教学课程资源库 更多>>

关键词： Medical imaging   Artificial intelligence   Computer science  

摘要： Contrast enhanced cardiac computed tomography angiography (CTA) is an important imaging modality for assessing the morphology of the heart and coronary arteries for diagnosing cardiovascular disease. Accurate heart segmentation in cardiac CTA images is an important task for quantitative assessment of aorta, myocardium, and different cardiac chambers. However, manual segmentation is time-consuming, prone to errors and variations between observers while automatic segmentation is very challenging due to large shape variations in the cardiac anatomy among different subjects, indistinct boundaries between substructures of the heart (e.g. right ventricle and right atrium) or between the heart and surrounding tissues (e.g. liver, ribs, sternum) and long processing time may not be efficient in clinical practice. This dissertation provides a new method that uses machine learning based approach to improve the accuracy and efficiency of the commonly used multi-atlas and deep neural network based approaches for 17 different cardiac and non-cardiac structures segmentation in cardiac CTA images. Specifically, an automatic framework combining multi-atlas and corrective segmentation to identify several cardiovascular and intrathoracic structures from cardiac CTA images is proposed. The proposed multi-atlas based framework significantly improves the segmentation accuracy without suffering from high computational burden existed in conventional multi-atlas based approach. In addition, large amounts of annotated data are required for deep neural networks to produce reliable results. This is a common limitation in deep learning-based approach, especially in medical field as the annotated data required efforts by domain experts and it can be very challenging and time consuming when multiple labels are desired in 3D volumetric data. To tackle this limitation, the proposed multi-atlas based framework can be used to generate reliable segmentations with less annotated samples to serve as tra

关键词： Statistics   Artificial intelligence   Computer science  

摘要： A deep generative classifier employs Short Run Markov Chain Monte Carlo inference with Langevin dynamics and backpropagation through time. In contrast to a convolutional neural network (ConvNet) with analogous architecture, the Short Run classifier approaches the same classification accuracy and (1) may synthesize data, (2) may learn unsupervised from additional unannotated data, and (3) exhibits robustness to adversarial attacks, due to the stochasticity of the Langevin equation and the top-down architecture of the generator network. The ConvNet classifier lacks the ability to perform (1) or (2) and possesses no defense against adversarial attacks, a critical concern for any deployed machine learning system. Meanwhile, the Short Run classifier demonstrates the capacity to improve in both classification accuracy and the quality of data synthesis, given additional unannotated data.

关键词： Computer engineering   Electrical engineering   Computer science  

摘要： Person re-identification (re-id) has broad appeal in practical applications such as video surveillance and criminal investigations. Given an image or image sequence of a person of interest from a “probe” camera view, person re-id aims to find the correct match from the candidates seen in a “gallery” camera view. Despite tremendous progress in re-id research, several problems still hinder the reliable, real-world use of person re-id. In this thesis, we study person re-id from different operational aspects, in order to design better real-world re-id systems. In typical re-id research, the probe image or image sequence is matched to a fixed gallery set, ignoring the arrival time of each person. In our first contribution, we designed and collected a new, large-scale, multi-camera re-id dataset called RPIfield. It preserves time-stamp information for every detected person in each video sequence, to better simulate real-world re-id operational scenarios. With this information about candidates' reappearances, re-id algorithms can be applied to provide instantaneous rank lists for probes that simulate a real-time re-id system. Based on this dataset, we propose a new evaluation methodology called the Rank Persistence Curve (RPC) for evaluating re-id algorithms in circumstances when the same person of interest can appear multiple times in the gallery, as well as when the performance over multiple persons of interest should be aggregated. We demonstrate the effectiveness of RPCs on RPIfield for allowing users to make informed choices about the expected performance of candidate re-id algorithms in real-world deployments. Next, we propose a novel deep Convolutional Neural Network (CNN) architecture for image-based person re-id, to deal with typical challenges such as viewpoint change, misalignment, and partially-cropped person images. We address these questions by means of a new attention-driven Siamese learning architecture, called the Consistent Attentive Siamese Network. Our

关键词： Electrical engineering   Medical imaging   Computer science  

摘要： Solving inverse problems remain an active research area in various fields to study the cause of a phenomenon by observing the effects. In particular, such efforts are well grounded in medical imaging applications where inverse problems naturally arise due to the imaging target being either inaccessible or invisible to human eyes. Non-regularity or ill-conditioning is a major challenge in such situations which is a direct consequence of limited observations/measurements being available. Medical imaging applications have classically incorporated domain-specific knowledge about the forward encoding operator to regularize the pertaining inverse problem. Deep learning has recently received high interest as an alternative approach for data-driven regularization. In principle, such regularization can be achieved by acquiring more measurements. However, sufficient measurements are usually unavailable due to practical considerations. For instance, in electromagnetic source imaging (ESI) using electroencephalography (EEG), number of recording electrodes can not be excessively increased due to subsequent distortions that electrodes contacts can cause in the medium. This limitation adversely affects the spatial resolution of EEG-based ESI. Another example is magnetic resonance imaging (MRI) where increasing imaging resolution trades off with noise properties, temporal resolution and scan duration. High-resolution MRI can become prohibitively long especially when a sequence of images need to be obtained. Thus, accelerated MRI techniques are required to reconstruct high-resolution images using less data. The present thesis contributes to solving regularization issues pertaining to inverse problems in medical imaging with a focus on high-resolution electromagnetic source imaging (ESI) using electroencephalography (EEG) and accelerated high resolution magnetic resonance imaging (MRI) with high fidelity. The ESI spatial resolution and the brain full coverage is shown to significantl

关键词： Computer science  

关键词： Computer science  

摘要： Human beings have always been interested in understanding themselves and their surroundings. Learning about the relationship between the two can reveal facts of the present and help predict the future, a critical part to live a better life. With the proliferation of IoT sensor devices, it is now possible to collect quality data for each individual and utilize this data for building personal models that can help to understand the self and environment. However, since this sensor data have different granularities and semantics, the semantic gap becomes even more formidable. Thus, there are challenges in aggregating, integrating, and synchronizing this heterogeneous data to a form such that it effectively describes the life experiences of each individual. In this dissertation, we design a personal chronicle, which contributes a solution to the aforementioned challenges, called Personicle, in which all kinds of personal data streams can be correlated with one another to form a model of a person. To implement the Personicle, we first attempt to bridge the semantic gap between the low- level multimedia logs and high-level semantics by developing a common daily event model through the data unobtrusively obtained from smart devices. To do this, we define an atomic interval, which brings together the scattered sources of heterogeneous data to partition the data into manageable pieces. This atomic interval lets us segment a day into sequences of similar patterns and use the segments for daily event recognition. Secondly, we design an event-triggered Ecological Momentary Assessment (EMA) to maximize the chance of aggregating the semantic data from the users. Unlike the traditional EMA process, which mainly depends on user initiative and intervention, we contribute to overcoming the problems endemic to persistent data collection, such as missing a logging moment or early abandonment, by initiating the EMA process from the system side at the right moment. Lastly, we propose a ful

关键词： Computer science  

摘要： For self-driving vehicles, aerial drones, and autonomous robots to be successfully deployed in the real-world, they must be able to navigate complex environments and track objects. While Artificial Intelligence and Machine Vision have made significant progress in dynamic scene understanding, they are not yet as robust and computationally efficient as humans or other primates in these tasks. For example, the current state-of-the-art visual tracking methods become inaccurate when applied to random test videos. We suggest that ideas from cortical visual processing can inspire real world solutions for motion perception and tracking that are robust and efficient. In this context, the following contributions are made in this dissertation. First, a method for estimating 6DoF ego-motion and pixel-wise object motion is introduced, based on a learned overcomplete motion field basis set. The method uses motion field constraints for training and a novel differentiable sparsity regularizer to achieve state-of-the-art ego and object-motion performances on benchmark datasets. Second, a Convolutional Neural Network (CNN) that learns hidden neural representations analogous to the response characteristics of dorsal Medial Superior Temporal area (MSTd) neurons for optic flow and object motion is presented. The findings suggest that goal driven training of CNNs might automatically result in the MSTd-like response properties of model neurons. Third, a recurrent neural network model of predictive smooth pursuit eye movements is presented that generates similar pursuit initiation and predictive pursuit behaviors as observed in humans. The model provides the computational mechanisms of formation and rapid update of an internal model of target velocity, commonly attributed to zero lag tracking and smooth pursuit of occluded objects. Finally, a spike based stereo depth algorithm is presented that reconstructs dynamic visual scenes at 400 frames-per-second with one watt of power consumption w

关键词： Information science   Artificial intelligence   Computer science  

摘要： Deep learning is an increasingly popular technology used for such tasks as image classification, speech recognition, and language translation. Deep learning technology is under active development, so that many innovative chipsets, useful frameworks, creative algorithms, and big data sets are emerging. Previous research using image data for measuring deep learning classifiers has usually focused on high quality image datasets. Thus, scientists usually have not tried to benchmark deep learning robustness with imperfect images. However, high quality images are not always available, and people also expect image classifier robustness in applications where image quality is not high. Therefore, in this research work, we plan to contribute to the currently still small body of work on benchmarking image quality enhancement, benchmarking deep learning robustness, and proposing two quantitative methods and a three-dimensional surface model for measuring robust deep learning classifiers, and how to increase deep learning robustness by the hybrid training. Two-factor corrupted image data will be tested as a tool for improving deep learning classifier robustness. In this dissertation, chapter 1 is an introduction. Chapter 2 is a literature review. Chapter 3 illustrates how to create the two-factor corruption method. In chapter 4, the three image quality metrics and sampling technologies are discussed. Then, in chapter 5, we provide how to benchmark the effects of image quality enhancement. In chapter 6, we suggest a statistical plot named mCV analysis, a quantitative index named normalized difference robustness index, and a 3D surface model for measuring robust deep learning classifiers. Moreover, the hybrid training for improving deep learning robustness is also discussed in chapter 6. Chapter 7 is conclusion and future work.

关键词： Computer science  

摘要： In recent years, a wide range of resource-constrained devices have been built and integrated into many networked systems. These devices collect and transfer data over the Internet in order for users to access the data or to control these devices remotely. However, the data also may contain sensitive information such as medical records or credit card numbers. This underscores the importance of protecting potentially sensitive data before it is transferred over the network. To provide security services such as data confidentiality and authentication, these devices must be provided with cryptographic keys to encrypt the data. Designing security schemes for resource-limited devices is a challenging task due to the inherit characteristics of these devices which are limited memory, processing power and battery life. In this dissertation, we propose lightweight polynomial-based cryptographic protocols in three environments that encompass resource-constrained devices which are Wireless Sensor Network (WSN), Fog Computing, and Blockchain Network. With polynomial-based schemes, we guarantee high network connectivity due to the existence of a shared pairwise key between every pair of nodes in the network. More importantly, the proposed schemes are lightweight which means they exhibit low memory, processing and communication overheads for resource-constrained devices compared with other schemes. The only problem with polynomial-based schemes is that they suffer from node-captured attacks. That is, when an attacker captured a specific number of nodes, the attacker could compromise the security of the whole network. In this dissertation, we propose, for the first time, polynomial-based schemes with probabilistic security in WSNs. That is, when the attacker captured a specific number of sensor nodes, there is a low probability the attacker could compromised the security of the whole network. We show how we can modify system’s parameters to lower such attacks.

关键词： African studies   Information technology   Public administration   Computer science  

摘要： Several technical challenges prevent the adoption of cloud computing by government organizations in Nigeria. Information technology (IT) leaders in the Nigerian government are concerned about this problem because the lack of cloud computing adoption may prevent the Nigerian government from taking advantage of cloud-based information systems to improve its service delivery to citizens and businesses. Grounded in the technology acceptance model, the purpose of this quantitative correlational study was to examine if IT administrators’ perception of data security and perception of fault tolerance can predict their intentions to adopt cloud computing. Data were collected from 79 IT administrators in government organizations in Nigeria. The results of the multiple regression were significant, F(2, 76) = 31.58, p < .001, R2 = 0.45, with IT administrators’ perception of data security (β = .72, p < .001) being the only significant predictor of IT administrators’ intention to adopt cloud computing; IT administrators’ perception of fault tolerance (β = .09, p = .37) was not a significant predictor of IT administrators’ intention to adopt cloud computing. The Nigerian government may use this study as a pedestal to measure cloud computing practice and maturity in all its organizations, improve existing cloud computing policies, and increase cloud computing training programs for its IT administrators. This study’s results might contribute to positive social change by helping the Nigerian government improve its service delivery to citizens and businesses through the increased adoption of cloud computing-based information systems.