Search results

Title: DEEP LEARNING FOR IMAGE PROCESSING WITH APPLICATIONS TO MEDICAL IMAGING
Creator: Zarshenas, Amin
Date: 2019
Description: Deep Learning is a subfield of machine learning concerned with algorithms that learn hierarchical data representations. Deep learning has...
Show moreDeep Learning is a subfield of machine learning concerned with algorithms that learn hierarchical data representations. Deep learning has proven extremely successful in many computer vision tasks including object detection and recognition. In this thesis, we aim to develop and design deep-learning models to better perform image processing and tackle three important problems: natural image denoising, computed tomography (CT) dose reduction, and bone suppression in chest radiography (“chest x-ray”: CXR). As the first contribution of this thesis, we aimed to answer to probably the most critical design questions, under the task of natural image denoising. To this end, we defined a class of deep learning models, called neural network convolution (NNC). We investigated several design modules for designing NNC for image processing. Based on our analysis, we design a deep residual NNC (R-NNC) for this task. One of the important challenges in image denoising regards to a scenario in which the images have varying noise levels. Our analysis showed that training a single R-NNC on images at multiple noise levels results in a network that cannot handle very high noise levels; and sometimes, it blurs the high-frequency information on less noisy areas. To address this problem, we designed and developed two new deep-learning structures, namely, noise-specific NNC (NS-NNC) and a DeepFloat model, for the task of image denoising at varying noise levels. Our models achieved the highest denoising performance comparing to the state-of-the-art techniques.As the second contribution of the thesis, we aimed to tackle the task of CT dose reduction by means of our NNC. Studies have shown that high dose of CT scans can increase the risk of radiation-induced cancer in patients dramatically; therefore, it is very important to reduce the radiation dose as much as possible. For this problem, we introduced a mixture of anatomy-specific (AS) NNC experts. The basic idea is to train multiple NNC models for different anatomic segments with different characteristics, and merge the predictions based on the segmentations. Our phantom and clinical analysis showed that more than 90% dose reduction would be achieved using our AS NNC model.We exploited our findings from image denoising and CT dose reduction, to tackle the challenging task of bone suppression in CXRs. Most lung nodules that are missed by radiologists as well as by computer-aided detection systems overlap with bones in CXRs. Our purpose was to develop an imaging system to virtually separate ribs and clavicles from lung nodules and soft-tissue in CXRs. To achieve this, we developed a mixture of anatomy-specific, orientation-frequency-specific (ASOFS) expert deep NNC model. While our model was able to decompose the CXRs, to achieve an even higher bone suppression performance, we employed our deep R-NNC for the bone suppression application. Our model was able to create bone and soft-tissue images from single CXRs, without requiring specialized equipment or increasing the radiation dose.
Show less

Title: DEEP LEARNING IN ENGINEERING MECHANICS: WAVE PROPAGATION AND DYNAMICS IMPLEMENTATIONS
Creator: Finol Berrueta, David
Date: 2019
Description: With the advent of Artiﬁcial Intelligence research in the 1960s, the need for intelligent systems that are able to truly comprehend the...
Show moreWith the advent of Artiﬁcial Intelligence research in the 1960s, the need for intelligent systems that are able to truly comprehend the physical world around them became relevant. Signiﬁcant milestones in the realm of machine learning and, in particular, deep learning during the past decade have led to advanced data-driven models that are able to approximate complex functions from pure observations. When it comes to the application of physics-based scenarios, the vast majority of these models rely on statistical and optimization constructs, leaving minimal room in their development for the physics-driven frameworks that more traditional engineering and science ﬁelds have been developing for centuries. On the other hand, the more traditional engineering ﬁelds, such as mechanics, have evolved on a diﬀerent set of modeling tools that are mostly based on physics driven assumptions and equations, typically aided by statistical tools for uncertainty handling. Deep learning models can provide signiﬁcant implementation advantages in commercial systems over traditional engineering modeling tools in the current economies of scale, but they tend to lack the strong reliability their counterparts naturally allow. The work presented in this thesis is aimed at assessing the potential of deep learning tools, such as Convolutional Neural Networks and Long Short-Term Memory Networks, as data-driven models in engineering mechanics, with a major focus on vibration problems. In particular, two implementation cases are presented: a data driven surrogate model to a Phononic eigenvalue problem, and a physics-learning model in rigid-body dynamics scenario. Through the applications presented, this work that shows select deep learning architectures can appropriately approximate complex functions found in engineering mechanics from a system’s time history or state and generalize to set expectations outside training domains. In spatio-temporal systems, it is also that shown local learning windows along space and time can provide improved model reliability in their approximation and generalization performance
Show less

Title: DAMAGE ASSESSMENT OF CIVIL STRUCTURES AFTER NATURAL DISASTERS USING DEEP LEARNING AND SATELLITE IMAGERY
Creator: Jones, Scott F
Date: 2019
Description: Since 1980, millions of people have been harmed by natural disasters that have cost communities across the world over three trillion dollars....
Show moreSince 1980, millions of people have been harmed by natural disasters that have cost communities across the world over three trillion dollars. After a natural disaster has occurred, the creation of maps that identify the damage to buildings and infrastructure is imperative. Currently, many organizations perform this task manually, using pre- and post-disaster images and well-trained professionals to determine the degree and extent of damage. This manual task can take days to complete. I propose to do this task automatically using post-disaster satellite imagery. I use a pre-trained neural network, SegNet, and replaced its last layer with a simple damage classification scheme. This final layer of the network is re-trained using cropped segments of the satellite image of the disaster. The data were obtained from a publicly accessible source, the Copernicus EMS system. They provided three channel (RGB) reference and damage grading maps that were used to create the images of the ground truth and the damaged terrain. I then retrained the final layer of the network to identify civil structures that had been damaged. The resulting network was 85% accurate at labelling the pixels in an image of the disaster from typhoon Haiyan. The test results show that it is possible to create these maps quickly and efficiently.
Show less

Title: IMPACT OF DATA SHAPE, FIDELITY, AND INTER-OBSERVER REPRODUCIBILITY ON CARDIAC MAGNETIC RESONANCE IMAGE PIPELINES
Creator: Obioma, Blessing Ngozi
Date: 2020
Description: Artificial Intelligence (AI) holds a great promise in the healthcare. It provides a variety of advantages with its application in clinical...
Show moreArtificial Intelligence (AI) holds a great promise in the healthcare. It provides a variety of advantages with its application in clinical diagnosis, disease prediction, and treatment, with such interests intensifying in the medical image field. AI can automate various cumbersome data processing techniques in medical imaging such as segmentation of left ventricular chambers and image-based classification of diseases. However, full clinical implementation and adaptation of emerging AI-based tools face challenges due to the inherently opaque nature of such AI algorithms based on Deep Neural Networks (DNN), for which computer-trained bias is not only difficult to detect by physician users but is also difficult to safely design in software development. In this work, we examine AI application in Cardiac Magnetic Resonance (CMR) using an automated image classification task, and thereby propose an AI quality control framework design that differentially evaluates the black-box DNN via carefully prepared input data with shape and fidelity variations to probe system responses to these variations. Two variants of the Visual Geometric Graphics with 19 neural layers (VGG19) was used for classification, with a total of 60,000 CMR images. Findings from this work provides insights on the importance of quality training data preparation and demonstrates the importance of data shape variability. It also provides gateway for computation performance optimization in training and validation time.
Show less

Title: AUTOMATION OF ULTRASONIC FLAW DETECTION APPLICATIONS USING DEEP LEARNING ALGORITHMS
Creator: Virupakshappa, Kushal
Date: 2021
Description: The Industrial Revolution-4.0 promises to integrate multiple technologies including but not limited to automation, cloud computing, robotics,...
Show moreThe Industrial Revolution-4.0 promises to integrate multiple technologies including but not limited to automation, cloud computing, robotics, and Artificial Intelligence. The non-Destructive Testing (NDT) industry has been shifting towards automation as well. For ultrasound-based NDT, these technological advancements facilitate smart systems hosting complex signal processing algorithms. Therefore, this thesis introduces the effective use of AI algorithms in challenging NDT scenarios. The first objective is to investigate and evaluate the performance of both supervised and unsupervised machine learning algorithms and optimize them for ultrasonic flaw detection utilizing Amplitude-scan (A-scan) data. Several inferences and optimization algorithms have been evaluated. It has been observed that proper choice of features for specific inference algorithms leads to accurate flaw detection. The second objective of this study is the hardware realization of the ultrasonic flaw detection algorithms on embedded systems. Support Vector Machine algorithm has been implemented on a Tegra K1 GPU platform and Supervised Machine Learning algorithms have been implemented on a Zynq FPGA for a comparative study. The third main objective is to introduce new deep learning architectures for more complex flaw detection applications including classification of flaw types and robust detection of multiple flaws in B-scan data. The proposed Deep Learning pipeline combines a novel grid-based localization architecture with meta-learning. This provides a generalized flaw detection solution wherein additional flaw types can be used for inference without retraining or changing the deep learning architecture. Results show that the proposed algorithm performs well in more complex scenarios with high clutter noise and the results are comparable with traditional CNN and achieve the goal of generality and robustness.
Show less

Title: IMPROVING DEEP LEARNING BASED SEMANTIC SEGMENTATION USING CONTEXT INFORMATION
Creator: Xia, Zhengyu
Date: 2021
Description: Semantic segmentation is an important but challenging task in computer vision because it aims to assign each pixel a category label accurately...
Show moreSemantic segmentation is an important but challenging task in computer vision because it aims to assign each pixel a category label accurately. Nowadays, applications such as autonomous driving, path navigation, image search engine, or augmented reality require accurate semantic analysis and efficient segmentation mechanisms. In this thesis, we propose multiple models to improve the performance of semantic segmentation. In the first part, we focus on the single-task network, which aims to improve the performance of semantic segmentation. Our research includes exploiting context information using mixed spatial pyramid pooling to extract dense context-embedded features in FCN-based semantic segmentation. We also propose a GAF module to generate a global context-based attention map to guide the shallow-layer feature maps for better pixel localization. In the second part, we focus on a multi-task network that incorporates semantic segmentation to improve other computer vision tasks such as object detection. Specifically, a multi-task network, along with a learning strategy is designed to let semantic segmentation and object detection assist each other since they are highly correlated. Also, we include weakly-supervised multi-label semantic segmentation learning to deal with the shortage of high-quality training examples and to improve the performance of cross-domain object detection. In the third part, we focus on improving the performance of video panoptic segmentation, which is a unified network that incorporates semantic segmentation and instance segmentation using video streams. We design a new ConvLSTM pyramid to transmit spatio-temporal contextual information in our video panoptic segmentation network. Specifically, we propose a modified ConvLSTM to generate temporal contextual information. Also, we design an MSTPP module to obtain mixed spatio-temporal context-embedded feature maps. Experimental results on different datasets show that our proposed method achieves better performance compared with the state-of-the-art methods.
Show less

Title: AI IN MEDICINE: ENABLING INTELLIGENT IMAGING, PROGNOSIS, AND MINIMALLY INVASIVE SURGERY
Creator: Getty, Neil
Date: 2022
Description: While an extremely rich research field, compared to other applications of AI such as natural language processing (NLP) and image processing...
Show moreWhile an extremely rich research field, compared to other applications of AI such as natural language processing (NLP) and image processing/generation, AI in medicine has been much slower to be applied in real-world clinical settings. Often the stakes of failure are more dire, the access of private and proprietary data more costly, and the burden of proof required by expert clinicians is much higher. Beyond these barriers, the often typical data-driven approach towards validation is interrupted by a need for expertise to analyze results. Whereas the results of a trained Imagenet or machine translation model are easily verified by a computational researcher, analysis in medicine can be much more multi-disciplinary demanding. AI in medicine is motivated by a great demand for progress in health-care, but an even greater responsibility for high accuracy, model transparency, and expert validation.This thesis develops machine and deep learning techniques for medical image enhancement, patient outcome prognosis, and minimally invasive robotic surgery awareness and augmentation. Each of the works presented were undertaken in di- rect collaboration with medical domain experts, and the efforts could not have been completed without them. Pursuing medical image enhancement we worked with radiologists, neuroscientists and a neurosurgeon. In patient outcome prognosis we worked with clinical neuropsychologists and a cardiovascular surgeon. For robotic surgery we worked with surgical residents and a surgeon expert in minimally invasive surgery. Each of these collaborations guided priorities for problem and model design, analysis, and long-term objectives that ground this thesis as a concerted effort towards clinically actionable medical AI. The contributions of this thesis focus on three specific medical domains. (1) Deep learning for medical brain scans: developed processing pipelines and deep learn- ing models for image annotation, registration, segmentation and diagnosis in both traumatic brain injury (TBI) and brain tumor cohorts. A major focus of these works is on the efficacy of low-data methods, and techniques for validation of results without any ground truth annotations. (2) Outcome prognosis for TBI and risk prediction for Cardiovascular Disease (CVD): we developed feature extraction pipelines and models for TBI and CVD patient clinical outcome prognosis and risk assessment. We design risk prediction models for CVD patients using traditional Cox modeling, machine learning, and deep learning techniques. In this works we conduct exhaustive data and model ablation study, with a focus on feature saliency analysis, model transparency, and usage of multi-modal data. (3) AI for enhanced and automated robotic surgery: we developed computer vision and deep learning techniques for understanding and augmenting minimally invasive robotic surgery scenes. We’ve developed models to recognize surgical actions from vision and kinematic data. Beyond model and techniques, we also curated novel datasets and prediction benchmarks from simulated and real endoscopic surgeries. We show the potential for self-supervised techniques in surgery, as well as multi-input and multi-task models.
Show less

Title: RADIAL MAP ASSESSMENT APPROACH FOR DEEP LEARNING DENOISED CARDIAC MAGNETIC RESONANCE RECONSTRUCTION SHARPNESS
Creator: Mo, Fei
Date: 2021
Description: Deep Learning (DL) and Artificial Intelligence (AI) play important roles in the computer-aided medical diagnostics and precision medicine...
Show moreDeep Learning (DL) and Artificial Intelligence (AI) play important roles in the computer-aided medical diagnostics and precision medicine fields, capable of complementing human operators in disease diagnosis and treatment but optimizing and streamlining medical image display. While incredibly powerful, images produced via Deep Learning or Artificial Intelligence should be analyzed critically in order to be cognizant of how the algorithms are producing the new image and what the new imagine is. One such opportunity arose in the form of a unique collaborative project: the technical development of an image assessment tool that would analyze outputs between DL-based and non DL-based Magnetic Resonance Imaging reconstruction methods.More specifically, we examine the operator input dependence of the existing reference method in terms of accuracy and precision performance, and subsequently propose a new metric approach that preserves the heuristics of the intended quantification, overcomes operator dependence, and provides a relative comparative scoring approach that may normalize for angular dependence of examined images. In chapter 2 of this thesis, we provide a background description pertaining to the two imaging science principles that yielded our proposed method description and study design. First, if treated naively, the examined linear measurement approach exhibits potential bias with respect to the coordinate lattice space of the examined image. Second, the examined DL-based image reconstruction methods used in this thesis warrants an elaborate and explicit description of the measured noise and signal present in the reconstructed images. This specific reconstruction approach employs an iterative scheme with an embedded DL-based substep or filter to which we are blinded. In chapters 3 and 4 of this thesis, the imaging and DL-based image reconstruction experiments are described. These experiments employ cardiac MRI datasets from multiple clinical centers. We first outline the clinical and technical background for this approach, and then examine the quality of DL-based reconstructed image sharpness by two alternative methods: 1) by employing the gold-standard method that addresses the lattice point irregularity using a ‘re-gridding’ method, and 2) by applying our novel proposed method inspired by radial MRI k-space sampling, which exploits the mathematical properties of uniform radial sampling to yield the target voxel counts in the ‘gridded’ polar coordinate system. This new measure of voxel counts is shown to overcome the limitation due to the operator-dependence for the conventional approach. Furthermore, we propose this metric as a relative and comparative index between two alternative reconstruction methods from the same MRI k-space.
Show less

Title: IMPACT OF DATA SHAPE, FIDELITY, AND INTER-OBSERVER REPRODUCIBILITY ON CARDIAC MAGNETIC RESONANCE IMAGE PIPELINES
Creator: Obioma, Blessing Ngozi
Date: 2020
Description: Artificial Intelligence (AI) holds a great promise in the healthcare. It provides a variety of advantages with its application in clinical...
Show moreArtificial Intelligence (AI) holds a great promise in the healthcare. It provides a variety of advantages with its application in clinical diagnosis, disease prediction, and treatment, with such interests intensifying in the medical image field. AI can automate various cumbersome data processing techniques in medical imaging such as segmentation of left ventricular chambers and image-based classification of diseases. However, full clinical implementation and adaptation of emerging AI-based tools face challenges due to the inherently opaque nature of such AI algorithms based on Deep Neural Networks (DNN), for which computer-trained bias is not only difficult to detect by physician users but is also difficult to safely design in software development. In this work, we examine AI application in Cardiac Magnetic Resonance (CMR) using an automated image classification task, and thereby propose an AI quality control framework design that differentially evaluates the black-box DNN via carefully prepared input data with shape and fidelity variations to probe system responses to these variations. Two variants of the Visual Geometric Graphics with 19 neural layers (VGG19) was used for classification, with a total of 60,000 CMR images. Findings from this work provides insights on the importance of quality training data preparation and demonstrates the importance of data shape variability. It also provides gateway for computation performance optimization in training and validation time.
Show less

repository.iit

Search the repository

Enabled Filters

Refine Results

Type

Date

Subject

Creator

Rights