Browsing by Subject "Reconstruction"

Cone-beam computed tomography (CBCT) is an X-ray-based imaging modality widely used in medical practices. Due to the ionizing imaging dose induced by CBCT, many studies were conducted to reduce the number of projections (sparse sampling) to lower the imaging dose while maintaining good image quality and fast reconstruction speed. Conventionally, a CBCT volume is reconstructed analytically with the Feldkamp Davis Kress (FDK) algorithm that backprojects filtered projections according to projection angles. However, the FDK algorithm requires a dense angular sampling that satisfies the Shannon-Nyquist theorem. The FDK algorithm reconstructs CBCT with a high speed but requires relatively high patient imaging dose. The iterative methods like algebraic reconstruction technique (ART) and compressed sensing (CS) methods are investigated to reduce patient imaging dose. These iterative methods update estimated images iteratively and the CS methods apply penalty terms to award desired features. Yet these methods are limited by the iterative design with substantially increased computation time and consumption of computation power. Scholars have also conducted research on bypassing the limit of Shannon-Nyquist theorem by interpolating densely sampled CBCT projections from sparsely sampled projections. However, blurred structures in reconstructed images remain to be a concern for analytical interpolation methods. As such, previous research indicates that it is hard to achieve the three goals of lowered patient imaging dose, good image quality, and fast reconstruction speed all at once.

As deep learning (DL) gained popularity in fields like computer vision and data science, scholars also applied DL techniques in medical image processing. Studies on DL-based CT image reconstruction have yielded encouraging results, but GPU memory limitation made it challenging to apply DL techniques on CBCT reconstruction.

In this dissertation, we hypothesize that the image quality of CBCT reconstructed from under-sampled projections (low-dose) using deep learning techniques can be comparable to that of CBCT reconstructed from fully sampled projections for treatment verification in radiation therapy. This dissertation proposes that by applying DL techniques in pre-processing, reconstruction, and post-processing stages, the challenge of improving CBCT image quality with low imaging dose and fast reconstruction speed can be mitigated.

The dissertation proposed a geometry-guided deep learning (GDL) technique, which is as the first technique to perform end-to-end CBCT reconstruction from sparsely sampled projections and demonstrated its feasibility for CBCT reconstruction from real patient projection data. In this study, we have found that incorporating geometry information into the DL technique can effectively reduce the model size, mitigating the memory limitation in CBCT reconstruction. The novel GDL technique is composed of a GDL reconstruction module and a post-processing module. The GDL reconstruction module learns and performs projection-to-image domain transformation by replacing the traditional single fully connected layer with an array of small fully connected layers in the network architecture based on the projection geometry. The additional deep learning post-processing module further improves image quality after reconstruction.

This dissertation further optimizes the number of beamlets used in the GDL technique through a geometry-guided multi-beamlet deep learning (GMDL) technique. In addition to connecting each pixel in the projection domain to beamlet points along the central beamlet in the image domain as GDL does, these smaller fully connected layers in GMDL connect each pixel to beamlets peripheral to the central beamlet based on the CT projection geometry. Due to the limitation of GPU VRAM, the proposed technique is demonstrated through low-dose CT image reconstruction and is compared with the GDL technique and a large fully connected layer-based reconstruction method.

In addition, the dissertation also investigates deep learning-based CBCT projection interpolation and proposes a patient-independent deep learning projection interpolation technique for CBCT reconstruction. Different from previous studies that interpolate phantom or simulated data, the proposed technique is demonstrated to work on real patient projection data with unevenly distributed projection angles. The proposed technique re-slices the stack of interpolated projections axially, and each acquired slice is processed by a deep residual U-Net (DRU) model to augment the slice’s image quality. The resulting slices are reassembled into a stack of densely-sampled projections to be reconstructed into a CBCT volume. A second DRU model further post-processes the reconstructed CBCT volume to improve the image quality.

In summary, a geometry-guided deep learning (GDL) technique was proposed as the first deep learning technique for end-to-end CBCT reconstruction from sparsely sampled real patient projection data. The geometry-guided multi-beamlet deep learning (GMDL) technique further optimizes the number of beamlets based on the GDL technique. A patient-independent deep learning projection interpolation technique was also proposed for the pre-processing and post-processing stage of CBCT reconstruction.

In conclusion, the work presented in this dissertation demonstrates the feasibility of improving CBCT image quality with low imaging dose and fast reconstruction speed. The techniques developed in this dissertation also have great potential for clinical applications to enhance CBCT imaging for radiation therapy.

In this dissertation I study the long-term effects of post-disaster reconstruction of infrastructure on economic and demographic outcomes. The effects on individuals and communities that result from shocks to existing infrastructure have not been widely explored in the economic and development literature. As some of the largest natural disasters in recent times have shown, massive destruction of infrastructure is followed by large influxes of resources aimed at the reconstruction of damaged property. For example, after the 2004 Indian Ocean tsunami, Indonesia alone received enough aid to deal with the estimated seven billion dollars in infrastructure losses. While there are studies that address how money was allocated, there is hardly any good empirical evidence that provides a causal estimate of the effect that large reconstruction programs have on targeted beneficiaries. In this dissertation I address this gap in the literature.

The context of my study is the 2004 Indian Ocean tsunami, one of the most devastating natural disasters in recent years. In particular, the location for this analysis is the Indonesian province of Aceh, which was the area hardest hit by the disaster (Chapter 2). One of the main reasons why long-term impacts of post-disaster reconstruction remain an understudied topic is the lack of access to data that tracks individuals over time and across space. Having longitudinal data of this type provides a more complete picture of beneficiaries of post-disaster aid, as well as the effects of reconstruction programs on economic outcomes and demographic processes, such as migration. My dissertation addresses this concern by using a unique, population representative panel of survivors of the Indian Ocean tsunami, the Study of the Tsunami Aftermath and Recovery (STAR), which collected extensive individual, household, and community data in Aceh, Indonesia, every year between 2005 and 2010, with an additional follow-up in 2015 (Chapter 3).

Using these data, the first question I explore empirically is an estimation of the causal effects of reconstruction of the housing stock on a multidimensional set of well-being measures (Chapter 4). First, I show that post-tsunami reconstruction was largely determined by the level of damage, regardless of pre-tsunami characteristics of communities, households, and individuals. Based on this finding, I identify the causal effects of housing reconstruction on post-disaster well-being using an individual fixed effects strategy. I show that housing reconstruction causes significant reductions in levels of post-traumatic stress reactivity, and significant increases in socioeconomic well-being. These effects are mainly concentrated after two years of housing tenure, and among those from highly damaged communities. Housing reconstruction has a positive relationship with self-rated physical health (although these estimates are not statistically significant). These results provide important causal evidence of how reconstruction of infrastructure after a natural disaster can have long-lasting, positive consequences for the recovery of survivors.

Next, I continue looking at the effects of rebuilding individual assets (i.e. the home) but turn to the analysis of migration, a key demographic process following natural disasters. Specifically, I look at migration and its relationship with housing reconstruction and well-being (Chapter 5). The 2004 Indian Ocean tsunami displaced large numbers of people. In Aceh, Indonesia, an estimated 500,000 people left their communities after the disaster. In this research, we provide a demographic perspective on displacement and longer-term adaptation and recovery after a disaster. We describe patterns of mobility among tsunami survivors, including those who did not return to their origin communities, those who did return, and those who never left. We also consider mobility among those living in communities that did not suffer tsunami damage. We then examine how the likelihood of receiving housing aid varies across these subgroups. Finally, we consider how measures of subjective well-being evolve after the disaster. Results show that predictors of relocation vary significantly across individuals depending on the level of exposure of communities to the physical damage of the tsunami. Relocation decisions, and in particular staying in the pre-tsunami community, are highly related to the likelihood of benefiting from housing aid. And, changes in subjective well-being not only depend on receipt of housing aid but also on interactions between relocation decisions.

The last empirical analysis changes the focus from the reconstruction of individual assets to the reconstruction of community infrastructure (Chapter 6}), an important component of post-disaster rebuilding programs. In the aftermath of the tsunami, it is estimated that a total of 2,600 km of roads and 119 bridges needed rebuilding. In less than four years a total of 3,700 km of roads and all the destroyed (or damaged) bridges had been rebuilt \citep{indonesia2010provincial}. Roads can be an important gateway to economic development, so in this analysis I focus on estimating the economic effects of road reconstruction in post-tsunami Aceh. First, I exploit variation in timing of road reconstruction projects at the community level and, using a fixed effects strategy, I show that road reconstruction may not be enough to cause significant economic effects, but that quality of road construction matters, specifically access to all-weather roads. Further, I also show that road reconstruction that happens in combination with public works programs has additional positive effects. I provide further evidence on the effects of road reconstruction by looking at the specific case of the reconstruction of the Banda Aceh-Meulaboh road. The Banda Aceh-Meulaboh road is a good example of a project that seeks to restore large public infrastructure after a major shock to the built environment under the assumption that it would contribute to restore economic activity in the area. Using a difference-in-differences strategy, I exploit changes in access to the road between 2005 and 2015. I show that gaining access to the road has positive and modest effects both on individuals and households and, in particular, on households in rural areas. I did not find any statistically significant negative effects of losing access to the road but results from this case study point that losing access may be hindering some progress, for example, to translate work opportunities into higher wages.

Taken together, results from the empirical analyses in this dissertation fill an important gap in our understanding of what happens to disaster victims in the long-run, how they benefit from reconstruction programs that rebuild both individual and community assets, and how these programs can have long-lasting consequences on economics and demographic trajectories of populations. As a result, my study not only represents an important contribution to existing literature, but it also underscores the importance of having data collection projects that account for the long-term nature of infrastructure reconstruction projects. Natural disasters are projected to become increasingly more common, and this type of data can result in empirical research, like this dissertation, that can improve our understanding of how disaster victims cope, which strategies work best and why, and create lessons that can inform disaster management and reconstruction policies that will result in successful post-disaster experiences.

Narrating Infanticide: Constructing the Modern Gendered State in Nineteenth-Century America traces how modern ideas about gender and race became embedded in the institutions of law and government between the Revolution and the end of Reconstruction. Contemporary understandings of gender and race actually consolidated only in the aftermath of the Civil War, as communities embraced beliefs that women and African Americans constituted distinctive groups with shared, innate characteristics related solely to the fact that they were female or racially different. People then applied these ideas about gender and race to all arenas of life, including the law.

Yet understanding the roles of women and African Americans through universalizing legal conceptions of gender and/or race--conceptions that crystallized in law only in the wake of the Civil War--elides the complexity of the ways in which antebellum communities responded to the interactions of women, the enslaved, and free blacks with the legal system. My study's focus on infanticide, a crime that could only be perpetrated by females, reveals how women--and men--of all races involved themselves in the day-to-day legal processes that shaped the daily lives of Americans during the early republic and antebellum periods. Communities responded to cases of infant death informed by understandings of motherhood and child mortality specific to that particular case and individual, rather than shaping outcomes--as they began to do so after the Civil War--based on broad assumptions about the race or gender of the offender. My conclusions are drawn from almost one hundred cases of infanticide and infant death between 1789 and 1877 gleaned primarily from court records and newspapers in Connecticut, Illinois, and North Carolina. In addition, the study draws on reports of other instances from around the nation, as narrated in sources such as diaries, periodicals, newspapers, crime pamphlets, and medical journals.

In this work, the simulation (forward problem) and reconstruction (inverse problem) in Thermoacoustic Tomography (TAT) are studied using a pseudospectral time-domain (PSTD) method with 4th-order time integration.

The objective of the TAT simulation is to solve for the thermoacoustic pressure field in an inhomogeneous medium. Using the PSTD method, the spatial derivatives of pressure field and particle velocity can be obtained using fast fourier transform (FFT). Since the Fourier transforms used to represent the spatial derivatives of smooth functions are exact, only 2 points per wavelength are needed in the spatial discretization. The time integration is achieved by a 4th-order method to effectively reduce the computational time. The results of the algorithm are validated by analytical solutions. Perfectly Matched Layers (PMLs) are applied to absorb the outgoing waves and avoid ``wraparound'' effect. The maximum attenuation coefficient of the PMLs has an optimum value to minimize the reflections due to discretization and wraparound effect for 2D and 3D problems. Different PML profiles are also compared, quadratic profile is chosen because it can minimize the overall reflection. Spatial smoothing is needed for PSTD to avoid Gibbs' phenomenon in the modeling of a point source, and the effect of the smoothing function is studied.

In the TAT reconstruction problem, the PSTD method is used to reconstruct the thermoacoustic sources by solving the thermoacoustic wave equations in a reversed temporal order within the framework of time reversal imaging. The back-propagated pressure waves then refocus at the spatial locations of the original sources. Most other TAT reconstruction algorithms are based on the assumption that the tissue medium is acoustically homogeneous. In practice, however, even the mild tissue inhomogeneity will cause large phase errors and cause spatial misplacement and distortion of the sources. The proposed PSTD method utilizes a two-step process to solve this problem. In the first step, a homogeneous time reversal reconstruction is performed. Since an inhomogeneity itself is usually a source because of spatially dependent electrical conductivity (thus microwave absorption), the spatial location and the shape of the inhomogeneity can be estimated. In the second step, the updated acoustic property map is loaded followed by an inhomogeneous reconstruction. Numerical results show that this method greatly improves the reconstruction results. Images with improved quality are reconstructed from experimental data.

A 3D PSTD algorithm is developed and validated. Numerical results show that the PSTD algorithm with the 4th-order time integration is capable of simulating large 3D acoustic problems accurately and efficiently. A 3D breast phantom model is used to study the inhomogeneous reconstruction in 3D. Improved results over the homogeneous method are observed.

A preliminary study of the Thermoacoustic Tomography (TAT) using continuous-wave (CW) modulated microwaves is summarized. The theoretical background, system configuration, experiment setup, and measurement results are presented.

Hyperpolarized (HP) 129Xe MRI has emerged as a viable tool for evaluating lung function without ionizing radiation. HP 129Xe has already been used to image ventilation and quantify ventilation defects. However, this thesis aims to further develop imaging techniques that are capable of imaging, not just ventilation, but also gas transfer within the lung. This ability to image gas transfer directly is enabled by the solubility and chemical shifts of 129Xe that provide separate MR signatures in the airspaces, barrier tissue, and red blood cells (RBCs).

While 129Xe in the airspace (referred to as gas-phase 129Xe) can be readily imaged with standard vendor-provided imaging sequences, 129Xe in the barrier and RBC compartments (collectively referred to as dissolved-phase 129Xe) has such a rapid T2* (<2 msec at 2T) that even simple gradient recalled echo (GRE) sequences are ineffective at imaging the limited signal before it decays. To minimize these losses from T2* decay, the 3D radial sequence offers much shorter TEs that can image the dissolved-phase 129Xe. Despite their ability to image dissolved-phase signal, however, 3D radial sequences have not yet been widely adopted within the hyperpolarized gas community. In order to demonstrate the potential of the 3D radial pulse sequence, chapter 3 uses standard 129Xe ventilation imaging to compare 3D radial image quality and defect conspicuity with that of the conventional GRE. Since the 3D radial sequence offered comparable performance in ventilation imaging, and also provided the ability to image dissolved-phase 129Xe, chapter 3 establishes that the 3D radial sequence is well-suited for imaging 129Xe in humans.

Though 3D radial acquisition offers clear advantages for functional 129Xe lung imaging, its non-Cartesian sampling of k-space complicates image reconstruction. Chapter 4 carefully explains the process of gridding-based reconstruction, and describes how problems arising from non-selective RF pulses and undersampling, both of which are commonly employed in hyperpolarized 129Xe imaging, can be avoided by using appropriate reconstruction techniques. Furthermore, we detail a generalized procedure to optimize reconstruction parameters, then demonstrate the benefits of our improved reconstruction methods across both 1H anatomical imaging as well as functional imaging of 129Xe in the gas- and dissolved-phases.

These dissolved-phase images are particularly interesting because they consist of separate contributions from 129Xe in the RBCs and barrier tissue. Once these two resonances are disentangled from one another, they provide a noninvasive means to measure gas exchange regionally. However, such decomposition of these two resonances is predicated on prior knowledge of their spectroscopic properties. To that end, chapter 5 describes a non-linear spectroscopic curve fitting toolbox that we developed to more accurately characterize the 129Xe spectrum in vivo. Though previously, only two dissolved-phase resonances have ever been described within the lung, our fitting tools were able to identify a third dissolved-phase resonance in both healthy volunteers and healthy controls. Furthermore, we describe several spectroscopic features that differ statistically between our healthy volunteers and IPF subjects to demonstrate that this technique is sensitive to even subtle functional changes within the lung. These spectroscopic measurements provide the basis for imaging gas transfer.

Describing lung function regionally requires phase-sensitive imaging techniques that can decompose the dissolved-phase signal into images that represent the contribution from the RBC and barrier resonances. To date, only two implementations have been demonstrated, and both suffered from poor SNR and challenges in quantifying gas transfer. Chapter 6 adds quantitative processing techniques that improve phase sensitive imaging of 129Xe gas transfer. These methods 1) normalize both the RBC and barrier uptake images by gas-phase magnetization so that intensities can be compared across subjects, 2) compress the dynamic range of these functional images to enhance their perceived SNR, and 3) derive colormap thresholds from a healthy reference population to give intensities meaningful context.

To show the value of our quantitative gas transfer imaging, chapter 7 applies these techniques to a cohort of healthy volunteers and another of IPF patients. Since patients with IPF exhibit a progressive thickening and hardening of the pulmonary interstitium that severely restricts the transport of gases between the lungs and blood, they represent an ideal population to prove out our methods. This analysis identifies several patterns to the RBC and barrier distributions which could potentially represent different stages of disease. Furthermore, we demonstrate that our MRI-based findings correlate well with DLCO and FVC, and to a lesser extent with the structural cues seen in CT. This suggests that 129Xe imaging offers complimentary functional information that can’t be derived from CT, while also describing its spatial distribution unlike PFTs.

The work in this thesis has transitioned our HP 129Xe gas transfer studies from a proof of concept to an optimized and quantitative imaging protocol with robust processing pipelines. Using these MRI methods, we have shown that we can directly and quantitatively probe pulmonary ventilation and gas transfer within a single breath hold. In IPF, such noninvasive imaging methods are desperately needed to monitor the efficacy of these new treatments to ensure that the associated medical expense is justified with positive changes in outcomes. Finally, these new functional contrasts will be useful in studying other cardiopulmonary diseases such as asthma, chronic obstructive pulmonary disease, and pulmonary arterial hypertension.

X-ray micro-CT is widely used for small animal imaging in preclinical studies of cardiopulmonary disease, but further development is needed to improve spatial resolution, temporal resolution, and material contrast. This study presents a set of tools that achieve these improvements. These tools include the mathematical formulation and computational implementation of algorithms for calibration, image reconstruction, and image analysis with our custom micro-CT system. These tools are tested in simulations and in experiments with live animals. With these tools, it is possible to visualize the distribution of a contrast agent throughout the body of a mouse as it changes over time, and produce 5-dimensional images (3 spatial dimensions + time + energy) of the cardiac cycle.