0 Informatik, Informationswissenschaft, allgemeine Werke
Refine
Document Type
- Article (5)
Keywords
- Adaptivität (1)
- Automatic Item Generation (1)
- Forschungsdaten-Policy (1)
- Forschungsdatenmanagement (1)
- Interaktivität (1)
- Lernaufgaben (1)
- Strategieentwicklung (1)
- Train-the-Trainer (1)
- abdominal imaging (1)
- digitale Medien (1)
Generative multi-adversarial network for striking the right balance in abdominal image segmentation
(2020)
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.
Hochschulen und deren Zentraleinrichtungen beschäftigen sich zunehmend mit dem Thema Forschungsdatenmanagement (FDM), um ihre Forschenden adäquat zu unterstützen. Nicht zuletzt aufgrund neuer Verlags- und Förderanforderungen wünschen sich Forschende Beratung und Orientierung, wie sie mit ihren Forschungsdaten umgehen sollen. Damit Hochschulen schnell und nachhaltig Lösungen zum institutionellen FDM etablieren können, haben fünf Berliner und Brandenburger Universitäten im gemeinsamen Verbundvorhaben FDMentor mit Förderung des Bundesministeriums für Bildung und Forschung (BMBF) entsprechende Leitfäden und Werkzeuge erarbeitet. Die innerhalb von zwei Jahren (2017–2019) entstandenen Ergebnisse in den Bereichen Strategieentwicklung, Forschungsdaten-Policy, rechtliche Aspekte und Kompetenzausbau finden über das Verbundprojekt hinaus ihre Anwendung.
Digitale Medien enthalten bislang vor allem Inhalte in verschiedenen Darstellungsformen. Dies allein erzeugt jedoch nur einen geringen Mehrwert zu klassischen Lernressourcen, da die Kriterien der Interaktivität und Adaptivität nicht mit einbezogen werden. Dies scheitert jedoch oft an dem damit verbundenen Erstellungsaufwand. Der folgende Beitrag zeigt, wie durch die automatische Erzeugung von Aufgaben ein hochwertiger Wissenserwerb mit digitalen Medien ermöglicht wird. Ferner werden Vor- und Nachteile der automatischen Erstellung von Aufgaben erörtert.
Automatisierungstechnik II
(2018)
The simultaneous switching activity in digital circuits challenges the design of mixed-signal SoCs. Rather than focusing on time-domain noise voltage minimization, this work optimizes switching noise in the frequency domain. A two-tier solution based on the on-chip clock scheduling is proposed. First, to cope with the switching noise at the fundamental clock frequency, which usually dominates in terms of noise power, a two-phase clocking scheme is employed for system timing. Second, on-chip clock latencies are manipulated to target harmonic peaks in specific frequency bands for the spectral noise optimization. An automated design flow, which allows for noise optimization in user-defined application-specific frequency bands, is developed. The effectiveness of our design solution is validated by measurements of substrate noise and conductive EMI (electromagnetic interference) noise on a test chip, which consists of four wireless sensor node baseband processors each addressing a distinct clock-tree-synthesis strategy. Compared to the reference synchronous design, the proposed clock scheduling solution substantially reduces noise in the target GSM-850 band, i.e., by 11.1 dB on the substrate noise and 12.9 dB on the EMI noise, along with dramatic noise peak drops measured at the 50-MHz clock frequency.