TY - JOUR A1 - Rezaei, Mina A1 - Yang, Haojin A1 - Meinel, Christoph T1 - Recurrent generative adversarial network for learning imbalanced medical image semantic segmentation JF - Multimedia tools and applications : an international journal N2 - We propose a new recurrent generative adversarial architecture named RNN-GAN to mitigate imbalance data problem in medical image semantic segmentation where the number of pixels belongs to the desired object are significantly lower than those belonging to the background. A model trained with imbalanced data tends to bias towards healthy data which is not desired in clinical applications and predicted outputs by these networks have high precision and low recall. To mitigate imbalanced training data impact, we train RNN-GAN with proposed complementary segmentation mask, in addition, ordinary segmentation masks. The RNN-GAN consists of two components: a generator and a discriminator. The generator is trained on the sequence of medical images to learn corresponding segmentation label map plus proposed complementary label both at a pixel level, while the discriminator is trained to distinguish a segmentation image coming from the ground truth or from the generator network. Both generator and discriminator substituted with bidirectional LSTM units to enhance temporal consistency and get inter and intra-slice representation of the features. We show evidence that the proposed framework is applicable to different types of medical images of varied sizes. In our experiments on ACDC-2017, HVSMR-2016, and LiTS-2017 benchmarks we find consistently improved results, demonstrating the efficacy of our approach. KW - Imbalanced medical image semantic segmentation KW - Recurrent generative KW - adversarial network Y1 - 2019 U6 - https://doi.org/10.1007/s11042-019-7305-1 SN - 1380-7501 SN - 1573-7721 VL - 79 IS - 21-22 SP - 15329 EP - 15348 PB - Springer CY - Dordrecht ER - TY - JOUR A1 - Rezaei, Mina A1 - Näppi, Janne J. A1 - Lippert, Christoph A1 - Meinel, Christoph A1 - Yoshida, Hiroyuki T1 - Generative multi-adversarial network for striking the right balance in abdominal image segmentation JF - International journal of computer assisted radiology and surgery N2 - Purpose: The identification of abnormalities that are relatively rare within otherwise normal anatomy is a major challenge for deep learning in the semantic segmentation of medical images. The small number of samples of the minority classes in the training data makes the learning of optimal classification challenging, while the more frequently occurring samples of the majority class hamper the generalization of the classification boundary between infrequently occurring target objects and classes. In this paper, we developed a novel generative multi-adversarial network, called Ensemble-GAN, for mitigating this class imbalance problem in the semantic segmentation of abdominal images. Method: The Ensemble-GAN framework is composed of a single-generator and a multi-discriminator variant for handling the class imbalance problem to provide a better generalization than existing approaches. The ensemble model aggregates the estimates of multiple models by training from different initializations and losses from various subsets of the training data. The single generator network analyzes the input image as a condition to predict a corresponding semantic segmentation image by use of feedback from the ensemble of discriminator networks. To evaluate the framework, we trained our framework on two public datasets, with different imbalance ratios and imaging modalities: the Chaos 2019 and the LiTS 2017. Result: In terms of the F1 score, the accuracies of the semantic segmentation of healthy spleen, liver, and left and right kidneys were 0.93, 0.96, 0.90 and 0.94, respectively. The overall F1 scores for simultaneous segmentation of the lesions and liver were 0.83 and 0.94, respectively. Conclusion: The proposed Ensemble-GAN framework demonstrated outstanding performance in the semantic segmentation of medical images in comparison with other approaches on popular abdominal imaging benchmarks. The Ensemble-GAN has the potential to segment abdominal images more accurately than human experts. KW - imbalanced learning KW - generative multi-discriminative networks KW - semantic KW - segmentation KW - abdominal imaging Y1 - 2020 U6 - https://doi.org/10.1007/s11548-020-02254-4 SN - 1861-6410 SN - 1861-6429 VL - 15 IS - 11 SP - 1847 EP - 1858 PB - Springer CY - Berlin ER - TY - THES A1 - Rezaei, Mina T1 - Deep representation learning from imbalanced medical imaging N2 - Medical imaging plays an important role in disease diagnosis, treatment planning, and clinical monitoring. One of the major challenges in medical image analysis is imbalanced training data, in which the class of interest is much rarer than the other classes. Canonical machine learning algorithms suppose that the number of samples from different classes in the training dataset is roughly similar or balance. Training a machine learning model on an imbalanced dataset can introduce unique challenges to the learning problem. A model learned from imbalanced training data is biased towards the high-frequency samples. The predicted results of such networks have low sensitivity and high precision. In medical applications, the cost of misclassification of the minority class could be more than the cost of misclassification of the majority class. For example, the risk of not detecting a tumor could be much higher than referring to a healthy subject to a doctor. The current Ph.D. thesis introduces several deep learning-based approaches for handling class imbalanced problems for learning multi-task such as disease classification and semantic segmentation. At the data-level, the objective is to balance the data distribution through re-sampling the data space: we propose novel approaches to correct internal bias towards fewer frequency samples. These approaches include patient-wise batch sampling, complimentary labels, supervised and unsupervised minority oversampling using generative adversarial networks for all. On the other hand, at algorithm-level, we modify the learning algorithm to alleviate the bias towards majority classes. In this regard, we propose different generative adversarial networks for cost-sensitive learning, ensemble learning, and mutual learning to deal with highly imbalanced imaging data. We show evidence that the proposed approaches are applicable to different types of medical images of varied sizes on different applications of routine clinical tasks, such as disease classification and semantic segmentation. Our various implemented algorithms have shown outstanding results on different medical imaging challenges. N2 - Medizinische Bildanalyse spielt eine wichtige Rolle bei der Diagnose von Krankheiten, der Behandlungsplanung, und der klinischen Überwachung. Eines der großen Probleme in der medizinischen Bildanalyse ist das Vorhandensein von nicht ausbalancierten Trainingsdaten, bei denen die Anzahl der Datenpunkte der Zielklasse in der Unterzahl ist. Die Aussagen eines Modells, welches auf einem unbalancierten Datensatz trainiert wurde, tendieren dazu Datenpunkte in die Klasse mit der Mehrzahl an Trainingsdaten einzuordnen. Die Aussagen eines solchen Modells haben eine geringe Sensitivität aber hohe Genauigkeit. Im medizinischen Anwendungsbereich kann die Einordnung eines Datenpunktes in eine falsche Klasse Schwerwiegende Ergebnisse mit sich bringen. In die Nichterkennung eines Tumors Beispielsweise brigt ein viel höheres Risiko für einen Patienten, als wenn ein gesunder Patient zum Artz geschickt wird. Das Problem des Lernens unter Nutzung von nicht ausbalancierten Trainingsdaten wird erst seit Kurzem bei der Klassifizierung von Krankheiten, der Entdeckung von Tumoren und beider Segmentierung von Tumoren untersucht. In der Literatur wird hier zwischen zwei verschiedenen Ansätzen unterschieden: datenbasierte und algorithmische Ansätze. Die vorliegende Arbeit behandelt das Lernen unter Nutzung von unbalancierten medizinischen Bilddatensätzen mittels datenbasierter und algorithmischer Ansätze. Bei den datenbasierten Ansätzen ist es unser Ziel, die Datenverteilung durch gezieltes Nutzen der vorliegenden Datenbasis auszubalancieren. Dazu schlagen wir neuartige Ansätze vor, um eine ausgeglichene Einordnung der Daten aus seltenen Klassen vornehmen zu können. Diese Ansätze sind unter anderem synthesize minority class sampling, patient-wise batch normalization, und die Erstellung von komplementären Labels unter Nutzung von generative adversarial networks. Auf der Seite der algorithmischen Ansätze verändern wir den Trainingsalgorithmus, um die Tendenz in Richtung der Klasse mit der Mehrzahl an Trainingsdaten zu verringern. Dafür schlagen wir verschiedene Algorithmen im Bereich des kostenintensiven Lernens, Ensemble-Lernens und des gemeinsamen Lernens vor, um mit stark unbalancierten Trainingsdaten umgehen zu können. Wir zeigen, dass unsere vorgeschlagenen Ansätze für verschiedenste Typen von medizinischen Bildern, mit variierender Größe, auf verschiedene Anwendungen im klinischen Alltag, z. B. Krankheitsklassifizierung, oder semantische Segmentierung, anwendbar sind. Weiterhin haben unsere Algorithmen hervorragende Ergebnisse bei unterschiedlichen Wettbewerben zur medizinischen Bildanalyse gezeigt. KW - machine learning KW - deep learning KW - computer vision KW - imbalanced learning KW - medical image analysis KW - Maschinenlernen KW - tiefes Lernen KW - unbalancierter Datensatz KW - Computervision KW - medizinische Bildanalyse Y1 - 2019 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:kobv:517-opus4-442759 ER - TY - BOOK A1 - Adriano, Christian A1 - Bleifuß, Tobias A1 - Cheng, Lung-Pan A1 - Diba, Kiarash A1 - Fricke, Andreas A1 - Grapentin, Andreas A1 - Jiang, Lan A1 - Kovacs, Robert A1 - Krejca, Martin Stefan A1 - Mandal, Sankalita A1 - Marwecki, Sebastian A1 - Matthies, Christoph A1 - Mattis, Toni A1 - Niephaus, Fabio A1 - Pirl, Lukas A1 - Quinzan, Francesco A1 - Ramson, Stefan A1 - Rezaei, Mina A1 - Risch, Julian A1 - Rothenberger, Ralf A1 - Roumen, Thijs A1 - Stojanovic, Vladeta A1 - Wolf, Johannes ED - Meinel, Christoph ED - Plattner, Hasso ED - Döllner, Jürgen Roland Friedrich ED - Weske, Mathias ED - Polze, Andreas ED - Hirschfeld, Robert ED - Naumann, Felix ED - Giese, Holger ED - Baudisch, Patrick ED - Friedrich, Tobias ED - Böttinger, Erwin ED - Lippert, Christoph T1 - Technical report BT - Fall Retreat 2018 N2 - Design and Implementation of service-oriented architectures imposes a huge number of research questions from the fields of software engineering, system analysis and modeling, adaptability, and application integration. Component orientation and web services are two approaches for design and realization of complex web-based system. Both approaches allow for dynamic application adaptation as well as integration of enterprise application. Commonly used technologies, such as J2EE and .NET, form de facto standards for the realization of complex distributed systems. Evolution of component systems has lead to web services and service-based architectures. This has been manifested in a multitude of industry standards and initiatives such as XML, WSDL UDDI, SOAP, etc. All these achievements lead to a new and promising paradigm in IT systems engineering which proposes to design complex software solutions as collaboration of contractually defined software services. Service-Oriented Systems Engineering represents a symbiosis of best practices in object-orientation, component-based development, distributed computing, and business process management. It provides integration of business and IT concerns. The annual Ph.D. Retreat of the Research School provides each member the opportunity to present his/her current state of their research and to give an outline of a prospective Ph.D. thesis. Due to the interdisciplinary structure of the research school, this technical report covers a wide range of topics. These include but are not limited to: Human Computer Interaction and Computer Vision as Service; Service-oriented Geovisualization Systems; Algorithm Engineering for Service-oriented Systems; Modeling and Verification of Self-adaptive Service-oriented Systems; Tools and Methods for Software Engineering in Service-oriented Systems; Security Engineering of Service-based IT Systems; Service-oriented Information Systems; Evolutionary Transition of Enterprise Applications to Service Orientation; Operating System Abstractions for Service-oriented Computing; and Services Specification, Composition, and Enactment. N2 - Der Entwurf und die Realisierung dienstbasierender Architekturen wirft eine Vielzahl von Forschungsfragestellungen aus den Gebieten der Softwaretechnik, der Systemmodellierung und -analyse, sowie der Adaptierbarkeit und Integration von Applikationen auf. Komponentenorientierung und WebServices sind zwei Ansätze für den effizienten Entwurf und die Realisierung komplexer Web-basierender Systeme. Sie ermöglichen die Reaktion auf wechselnde Anforderungen ebenso, wie die Integration großer komplexer Softwaresysteme. Heute übliche Technologien, wie J2EE und .NET, sind de facto Standards für die Entwicklung großer verteilter Systeme. Die Evolution solcher Komponentensysteme führt über WebServices zu dienstbasierenden Architekturen. Dies manifestiert sich in einer Vielzahl von Industriestandards und Initiativen wie XML, WSDL, UDDI, SOAP. All diese Schritte führen letztlich zu einem neuen, vielversprechenden Paradigma für IT Systeme, nach dem komplexe Softwarelösungen durch die Integration vertraglich vereinbarter Software-Dienste aufgebaut werden sollen. "Service-Oriented Systems Engineering" repräsentiert die Symbiose bewährter Praktiken aus den Gebieten der Objektorientierung, der Komponentenprogrammierung, des verteilten Rechnen sowie der Geschäftsprozesse und berücksichtigt auch die Integration von Geschäftsanliegen und Informationstechnologien. Die Klausurtagung des Forschungskollegs "Service-oriented Systems Engineering" findet einmal jährlich statt und bietet allen Kollegiaten die Möglichkeit den Stand ihrer aktuellen Forschung darzulegen. Bedingt durch die Querschnittstruktur des Kollegs deckt dieser Bericht ein weites Spektrum aktueller Forschungsthemen ab. Dazu zählen unter anderem Human Computer Interaction and Computer Vision as Service; Service-oriented Geovisualization Systems; Algorithm Engineering for Service-oriented Systems; Modeling and Verification of Self-adaptive Service-oriented Systems; Tools and Methods for Software Engineering in Service-oriented Systems; Security Engineering of Service-based IT Systems; Service-oriented Information Systems; Evolutionary Transition of Enterprise Applications to Service Orientation; Operating System Abstractions for Service-oriented Computing; sowie Services Specification, Composition, and Enactment. T3 - Technische Berichte des Hasso-Plattner-Instituts für Digital Engineering an der Universität Potsdam - 129 KW - Hasso Plattner Institute KW - research school KW - Ph.D. retreat KW - service-oriented systems engineering KW - Hasso-Plattner-Institut KW - Forschungskolleg KW - Klausurtagung KW - Service-oriented Systems Engineering Y1 - 2019 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:kobv:517-opus4-427535 SN - 978-3-86956-465-4 SN - 1613-5652 SN - 2191-1665 IS - 129 PB - Universitätsverlag Potsdam CY - Potsdam ER -