus of this research is the synthesis engine, the layer in the DSVM which transforms i-DSML models into executable scripts for the next lower layer to process. The appeal of an i-DSML is constrained as it possesses unique semantics contained within the DSVM. Existing DSVMs for i-DSMLs exhibit tight coupling between the implicit model of execution and the semantics of the domain, making it difficult to develop DSVMs for new i-DSMLs without a significant investment in resources. At the onset of this research only one i-DSML had been created for the user- centric communication domain using the aforementioned approach. This i-DSML is the Communication Modeling Language (CML) and its DSVM is the Communication Virtual machine (CVM). A major problem with the CVM’s synthesis engine is that the domain-specific knowledge (DSK) and the model of execution (MoE) are tightly interwoven consequently subsequent DSVMs would need to be developed from inception with no reuse of expertise. This dissertation investigates how to decouple the DSK from the MoE and sub- sequently producing a generic model of execution (GMoE) from the remaining appli- cation logic. This GMoE can be reused to instantiate synthesis engines for DSVMs in other domains. The generalized approach to developing the model synthesis com- ponent of i-DSML interpreters utilizes a reusable framework loosely coupled to DSK as swappable framework extensions. This approach involves first creating an i-DSML and its DSVM for a second do- main, demand-side smartgrid, or microgrid energy management, and designing the synthesis engine so that the DSK and MoE are easily decoupled. To validate the utility of the approach, the SEs are instantiated using the GMoE and DSKs of the two aforementioned domains and an empirical study to support our claim of reduced developmental effort is performed.

3D considerations for motion perception and visuomotor transformations

Moving through a stable, three-dimensional world is a hallmark of our motor and perceptual experience. This stability is constantly being challenged by movements of the eyes and head, inducing retinal blur and retino-spatial misalignments for which the brain must compensate. To do so, the brain must account for eye and head kinematics to transform two-dimensional retinal input into the reference frame necessary for movement or perception. The four studies in this thesis used both computational and psychophysical approaches to investigate several aspects of this reference frame transformation. In the first study, we examined the neural mechanism underlying the visuomotor transformation for smooth pursuit using a feedforward neural network model. After training, the model performed the general, three-dimensional transformation using gain modulation. This gave mechanistic significance to gain modulation observed in cortical pursuit areas while also providing several testable hypotheses for future electrophysiological work. In the second study, we asked how anticipatory pursuit, which is driven by memorized signals, accounts for eye and head geometry using a novel head-roll updating paradigm. We showed that the velocity memory driving anticipatory smooth pursuit relies on retinal signals, but is updated for the current head orientation. In the third study, we asked how forcing retinal motion to undergo a reference frame transformation influences perceptual decision making. We found that simply rolling one's head impairs perceptual decision making in a way captured by stochastic reference frame transformations. In the final study, we asked how torsional shifts of the retinal projection occurring with almost every eye movement influence orientation perception across saccades. We found a pre-saccadic, predictive remapping consistent with maintaining a purely retinal (but spatially inaccurate) orientation perception throughout the movement. Together these studies suggest that, despite their spatial inaccuracy, retinal signals play a surprisingly large role in our seamless visual experience. This work therefore represents a significant advance in our understanding of how the brain performs one of its most fundamental functions.

Multi-Scale Visualization and Interpretation of Geological Relationships

Visualization and interpretation of geological observations into a cohesive geological model are essential to Earth sciences and related fields. Various emerging technologies offer approaches to multi-scale visualization of heterogeneous data, providing new opportunities that facilitate model development and interpretation processes. These include increased accessibility to 3D scanning technology, global connectivity, and Web-based interactive platforms. The geological sciences and geological engineering disciplines are adopting these technologies as volumes of data and physical samples greatly increase. However, a standardized and universally agreed upon workflow and approach have yet to properly be developed. In this thesis, the 3D scanning workflow is presented as a foundation for a virtual geological database. This database provides augmented levels of tangibility to students and researchers who have little to no access to locations that are remote or inaccessible. A Web-GIS platform was utilized jointly with customized widgets developed throughout the course of this research to aid in visualizing hand-sized/meso-scale geological samples within a geologic and geospatial context. This context is provided as a macro-scale GIS interface, where geophysical and geodetic images and data are visualized. Specifically, an interactive interface is developed that allows for simultaneous visualization to improve the understanding of geological trends and relationships. These developed tools will allow for rapid data access and global sharing, and will facilitate comprehension of geological models using multi-scale heterogeneous observations.

Fast 2D/3D object representation with growing neural gas

This work presents the design of a real-time system to model visual objects with the use of self-organising networks. The architecture of the system addresses multiple computer vision tasks such as image segmentation, optimal parameter estimation and object representation. We first develop a framework for building non-rigid shapes using the growth mechanism of the self-organising maps, and then we define an optimal number of nodes without overfitting or underfitting the network based on the knowledge obtained from information-theoretic considerations. We present experimental results for hands and faces, and we quantitatively evaluate the matching capabilities of the proposed method with the topographic product. The proposed method is easily extensible to 3D objects, as it offers similar features for efficient mesh reconstruction.