Melanoma, a severe and life-threatening skin cancer, is commonly misdiagnosed or left undiagnosed. Advances in artificial intelligence, particularly deep learning, have enabled the design and implementation of intelligent solutions to skin lesion detection and classification from visible light images, which are capable of performing early and accurate diagnosis of melanoma and other types of skin diseases. This work presents solutions to the problems of skin lesion segmentation and classification. The proposed classification approach leverages convolutional neural networks and transfer learning. Additionally, the impact of segmentation (i.e., isolating the lesion from the rest of the image) on the performance of the classifier is investigated, leading to the conclusion that there is an optimal region between “dermatologist segmented” and “not segmented” that produces best results, suggesting that the context around a lesion is helpful as the model is trained and built. Generative adversarial networks, in the context of extending limited datasets by creating synthetic samples of skin lesions, are also explored. The robustness and security of skin lesion classifiers using convolutional neural networks are examined and stress-tested by implementing adversarial examples. / Includes bibliography. / Thesis (M.S.)--Florida Atlantic University, 2018. / FAU Electronic Theses and Dissertations Collection

The death of one�s parents, irrespective of the age at which it occurs, is generally regarded as a life experience of considerable significance. The last few years of an elderly person�s life are often characterized by increasing frailty, declining health and loss of independence. Responsibility for the spiritual and physical care of parents during that period is undertaken by many adult children. Current research in this area is generally informed by the requirements of social policy, which, by identifying and addressing the inherent difficulties in this so-called informal caring, is designed to support carers in the community. The research reported in this thesis represents a departure from this mode of inquiry and seeks, rather, to explore the existential aspects of caring in this particular situation, from the carer�s perspective. To achieve this objective, an existential phenomenological approach informed principally by the philosophies of Heidegger and Merleau-Ponty, and the adaptation of these philosophies by Schutz, Giorgi and van Manen to social science research, was developed to suit the particular requirements of the topic. In addition to the author�s autobiographic material, primary sources include conversations with five people who had been principal carers for their parents during their final illnesses. In all cases caring had ended with the parent�s death at least one year before the conversations took place. The principal secondary sources are Simone de Beauvoir�s memoir, A Very Easy Death, and Philip Roth�s account of his father�s illness and death, Patrimony: A True Story. In addition, the argument is supported throughout by reference to other literary works. From these sources a number of major existential themes, including temporality, hope, suffering, and knowing the body, have been explored in depth, in conjunction with relevant existential theories. Synthesis of these topics suggests that in this particular circumstance, for the people involved in the study, the phenomenon of caring can be understood as an unconditional engagement with the life and concerns of their parent at the end of life, and can be interpreted within an existential framework as representing an authentic way of Being.

In this paper we examine the reconstruction of functions of any dimension from hyperplanar projections. This is a generalization of a problem that has generated much interest recently, especially in the field of medical imaging. Computed Axial Tomography (CAT) and Nuclear Magnetic Resonance (NMR) are two medical techniques that fall in this framework. CAT scans measure the hydrogen density along planes through the body. Here we will examine reconstruction methods that involve backprojecting the projection data and summing this over the entire region of interest. There are two methods for doing this. One method is to filter the projection data first, and then backproject this filtered data and sum over all projection directions. The other method is to backproject and sum the projection data first, and then filter. The two methods are mathematically equivalent, producing very similar equations. We will derive the reconstruction formulas for both methods for any number of dimensions. We will examine the cases of two and three dimensions, since these are the only ones encountered in practice. The equations are very different for these cases. In general, the equations are very different for even and odd dimensionality. We will discuss why this is so, and show that the equations for even and odd dimensionality are related by the Hilbert Transform.

This research is primarily focused on laser-based non-invasive photoacoustic tomography of small animal brains. Photoacoustic tomography, a novel imaging modality, was applied to visualize the distribution of optical absorptions in small-animal brains through the skin and skull. This technique combines the high-contrast advantage of optical imaging with the high-resolution advantage of ultrasonic imaging. Based on the intrinsic optical contrast, this imaging system successfully visualized three-dimensional tissue structures in intact brains, including lesions and tumors in brain cerebral cortex. Physiological changes and functional activities in brains, including cerebral blood volume and blood oxygenation in addition to anatomical information, were also satisfactorily monitored. This technique successfully imaged the dynamic distributions of exogenous contrast agents in small-animal brains. Photoacoustic angiography in small-animal brains yielding high contrast and high spatial resolution was implemented noninvasively using intravenously injected absorbing dyes. In the appendix, the theory of Monte Carlo simulation of polarized light propagation in scattering media was briefly summarized.

Heart diseases are the leading cause of death in developed countries, hence, understanding of the hemodynamics of the heart is critically important to enable methods for cardiovascular diagnosis. Assessment of intracardiac blood pressure is highly interesting as blood flow is driven by the pressure differences. Phase-contrast magnetic resonance imaging (PC-MRI) is a tool for measuring blood flow and has a wide range of cardiovascular applications. Based on previous studies, an approach to estimate the relative pressure fields in the human heart from three-dimensional time-resolved PC-MRI velocity data was implemented and evaluated. The relative pressure fields were obtained by solving the pressure Poisson equation, using a multi-grid approach. The method was evaluated on a numerical phantom and on PC-MRI data from one healthy subject and one patient with dilated cardiomyopathy. The pressure field was visualized in combination with blood flow data and morphological images. Results indicate that the used approach works well for cardiac relative pressure estimation and are in agreement with findings from previous research. The complete spatial and temporal coverage of relative pressure enables a higher understanding of physiology and pathophysiology of the human heart and is expected to give new insights for clinical investigations.

The percutaneous revascularization of chronic total occlusions represents a major challenge to interventional cardiologists. Procedural success is currently limited by the inadequate soft-tissue contrast of x-ray fluoroscopy and the inability to visualize the position and orientation of a revascularization device with respect to the lesion and the vessel wall. In this thesis methods are developed that enable the percutaneous revascularization of occlusive lesions to be guided using magnetic resonance (MR) imaging. Unlike x-rays, MR has excellent soft-tissue contrast and this can be exploited to provide valuable information regarding the composition and geometry of the lesion. The first method is a robust and redundant technique for determining the position and orientation of a catheter inside an MR scanner. The technique uses phase information introduced into the MR signal by a small receive coil located at the distal tip of the catheter. The technique is developed theoretically and is demonstrated with a feasibility experiment. A forward-looking intravascular imaging catheter is then presented that is capable of acquiring of high-resolution MR images of occlusive lesions and the vessel wall in front of the catheter. The imaging catheter consists of two orthogonal receive coils located at the distal tip of the catheter. The use of the imaging catheter is demonstrated in phantoms and in vivo. A third method enables active visualization of MR compatible guidewires. The method utilizes a catheter-based pick-up coil that is magnetically coupled to the guidewire. The proposed technique enables one to concentrate all active components on a catheter thereby facilitating the use of safety features. Complete characterization is presented theoretically and validated experimentally. In addition, the use of a practical catheter device is demonstrated in an in situ environment. Finally, future work required for the development of an integrated catheter-based device for the MR-guided revascularization of chronic total occlusions is discussed.

The percutaneous revascularization of chronic total occlusions represents a major challenge to interventional cardiologists. Procedural success is currently limited by the inadequate soft-tissue contrast of x-ray fluoroscopy and the inability to visualize the position and orientation of a revascularization device with respect to the lesion and the vessel wall. In this thesis methods are developed that enable the percutaneous revascularization of occlusive lesions to be guided using magnetic resonance (MR) imaging. Unlike x-rays, MR has excellent soft-tissue contrast and this can be exploited to provide valuable information regarding the composition and geometry of the lesion. The first method is a robust and redundant technique for determining the position and orientation of a catheter inside an MR scanner. The technique uses phase information introduced into the MR signal by a small receive coil located at the distal tip of the catheter. The technique is developed theoretically and is demonstrated with a feasibility experiment. A forward-looking intravascular imaging catheter is then presented that is capable of acquiring of high-resolution MR images of occlusive lesions and the vessel wall in front of the catheter. The imaging catheter consists of two orthogonal receive coils located at the distal tip of the catheter. The use of the imaging catheter is demonstrated in phantoms and in vivo. A third method enables active visualization of MR compatible guidewires. The method utilizes a catheter-based pick-up coil that is magnetically coupled to the guidewire. The proposed technique enables one to concentrate all active components on a catheter thereby facilitating the use of safety features. Complete characterization is presented theoretically and validated experimentally. In addition, the use of a practical catheter device is demonstrated in an in situ environment. Finally, future work required for the development of an integrated catheter-based device for the MR-guided revascularization of chronic total occlusions is discussed.

A Positive Electron Tomography/Computed Tomography devise was installed in theX-ray section at US Linköping in May 2007. Positive Electron Tomography examinations with 18F-fluoro-deoxy-glucose are mainly used for tumor examinations. During 2007 occurred approximately 200 examinations and in 2008 600 are planned. Today there are two reconstruction methods commercially available, Filtered Back projection and Maximum Likelihood Expectation Maximiza tion, used in the faster version called Ordered Subsets Expectation Maximization. The image quality in Positive Electron Tomography depends on the choice of reconstruction method and the settings of its parameters. We have performed a physical phantom study with Positive Electron Tomography to determine optimal parameters for the iterative reconstruction algorithm Ordered Subsets Expectation Maximization. To find out whether or not the quality of the image can be improved, so that the patient received radiation dose and/or examination time can be lowered. The phantom used was a NEMA IEC Body PhantomTM, designed to mimic smallhot lesions typical in 18F, Fluorine-18 PET, and all calculations were done according to the NEMA NU2-2001 protocol. The main conclusion from this project is that a higher level of contrast can be reached, compared to the one clinically obtained today. Using more iterations then recommended from the manufacturer. / I maj 2007 installerades en Positiv Elektron Tomografi /DatoriseradTomografi-kamera, PET/CT, i Röntgenavdelningen Linköping US. PET med 18F-fluoro-deoxy-glucose används huvudsakligen för tumörundersökningar.2007 genomfördes ca 200 undersökningar, och för 2008 är ytterligare 600 planerade. Idag finns två olika bildrekonstruktionsmetoder kliniskt tillgängliga; Filtered Back projection och Maximum Likelihood Expectation Maximization, där den  vidareutvecklade versionen kallad Ordered Subsets Expectation Maximizationanvänds. Bildkvalitén från en Positive Electron Tomographykamera påverkas av valet av rekonstruktionsmetod och dess ingående parametrar. I detta projekt har en fantomstudie genomförts med syfte att bestämma de optimala parametrarna för den iterativa metoden Ordered Subsets Expectation Maximization. För att utreda huruvida stråldosen och/eller undersöknings tiden kan minskas. Det testfantom som användes var en NEMA IEC Body PhantomTM. Projektet följde metoden angiven i NEMA NU2-2001 protokollet. Det resultaten visar är att de rekommenderade inställningarna från tillverkaren inte är de optimala.

A Positive Electron Tomography/Computed Tomography devise was installed in the X-ray section at US Linköping in May 2007. Positive Electron Tomography examinations with 18F-fluoro-deoxyglucose are mainly used for tumor examinations. During 2007 occurred approximately 200 examinations and in 2008 600 are planned.Today there are two reconstruction methods commercially available, Filtered Back projection and Maximum Likelihood Expectation Maximiza tion, used in the faster version called Ordered Subsets Expectation Maximization. The image quality in Positive Electron Tomography depends on the choice of reconstruction method and the settings of its parameters. We have performed a physical phantom study with Positive Electron Tomography to determine optimal parameters for the iterativereconstruction algorithm Ordered Subsets Expectation Maximization. To find out whether or not the quality of the image can be improved, so that the patient received radiation dose and/or examination time can be lowered. The phantom used was a NEMA IEC Body PhantomTM, designed to mimic small hot lesions typicalin 18F, Fluorine-18 PET, and all calculations were done according to the NEMA NU2-2001 protocol. The main conclusion from this project is that a higher level of contrastcan be reached, compared to the one clinically obtained today. Using more iterations then recommended from the manufacturer.

<p>In the United States, heart failure is a leading cause of hospitalization. The medical industry places great emphasis on diagnosing heart disease through cardiac metrics like ejection fraction. Left ventricular ejection fraction is a commonly used diagnostic indicator for heart efficiency and is measured with echocardiography through different volume calculation techniques. However, ejection fraction results can drastically vary from one examiner to another. Generally cardiologists still give ejection fraction measurements a plus or minus 10 percent error range. </p><p>A program developed at Duke called 4DViz is robust enough for users to process 3D ultrasound data. 4DViz allows examiners to determine heart chamber volumes by constructing a surface model over an imaged heart chamber with many mouse click inputs. Through 4DViz programming, a viable approach for calculating ejection fraction is attempted in this thesis. Using feature tracking, surface drawing, and voxel filling, the new approach aims to reduce examiner input and improve ejection fraction consistency. Water filled balloons were used to calibrate the algorithm's parameters. In testing, several volunteers were asked to use the 4DViz. Their results are compared to volume measurements where user input was standard. The results show promise and may remove some of the inconsistency behind ejection fraction measurements.</p> / Thesis