Accurate image reconstruction is fundamental to medical imaging diagnostics, involving the transformation of data from the sensor domain to the image domain by solving an inverse problem. In Magnetic Resonance Imaging (MRI), measurements are acquired in the k-space spatial frequency domain, and the inverse Fourier Transform is applied to reconstruct the image for diagnosis. However, exact solutions to the inverse problem using analytical models are often not possible. Partial measurements are often acquired to decrease scanning time, resulting in ill-posed inverse problems that necessitate a series of signal processing steps for optimal reconstructions. Supervised deep learning approaches have been explored for solving such inverse problems, including image reconstruction. While deep learning can tackle these challenges in a single reconstruction step, training deployable models can be challenging due to encountering unseen data distributions that deviate from the training data in real-world scenarios. In this dissertation, we first investigate the impact of complex input data design, data augmentations, adversarial noise, and hallucinations on reconstruction accuracy and robustness of deep learning-based image reconstruction methods. We illustrate how the complex input data design and architectural modifications can notably enhance performance accuracy. We showcase the emergence of artifacts when training lacks proper data augmentations such as multiple field-of-views in the dataset. Additionally, we study the effectiveness of deep learning when exposed to Gaussian versus engineered adversarial noise, proposing a technique to adapt the numerical properties of the training dataset for resilience against adversarial noise. Finally, we investigate the occurrence of hallucinations on undersampled out-of-distribution (OOD) data reconstructions and propose a method for quantifying and mitigating them through domain adaptation techniques. Due to encountering OOD data in real-world settings, it is essential to assess whether a given input falls within the training data distribution, in-distribution (ID), particularly when reconstructing medical images for diagnostic purposes. We propose a single model variance method based on the local Lipschitz metric to distinguish OOD images from ID. Our method achieves an impressive area under the curve of 99.94% for True Positive Rate versus False Positive Rate. Empirically, we demonstrate a very strong relationship between the local Lipschitz value and mean absolute error (MAE), supported by a high Spearman's rank correlation coefficient of 0.8475. Through selective prediction, we demonstrate a method to determine the local Lipschitz threshold for uncertainty as it relates to optimal model performance. Our study was validated using the AUTOMAP architecture for sensor-to-image domain MRI reconstruction. We compare our proposed approach with baseline methods of Monte-Carlo dropout and deep ensembles as well as the state-of-the-art Mean Variance Estimation (MVE) network approach. Furthermore, we showcase the versatility of our approach to other architectures and learned functions, including the UNET architecture for MRI denoising and Computed Tomography (CT) sparse-to-full view reconstruction applications. Lastly, we expand the field of deep learning to solve inverse problems to Nitrogen-vacancy (NV) center diamond magnetometry, a quantum sensing technique that measures the magnetic field produced by circuits using the NV center optical defect. We designed a MAGnetic Inverse Calculation UNET (MAGIC-UNET) to reconstruct current density images using magnetic fields as input by solving the inverse Biot-Savart law and compared it to the analytical Fourier Method. We find that the deep learning solution using the MAGIC-UNET has greater accuracy on simulated and NV-diamond magnetometry experimental data compared to the analytical Fourier Method. It also significantly reduces the magnetometer collection time due to requiring fewer signal averages. These results expand the application scope of NV-diamond magnetometry to weak current sources and the use of DL to solve inverse problems to the quantum sensor domain.

Estágio realizado na Flupol / Tese de mestrado integrado. Engenharia Electrotécnica e de Computadores. Faculdade de Engenharia. Universidade do Porto. 2008

No description available.

Thesis (M.S.)--Virginia Polytechnic Institute and State University, 1992. / Vita. Abstract. Includes bibliographical references (leaves 93-101). Also available via the Internet.

Thesis (M. App. Sc.)--Carleton University, 2003. / Includes bibliographical references (p. 101-102). Also available in electronic format on the Internet.

Thesis (Ph. D.)--University of Texas at Austin, 2004. / Supervisor: E. Glenn Lightsey. Vita. Includes bibliographical references.

Thesis (M.S. in Electrical Engineering) Naval Postgraduate School, December 1996. / "December 1996." Thesis advisor(s): Tri T. Ha and Vicente Garcia. Includes bibliographical references (p. 95-97). Also available online.

Thesis (M.S.)--University of Missouri--Rolla, 2007. / Vita. The entire thesis text is included in file. Title from title screen of thesis/dissertation PDF file (viewed May 16, 2007) Includes bibliographical references (p. 96-97).

Orientador: Heloisa Vieira da Rocha / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Computação / Made available in DSpace on 2018-07-22T21:22:10Z (GMT). No. of bitstreams: 1 Brenner_MaurenFernandaMeira_M.pdf: 3007263 bytes, checksum: c27f5a627a42ec00450c675feab2f59b (MD5) Previous issue date: 1996 / Resumo: A Sociedade da Mente é o nome dado a um conjunto de propostas de estruturas e mecanismos subjacentes à mente e à inteligência. Estas propostas foram desenvolvidas por Marvin Minsky e apresentadas principalmente durante a década de 80. Entre elas está a noção de que a mente é constituída por unidades desprovidas de inteligência chamadas agentes, unidades estas que podem ser identificadas em diversos níveis de abstração, sendo o mais "baixo" deles o nível equivalente ao neural. A noção de agente é usada para definir os conceitos e mecanismos relativos ao funcionamento integrado da mente como um todo, ou seja, relativos à coordenação dos agentes. O conceito central é o de estado mental parcial, que corresponde ao estado de um subconjunto dos agentes da mente. Estados mentais parciais constituem o meio de comunicação entre grupos de agentes, que na Sociedade da Mente são chamados agências, e são controlados através de agentes especiais que implementam determinados mecanismos de coordenação. Neste trabalho, a Sociedade da Mente é usada como fundamento para uma arquitetura para agentes inteligentes. Uma vez que na Sociedade da Mente a mente é composta por agentes, segue-se que o próprio agente também deve ser constituído de agentes menores, coordenados mediante os mecanismos propostos na Sociedade da Mente. Assim, projetamos uma arquitetura para agentes inteligentes baseada na Sociedade da Mente como uma arquitetura de um sistema multi-agente (um sistema de Inteligência Artificial Distribuída), onde a comunicação e coordenação entre os agentes são feita de uma forma inspirada no conceito de estados mentais parciais e nos mecanismos de coordenação da Sociedade da Mente. Descrevemos o processo de desenvolvimento no qual a arquitetura foi transformada desde um mapeamento direto entre agentes e agências da Sociedade da Mente e agentes da arquitetura multi-agentes, até a sua versão final baseada no modelo de blackboard, o qual constitui um paradigma bem conhecido em Inteligência Artificial Distribuída. Na arquitetura resultante, os mecanismos de coordenação da Sociedade da Mente servem como modelo para objetos organizados nas diferentes seções do blackboard. Esses objetos são "ativos", no sentido de que não constituem apenas informações a serem manipuladas, mas possuem também funcionalidades específicas, segundo as quais atuam como os mecanismos de coordenação da Sociedade da Mente. Mencionamos algumas possibilidades de aplicações usando a arquitetura desenvolvida e descrevemos em detalhe a implementação de uma dessas aplicações, que consiste em um agente que conta uma história. Esta implementação constituiu em uma experiência com o intuito de testar não somente a viabilidade da arquitetura proposta, mas também a própria metodologia deste trabalho, que foi a de tomar o modelo teórico da Sociedade da Mente como ponto de partida em lugar de projetar uma arquitetura para realizar alguma tarefa específica. / Abstract: The Society of Mind is a collection of proposals concerning structures and mechanisms underlying mind and intelligence. These proposals were developed by Marvin Minsky and presented mostly during the 80s. Among them, there is the notion that the mind is made up of mindless units called agents. Such units can be identified at many levels, the lowest of them being the equivalent to the neural level. The Society of Mind relies upon the notion of agent to define the concepts and mechanisms concerning the working of the mind as a whole, i.e. concerning the coordination of agents. The central concept is the partial mental state, which corresponds to the description of the states of some agents of the mind. Groups of agents, which are called agencies in the Society of Mind, communicate through partial mental states, which are controlled by special agents that implement certain coordination mechanisms. In this work, the Society of Mind is taken as a foundation for an intelligent agent architecture. Since in the Society of Mind the mind itself comprises many agents, it follows that the intelligent agent should be made up from smaller agents, coordinated through the mechanisms of the Society of Mind. Thus, we have designed an intelligent agent architecture based on the Society of Mind as an architecture of a multi-agent system (a Distributed Artificial Intelligence system), where the communication and coordination between agents is done in a way which is inspired on the concept of partial mental states and on the coordination mechanisms of the Society of Mind. We have described the design process through which the architecture was transformed from a direct mapping between agents and agencies of the Society of Mind to its final version based on the blackboard model, which is a well-known Distributed Artificial intelligence paradigm. The coordination mechanisms of the Society of Mind work as models for objects placed in the different sections of the blackboard. These objects are "active" in the sense that they do not only contain information to be operated upon, but also have specific functionalities according to which they act like the coordination mechanisms of the Society of Mind. We have mentioned some possibilities of applications using the architecture we developed, and described in some detail the implementation of one of those applications: a story-telling agent. This implementation. Was an experience whose objective was to test the feasibility of the proposed architecture as well as the methodology of this work itself which consisted of taking the theoretical model of the Society of Mind as a starting point instead of designing an architecture to perform some particular task. / Mestrado / Mestre em Ciência da Computação

No description available.