检索结果-内蒙古大学图书馆

Exploit fully automatic low-level segmented PET data for training high-level deep learning algorithms for the corresponding CT data

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Gsaxner, Christina Roth, Peter M. Wallner, Jürgen Egger, Jan Institute for Computer Graphics and Vision Faculty of Computer Science and Biomedical Engineering Graz University of Technology Graz Austria Computer Algorithms for Medicine Laboratory Graz Austria Department of Oral & Maxillofacial Surgery Medical University of Graz Auenbruggerplatz Styria Austria

We present an approach for fully automatic urinary bladder segmentation in CT images with artificial neural networks in this study. Automatic medical image analysis has become an invaluable tool in the different treatment stages of diseases. Especially medical image segmentation plays a vital role, since segmentation is often the initial step in an image analysis pipeline. Since deep neural networks have made a large impact on the field of image processing in the past years, we use two different deep learning architectures to segment the urinary bladder. Both of these architectures are based on pre-trained classification networks that are adapted to perform semantic segmentation. Since deep neural networks require a large amount of training data, specifically images and corresponding ground truth labels, we furthermore propose a method to generate such a suitable training data set from Positron Emission Tomography/Computed Tomography image data. This is done by applying thresholding to the Positron Emission Tomography data for obtaining a ground truth and by utilizing data augmentation to enlarge the dataset. In this study, we discuss the influence of data augmentation on the segmentation results, and compare and evaluate the proposed architectures in terms of qualitative and quantitative segmentation performance. The results presented in this study allow concluding that deep neural networks can be considered a promising approach to segment the urinary bladder in CT images. Copyright © 2019, The Authors. All rights reserved.

关键词： Deep neural networks

A comparative study on affordable photogrammetry tools 125

学校读者我要写书评

暂无评论

A comparative study on affordable photogrammetry tools

125th ASEE Annual Conference and Exposition

作者： Fraley, John Imeri, Astrit Fidan, Ismail Chandramouli, Magesh Mechanical Engineering Department Tennessee Tech. University United States Department of Computer Graphics Technology Purdue University Northwest United States

The objective of the Project MANEUVER (Manufacturing Education Using Virtual Environment Resources)1 is to develop an affordable virtual reality (VR) framework to address the imminent demand for well-trained digital manufacturing (DM) professionals. One important part of Project MANEUVER involves studying, evaluating, and identifying cost-efficient ways to generate 3D solid models for use in VR frameworks. To this end, this paper explains the research effort to find alternative ways so that 3D solid model could easily be generated without using any costly 3D scanning technology. In this study, the project team identified two software tools that could help the manufacturing professionals and educators generate a solid model of several parts. These two software tools namely, Qlone and 3DF Zephyr Free were selected for this study based on factors such as ease-of-use, cost-effectiveness, and the cognitive load on users. Using casestudies these two software tools were used to generate 3D solid models and prototypes. Finally, their pros and cons collected throughout this study were reported. © American Society for Engineering Education, 2018.

关键词：

Revealing the Invisible: Visual Analytics and Explanatory Storytelling for Advanced Team Sport Analysis

学校读者我要写书评

暂无评论

Revealing the Invisible: Visual Analytics and Explanatory St...

2018 International Symposium on Big Data Visual and Immersive Analytics, BDVA 2018

作者： Stein, Manuel Breitkreutz, Thorsten Haussler, Johannes Seebacher, Daniel Niederberger, Christoph Schreck, Tobias Grossniklaus, Michael Keim, Daniel Janetzko, Halldor Department of Computer and Information Science University of Konstanz Germany VIZRT Switzerland Institute for Computer Graphics and Knowledge Visualization Graz University of Technology Austria Department of Geography University of Zurich Switzerland

ISBN: (纸本)9781538691946

The analysis of invasive team sports often concentrates on cooperative and competitive aspects of collective movement behavior. A main goal is the identification and explanation of strategies, and eventually the development of new strategies. In visual sports analytics, a range of different visual-interactive analysis techniques have been proposed, e.g., based on visualization using for example trajectories, graphs, heatmaps, and animations. Identifying suitable visualizations for a specific situation is key to a successful analysis. Existing systems enable the interactive selection of different visualization facets to support the analysis process. However, an interactive selection of appropriate visualizations is a difficult, complex, and time-consuming task. In this paper, we propose a four-step analytics conceptual workflow for an automatic selection of appropriate views for key situations in soccer games. Our concept covers classification, specification, explanation, and alteration of match situations, effectively enabling the analysts to focus on important game situations and the determination of alternative moves. Combining abstract visualizations with real world video recordings by Immersive Visual Analytics and descriptive storylines, we support domain experts in understanding key situations. We demonstrate the usefulness of our proposed conceptual workflow via two proofs of concept and evaluate our system by comparing our results to manual video annotations by domain experts. Initial expert feedback shows that our proposed concept improves the understanding of competitive sports and leads to a more efficient data analysis. © 2018 IEEE.

关键词： Visualization

Deep learning - A first meta-survey of selected reviews across scientific disciplines, their commonalities, challenges and research impact

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Egger, Jan Pepe, Antonio Gsaxner, Christina Jin, Yuan Li, Jianning Kern, Roman Institute of Computer Graphics and Vision Faculty of Computer Science and Biomedical Engineering Graz University of Technology Graz Austria Computer Algorithms for Medicine Laboratory Graz Austria Department of Oral and Maxillofacial Surgery Medical University of Graz Graz Austria University Medicine Essen Essen Germany Research Center for Connected Healthcare Big Data Zhejiang Lab Zhejiang Hangzhou China Research Unit Experimental Neurotraumatology Department of Neurosurgery Medical University of Graz Graz Austria Knowledge Discovery Know-Center Graz Austria Institute of Interactive Systems and Data Science Graz University of Technology Graz Austria

Deep learning belongs to the field of artificial intelligence, where machines perform tasks that typically require some kind of human intelligence. Deep learning tries to achieve this by drawing inspiration from the learning of a human brain. Similar to the basic structure of a brain, which consists of (billions of) neurons and connections between them, a deep learning algorithm consists of an artificial neural network, which resembles the biological brain structure. Mimicking the learning process of humans with their senses, deep learning networks are fed with (sensory) data, like texts, images, videos or sounds. These networks outperform the state-of-the-art methods in different tasks and, because of this, the whole field saw an exponential growth during the last years. This growth resulted in way over 10,000 publications per year in the last years. For example, the search engine PubMed alone, which covers only a sub-set of all publications in the medical field, provides already over 11,000 results in Q3 2020 for the search term 'deep learning', and around 90% of these results are from the last three years. Consequently, a complete overview over the field of deep learning is already impossible to obtain and, in the near future, it will potentially become difficult to obtain an overview over a subfield. However, there are several review articles about deep learning, which are focused on specific scientific fields or applications, for example deep learning advances in computer vision or in specific tasks like object detection. With these surveys as a foundation, the aim of this contribution is to provide a first high-level, categorized meta-survey of selected reviews on deep learning across different scientific disciplines and outline the research impact that they already have during a short period of time. The categories (computer vision, language processing, medical informatics and additional works) have been chosen according to the underlying data sources (image,

关键词： Surveys

Single trial ERP amplitudes reveal the time course of acquiring representations of novel faces in individual participants

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Sommer, Werner Stapor, Katarzyna Kończak, Grzegorz Kotowski, Krzysztof Fabian, Piotr Ochab, Jeremi Bereś, Anna Ślusarczyk, Grażyna Humboldt-University at Berlin Department of Psychology Germany Silesian University of Technology Institute of Computer Science Gliwice Poland University of Economics in Katowice Department of Statistics Econometrics and Mathematics Poland Silesian University of Technology Department of Applied Informatics Gliwice Poland Silesian University of Technology Department of Algorithmics and Software Gliwice Poland Jagiellonian University Department of Theory of Complex Systems Krakow Poland Jagiellonian University Department of Cognitive Neuroscience and Neuroergonomics Krakow Poland Jagiellonian University Department of Design and Computer Graphics Krakow Poland Institut für Psychologie Humboldt-Universität zu Berlin Rudower Chaussee 18 Berlin12489 Germany Faculty of Automatic Control Electronics and Computer Science Department of Applied Informatics Silesian University of Technology Akademicka 16 Gliwice44-100 Poland

The neural correlates of face individuation – the acquisition of memory representations for novel faces – have been studied only in coarse detail and disregarding individual differences between learners. In their seminal study, (Tanaka, Curran, Porterfield, & Collins, 2006) required the identification of a particular novel face across 70 trials and found that the N250 component in the ERP became more negative from the first to the second half of the experiment, where it reached a similar amplitude as a well-known face. We were unable to directly replicate this finding in our study when we used the original spilt of trials. However, when we applied a different split of trials we observed very similar changes in N250 amplitude. Then, we developed and applied a new two-step explorative-confirmative non-parametric method based on permutation testing to determine the time course of face individuation in individual participants based on single trial N250 amplitudes. We show that the assumption of a steep initial increase of N250 amplitude across multiple presentations of the target face, followed by a plateau, yields plausible results in fitting linear trends for most participants. The transition point from initial acquisition to the plateau phase differed strongly between participants and tended to be earlier when performance in target face recognition was better. Hence, face individuation may be accounted for by a biphasic process of early, fast acquisition, followed by a slower, asymptotic consolidation or maintenance phase. The current approach might be fruitfully applied to further investigations into face individuation and their neural correlates. © 2020, CC BY-NC-ND.

关键词： Face recognition

Virtual reality education modules for digital manufacturing instruction 125

学校读者我要写书评

暂无评论

Virtual reality education modules for digital manufacturing ...

125th ASEE Annual Conference and Exposition

作者： Chandramouli, Magesh Jin, Ge Heffron, Justin David Fidan, Ismail Cossette, Mel Welsch, Cheryl A. Merrell, Wayne Department of Computer Graphics Technology Purdue University Northwest United States Purdue University Northwest United States Department of Computer Information Technology and Graphics Purdue University Calumet United States Department of Manufacturing and Engineering Technology Tennessee Technological University United States Edmonds Community College LynnwoodWA United States Eastern Iowa Community College District United States

There is an imminent need to remedy the 'skills gaps' in the digital manufacturing (DM) sector as evident from the Bureau of Labor Statistics projections pointing to a decline in traditional manufacturing jobs accompanied by marked growth in digital- and computer-driven manufacturing jobs. With proven advantages such as cost benefits, material conservation, minimized labor, and enhanced precision, manufacturing industries worldwide are adapting to digital manufacturing standards on a large scale. In an effort to remedy the lack of well-defined DM career pathways and instructional framework, our NSF ATE (Advanced Technological Education) project MANEUVER (Manufacturing Education Using Virtual Environment Resources) is developing an innovative pedagogical approach using virtual reality (VR). This multimodal VR framework DM instruction targeted at 2-year and 4-year manufacturing programs, facilitates the development of VR modules for multiple modes such as desktop VR, Augmented VR, and Immersive VR. The advantages of the virtual reality framework for digital manufacturing education include: significant cost reduction, reduction in equipment and maintenance costs, ability to pre-visualize the product before manufacturing. This paper introduces the design and development process of VR education tool to simulate three different additive manufacturing machines, e.g., LutzBot™, FormLabs™, and UPrint™ and different 3D printing technologies e.g., fused deposition modeling, and selective laser sintering to allow the students experience the materials and equipment needed to create the same part using different types of equipment and different types of technology. © American Society for Engineering Education, 2018.

关键词： Augmented VR Desktop VR Digital manufacturing Virtual reality

PET-Train: Automatic Ground Truth Generation from PET Acquisitions for Urinary Bladder Segmentation in CT Images using Deep Learning 11

学校读者我要写书评

暂无评论

PET-Train: Automatic Ground Truth Generation from PET Acquis...

11th Biomedical Engineering International Conference, BMEiCON 2018

作者： Gsaxner, Christina Pfarrkirchner, Birgit Lindner, Lydia Pepe, Antonio Roth, Peter M. Egger, Jan Wallner, Jurgen Inst. of Computer Graphics and Vision Graz University of Technology Graz Austria Department of Maxillofacial Surgery Medical University of Graz Graz Austria Computer Algorithms for Medicine Laboratory Graz Austria

ISBN: (纸本)9781538657249

In this contribution, we propose an automatic ground truth generation approach that utilizes Positron Emission Tomography (PET) acquisitions to train neural networks for automatic urinary bladder segmentation in Computed Tomography (CT) images. We evaluated different deep learning architectures to segment the urinary bladder. However, deep neural networks require a large amount of training data, which is currently the main bottleneck in the medical field, because ground truth labels have to be created by medical experts on a time-consuming slice-by-slice basis. To overcome this problem, we generate the training data set from the PET data of combined PET/CT acquisitions. This can be achieved by applying simple thresholding to the PET data, where the radiotracer accumulates very distinct in the urinary bladder. However, the ultimate goal is to entirely skip PET imaging and its additional radiation exposure in the future, and only use CT images for segmentation. © 2018 IEEE.

关键词： Positron emission tomography

Medical deep learning-A systematic meta-review

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Egger, Jan Gsaxner, Christina Pepe, Antonio Pomykala, Kelsey L. Jonske, Frederic Kurz, Manuel Li, Jianning Kleesiek, Jens Institute of Computer Graphics and Vision Faculty of Computer Science and Biomedical Engineering Graz University of Technology Inffeldgasse 16 Styria Graz8010 Austria Department of Oral & Maxillofacial Surgery Medical University of Graz Auenbruggerplatz 5/1 Styria Graz8036 Austria Computer Algorithms for Medicine Laboratory Styria Graz Austria University Medicine Essen Girardetstraße 2 Essen45131 Germany University Medicine Essen Hufelandstraße 55 Essen45147 Germany Partner Site Essen Hufelandstraße 55 Essen45147 Germany

Deep learning has remarkably impacted several different scientific disciplines over the last few years. For example, in image processing and analysis, deep learning algorithms were able to outperform other cutting-edge methods. Additionally, deep learning has delivered state-of-the-art results in tasks like autonomous driving, outclassing previous attempts. There are even instances where deep learning outperformed humans, for example with object recognition and gaming. Deep learning is also showing vast potential in the medical domain. With the collection of large quantities of patient records and data, and a trend towards personalized treatments, there is a great need for automated and reliable processing and analysis of health information. Patient data is not only collected in clinical centers, like hospitals and private practices, but also by mobile healthcare apps or online websites. The abundance of collected patient data and the recent growth in the deep learning field has resulted in a large increase in research effort s. In Q2/2020, the search engine PubMed returned already over 11,000 results for the search term 'deep learning', and around 90% of these publications are from the last three years. However, even though PubMed represents the largest search engine in the medical field, it does not cover all medical-related publications. Hence, a complete overview of the field of 'medical deep learning' is almost impossible to obtain and acquiring a full overview of medical sub-fields is becoming increasingly more difficult. Nevertheless, several review and survey articles about medical deep learning have been published within the last few years. They focus, in general, on specific medical scenarios, like the analysis of medical images containing specific pathologies. With these surveys as a foundation, the aim of this article is to provide the first high-level, systematic meta-review of medical deep learning surveys. Copyright © 2020, The Authors. All rights res

关键词： Generative adversarial networks

SELF-SUPERVISED LEARNING FOR STEREO RECONSTRUCTION ON AERIAL IMAGES

学校读者我要写书评

暂无评论

SELF-SUPERVISED LEARNING FOR STEREO RECONSTRUCTION ON AERIAL...

IEEE International Geoscience and Remote Sensing Symposium

作者： Patrick Knobelreiter Christoph Vogel Thomas Pock Institute for Computer Graphics and Vision Graz University of Technology

Recent developments established deep learning as an inevitable tool to boost the performance of dense matching and stereo estimation. On the downside, learning these networks requires a substantial amount of training data to be successful. Consequently, the application of these models outside of the laboratory is far from straight forward. In this work we propose a self-supervised training procedure that allows us to adapt our network to the specific (imaging) characteristics of the dataset at hand, without the requirement of external ground truth data. We instead generate interim training data by running our intermediate network on the whole dataset, followed by conservative outlier filtering. Bootstrapped from a pre-trained version of our hybrid CNN-CRF model, we alternate the generation of training data and network training. With this simple concept we are able to lift the completeness and accuracy of the pre-trained version significantly. We also show that our final model compares favorably to other popular stereo estimation algorithms on an aerial dataset.

关键词： Large scale 3D Dense matching CNN