Electrical & Computer Engineering Theses and Dissertations
Permanent URI for this collectionhttp://hdl.handle.net/1903/2765
Browse
3 results
Search Results
Item DEEP LEARNING APPLICATIONS IN BONE MINERAL DENSITY ESTIMATION, SPINE VERTEBRA DETECTION, AND LIVER TUMOR SEGMENTATION(2023) Wang, Fakai; Wu, Min; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)As the aging population and related health concerns emerge in more countries than ever, we face many challenges such as the availability, quality, and cost of medical resources. Thanks to the development of machine learning and computer vision in recent years, Deep Learning (DL) can help solve some medical problems. The diagnosis of various diseases (such as spine disorders, low bone mineral density, and liver cancer) relies on X-rays or Computed Tomography (CT). DL models could automatically analyze these radiography scans and help with the diagnosis. Different organs and diseases have distinct characteristics, requiring customized algorithms and models. In this dissertation, we investigate several Computer Aided-Diagnosis (CAD) tasks and present corresponding DL solutions. Deep Learning has multiple advantages. Firstly, DL models could uncover underlying health issues invisible to humans. One example is the opportunistic screening of Osteoporosis through chest X-ray. We develop DL models, utilizing chest film to predict bone mineral density, which helps prevent bone fractures. Humans could not tell anything about bone density in the chest film, but DL models could reliably make the prediction. The second advantage is accuracy and efficiency. Reading radiography is tedious, requiring years of expertise. This is particularly true when a radiologist needs to localize potential liver tumors by looking through tens of CT slices, spending several minutes. Deep learning models could localize and identify the tumors within seconds, greatly reducing human labor. Experiments show DL models can pick up small tumors, which are hardly seen by the naked eye. Attention should be paid to deep learning limitations. Firstly, DL models lack explainability. Deep learning models store diagnostic knowledge and statistical patterns in their parameters, which are obscure to humans. Secondly, uncertainty exists for rare diseases. If not exposed to rare cases, the models would yield uncertain outcomes. Thirdly, training AI models are subject to high-quality data but the labeling quality varies in clinical practice. Despite the challenges and issues, deep learning models are promising to promote medical diagnosis in society.Item Spectral Methods for Neural Network Designs(2022) Su, Jiahao; Huang, Furong; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Neural networks are general-purpose function approximators. Given a problem, engineers or scientists select a hypothesis space of functions with specific properties by designing the network architecture. However, mainstream designs are often ad-hoc, which could suffer from numerous undesired properties. Most prominently, the network architectures are gigantic, where most parameters are redundant while consuming computational resources. Furthermore, the learned networks are sensitive to adversarial perturbation and tend to underestimate the predictive uncertainty. We aim to understand and address these problems using spectral methods --- while these undesired properties are hard to interpret from network parameters in the original domain, we could establish their relationship when we represent the parameters in a spectral domain. These relationships allow us to design networks with certified properties via the spectral representation of parameters.Item DESIGN SPACE EXPLORATION FOR SIGNAL PROCESSING SYSTEMS USING LIGHTWEIGHT DATAFLOW GRAPHS(2018) Li, Lin; Bhattacharyya, Shuvra S; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Digital signal processing (DSP) is widely used in many types of devices, including mobile phones, tablets, personal computers, and numerous forms of embedded systems. Implementation of modern DSP applications is very challenging in part due to the complex design spaces that are involved. These design spaces involve many kinds of configurable parameters associated with the signal processing algorithms that are used, as well as different ways of mapping the algorithms onto the targeted platforms. In this thesis, we develop new algorithms, software tools and design methodologies to systematically explore the complex design spaces that are involved in design and implementation of signal processing systems. To improve the efficiency of design space exploration, we develop and apply compact system level models, which are carefully formulated to concisely capture key properties of signal processing algorithms, target platforms, and algorithm-platform interactions. Throughout the thesis, we develop design methodologies and tools for integrating new compact system level models and design space exploration methods with lightweight dataflow (LWDF) techniques for design and implementation of signal processing systems. LWDF is a previously-introduced approach for integrating new forms of design space exploration and system-level optimization into design processes for DSP systems. LWDF provides a compact set of retargetable application programming interfaces (APIs) that facilitates the integration of dataflow-based models and methods. Dataflow provides an important formal foundation for advanced DSP system design, and the flexible support for dataflow in LWDF facilitates experimentation with and application of novel design methods that are founded in dataflow concepts. Our developed methodologies apply LWDF programming to facilitate their application to different types of platforms and their efficient integration with platform-based tools for hardware/software implementation. Additionally, we introduce novel extensions to LWDF to improve its utility for digital hardware design and adaptive signal processing implementation. To address the aforementioned challenges of design space exploration and system optimization, we present a systematic multiobjective optimization framework for dataflow-based architectures. This framework builds on the methodology of multiobjective evolutionary algorithms and derives key system parameters subject to time-varying and multidimensional constraints on system performance. We demonstrate the framework by applying LWDF techniques to develop a dataflow-based architecture that can be dynamically reconfigured to realize strategic configurations in the underlying parameter space based on changing operational requirements. Secondly, we apply Markov decision processes (MDPs) for design space exploration in adaptive embedded signal processing systems. We propose a framework, known as the Hierarchical MDP framework for Compact System-level Modeling (HMCSM), which embraces MDPs to enable autonomous adaptation of embedded signal processing under multidimensional constraints and optimization objectives. The framework integrates automated, MDP-based generation of optimal reconfiguration policies, dataflow-based application modeling, and implementation of embedded control software that carries out the generated reconfiguration policies. Third, we present a new methodology for design and implementation of signal processing systems that are targeted to system-on-chip (SoC) platforms. The methodology is centered on the use of LWDF concepts and methods for applying principles of dataflow design at different layers of abstraction. The development processes integrated in our approach are software implementation, hardware implementation, hardware-software co-design, and optimized application mapping. The proposed methodology facilitates development and integration of signal processing hardware and software modules that involve heterogeneous programming languages and platforms. Through three case studies involving complex applications, we demonstrate the effectiveness of the proposed contributions for compact system level design and design space exploration: a digital predistortion (DPD) system, a reconfigurable channelizer for wireless communication, and a deep neural network (DNN) for vehicle classification.