@misc{MeinikheimMendelProbstetal., author = {Meinikheim, Michael and Mendel, Robert and Probst, Andreas and Scheppach, Markus W. and Messmann, Helmut and Palm, Christoph and Ebigbo, Alanna}, title = {Barrett-Ampel}, series = {Zeitschrift f{\"u}r Gastroenterologie}, volume = {60}, journal = {Zeitschrift f{\"u}r Gastroenterologie}, number = {08}, publisher = {Georg Thieme Verlag}, address = {Stuttgart}, doi = {10.1055/s-0042-1755109}, abstract = {Hintergrund Adenokarzinome des {\"O}sophagus sind bis heute mit einer infausten Prognose vergesellschaftet (1). Obwohl Endoskopiker mit Barrett-{\"O}sophagus als Pr{\"a}kanzerose konfrontiert werden, ist vor allem f{\"u}r nicht-Experten die Differenzierung zwischen Barrett-{\"O}sophagus ohne Dysplasie und assoziierten Neoplasien mitunter schwierig. Existierende Biopsieprotokolle (z.B. Seattle Protokoll) sind oftmals unzuverl{\"a}ssig (2). Eine fr{\"u}hzeitige Diagnose des Adenokarzinoms ist allerdings von fundamentaler Bedeutung f{\"u}r die Prognose des Patienten. Forschungsansatz Auf der Grundlage dieser Problematik, entwickelten wir in Kooperation mit dem Forschungslabor „Regensburg Medical Image Computing (ReMIC)" der OTH Regensburg ein auf k{\"u}nstlicher Intelligenz (KI) basiertes Entscheidungsunterst{\"u}tzungssystem (CDSS). Das auf einer DeepLabv3+ neuronalen Netzwerkarchitektur basierende CDSS differenziert mittels Mustererkennung Barrett- {\"O}sophagus ohne Dysplasie von Barrett-{\"O}sophagus mit Dysplasie bzw. Neoplasie („Klassifizierung"). Hierbei werden gemittelte Ausgabewahrscheinlichkeiten mit einem vom Benutzer definierten Schwellenwert verglichen. F{\"u}r Vorhersagen, die den Schwellenwert {\"u}berschreiten, berechnen wir die Kontur der Region und die Fl{\"a}che. Sobald die vorhergesagte L{\"a}sion eine bestimmte Gr{\"o}ße in der Eingabe {\"u}berschreitet, heben wir sie und ihren Umriss hervor. So erm{\"o}glicht eine farbkodierte Visualisierung eine Abgrenzung zwischen Dysplasie bzw. Neoplasie und normalem Barrett-Epithel („Segmentierung"). In einer Studie an Bildern in „Weißlicht" (WL) und „Narrow Band Imaging" (NBI) demonstrierten wir eine Sensitivit{\"a}t von mehr als 90\% und eine Spezifit{\"a}t von mehr als 80\% (3). In einem n{\"a}chsten Schritt, differenzierte unser KI-Algorithmus Barrett- Metaplasien von assoziierten Neoplasien anhand von zuf{\"a}llig abgegriffenen Bildern in Echtzeit mit einer Accuracy von 89.9\% (4). Darauf folgend, entwickelten wir unser System dahingehend weiter, dass unser Algorithmus nun auch dazu in der Lage ist, Untersuchungsvideos in WL, NBI und „Texture and Color Enhancement Imaging" (TXI) in Echtzeit zu analysieren (5). Aktuell f{\"u}hren wir eine Studie in einem randomisiert-kontrollierten Ansatz an unver{\"a}nderten Untersuchungsvideos in WL, NBI und TXI durch. Ausblick Um Patienten mit aus Barrett-Metaplasien resultierenden Neoplasien fr{\"u}hestm{\"o}glich an „High-Volume"-Zentren {\"u}berweisen zu k{\"o}nnen, soll unser KI-Algorithmus zuk{\"u}nftig vor allem Endoskopiker ohne extensive Erfahrung bei der Beurteilung von Barrett- {\"O}sophagus in der Krebsfr{\"u}herkennung unterst{\"u}tzen.}, subject = {Speiser{\"o}hrenkrebs}, language = {de} } @misc{ScheppachMendelProbstetal., author = {Scheppach, Markus W. and Mendel, Robert and Probst, Andreas and Meinikheim, Michael and Palm, Christoph and Messmann, Helmut and Ebigbo, Alanna}, title = {Intraprozedurale Strukturerkennung bei Third-Space Endoskopie mithilfe eines Deep-Learning Algorithmus}, series = {Zeitschrift f{\"u}r Gastroenterologie}, volume = {60}, journal = {Zeitschrift f{\"u}r Gastroenterologie}, number = {04}, publisher = {Thieme}, address = {Stuttgart}, doi = {10.1055/s-0042-1745652}, pages = {e250-e251}, abstract = {Einleitung Third-Space Interventionen wie die endoskopische Submukosadissektion (ESD) und die perorale endoskopische Myotomie (POEM) sind technisch anspruchsvoll und mit einem erh{\"o}hten Risiko f{\"u}r intraprozedurale Komplikationen wie Blutung oder Perforation assoziiert. Moderne Computerprogramme zur Unterst{\"u}tzung bei diagnostischen Entscheidungen werden unter Einsatz von k{\"u}nstlicher Intelligenz (KI) in der Endoskopie bereits erfolgreich eingesetzt. Ziel der vorliegenden Arbeit war es, relevante anatomische Strukturen mithilfe eines Deep-Learning Algorithmus zu detektieren und segmentieren, um die Sicherheit und Anwendbarkeit von ESD und POEM zu erh{\"o}hen. Methoden Zw{\"o}lf Videoaufnahmen in voller L{\"a}nge von Third-Space Endoskopien wurden aus der Datenbank des Universit{\"a}tsklinikums Augsburg extrahiert. 1686 Einzelbilder wurden f{\"u}r die Kategorien Submukosa, Blutgef{\"a}ß, Dissektionsmesser und endoskopisches Instrument annotiert und segmentiert. Mit diesem Datensatz wurde ein DeepLabv3+neuronales Netzwerk auf der Basis eines ResNet mit 101 Schichten trainiert und intern anhand der Parameter Intersection over Union (IoU), Dice Score und Pixel Accuracy validiert. Die F{\"a}higkeit des Algorithmus zur Gef{\"a}ßdetektion wurde anhand von 24 Videoclips mit einer Spieldauer von 7 bis 46 Sekunden mit 33 vordefinierten Gef{\"a}ßen evaluiert. Anhand dieses Tests wurde auch die Gef{\"a}ßdetektionsrate eines Experten in der Third-Space Endoskopie ermittelt. Ergebnisse Der Algorithmus zeigte eine Gef{\"a}ßdetektionsrate von 93,94\% mit einer mittleren Rate an falsch positiven Signalen von 1,87 pro Minute. Die Gef{\"a}ßdetektionsrate des Experten lag bei 90,1\% ohne falsch positive Ergebnisse. In der internen Validierung an Einzelbildern wurde eine IoU von 63,47\%, ein mittlerer Dice Score von 76,18\% und eine Pixel Accuracy von 86,61\% ermittelt. Zusammenfassung Dies ist der erste KI-Algorithmus, der f{\"u}r den Einsatz in der therapeutischen Endoskopie entwickelt wurde. Pr{\"a}limin{\"a}re Ergebnisse deuten auf eine mit Experten vergleichbare Detektion von Gef{\"a}ßen w{\"a}hrend der Untersuchung hin. Weitere Untersuchungen sind n{\"o}tig, um die Leistung des Algorithmus im Vergleich zum Experten genauer zu eruieren sowie einen m{\"o}glichen klinischen Nutzen zu ermitteln.}, language = {de} } @article{SouzaJrPachecoPassosetal., author = {Souza Jr., Luis Antonio de and Pacheco, Andr{\´e} G.C. and Passos, Leandro A. and Santana, Marcos Cleison S. and Mendel, Robert and Ebigbo, Alanna and Probst, Andreas and Messmann, Helmut and Palm, Christoph and Papa, Jo{\~a}o Paulo}, title = {DeepCraftFuse: visual and deeply-learnable features work better together for esophageal cancer detection in patients with Barrett's esophagus}, series = {Neural Computing and Applications}, volume = {36}, journal = {Neural Computing and Applications}, publisher = {Springer}, address = {London}, doi = {10.1007/s00521-024-09615-z}, pages = {10445 -- 10459}, abstract = {Limitations in computer-assisted diagnosis include lack of labeled data and inability to model the relation between what experts see and what computers learn. Even though artificial intelligence and machine learning have demonstrated remarkable performances in medical image computing, their accountability and transparency level must be improved to transfer this success into clinical practice. The reliability of machine learning decisions must be explained and interpreted, especially for supporting the medical diagnosis. While deep learning techniques are broad so that unseen information might help learn patterns of interest, human insights to describe objects of interest help in decision-making. This paper proposes a novel approach, DeepCraftFuse, to address the challenge of combining information provided by deep networks with visual-based features to significantly enhance the correct identification of cancerous tissues in patients affected with Barrett's esophagus (BE). We demonstrate that DeepCraftFuse outperforms state-of-the-art techniques on private and public datasets, reaching results of around 95\% when distinguishing patients affected by BE that is either positive or negative to esophageal cancer.}, subject = {Deep Learning}, language = {en} } @article{SouzaPachecodeSouzaetal., author = {Souza, Luis A. and Pacheco, Andr{\´e} G.C. and de Souza, Alberto F. and Oliveira-Santos, Thiago and Badue, Claudine and Palm, Christoph and Papa, Jo{\~a}o Paulo}, title = {TransConv: a lightweight architecture based on transformers and convolutional neural networks for adenocarcinoma and Barrett's esophagus identification}, series = {Neural Computing and Applications}, journal = {Neural Computing and Applications}, number = {37}, publisher = {Springer}, doi = {10.1007/s00521-025-11299-y}, pages = {15535 -- 15546}, abstract = {Barrett's esophagus, also known as BE, is commonly associated with repeated exposure to stomach acid. If not treated properly, it may evolve into esophageal adenocarcinoma, aka esophageal cancer. This paper proposes TransConv, a hybrid architecture that benefits from features learned by pre-trained vision transformers (ViTs) and convolutional neural networks (CNNs), followed by a shallow neural network composed of three normalizations, ReLU activations, and fully connected layers, and a SoftMax head to distinguish between BE and esophageal cancer. TransConv is designed to be training-lightweight, and for the ViT and CNN backbone models, weights are kept frozen during training, i.e., the primary goal of TransConv is to learn the weights of the fully connected layer from both backbones only, avoiding the burden of updating their weights but still learning their final descriptions for the lightweight convolutional model. We report promising results with low computational training costs in two datasets, one public and another private. From our achievements, TransConv was able to deliver balanced accuracy results around 85\% and 86\% for each evaluated dataset, respectively, in a design that required only 50 epochs of model training, a very reduced number compared to state-of-the-art conducted studies in the same domain.}, language = {en} } @unpublished{RueckertRueckertPalm, author = {R{\"u}ckert, Tobias and R{\"u}ckert, Daniel and Palm, Christoph}, title = {Methods and datasets for segmentation of minimally invasive surgical instruments in endoscopic images and videos: A review of the state of the art}, doi = {10.48550/arXiv.2304.13014}, pages = {25}, abstract = {In the field of computer- and robot-assisted minimally invasive surgery, enormous progress has been made in recent years based on the recognition of surgical instruments in endoscopic images. Especially the determination of the position and type of the instruments is of great interest here. Current work involves both spatial and temporal information with the idea, that the prediction of movement of surgical tools over time may improve the quality of final segmentations. The provision of publicly available datasets has recently encouraged the development of new methods, mainly based on deep learning. In this review, we identify datasets used for method development and evaluation, as well as quantify their frequency of use in the literature. We further present an overview of the current state of research regarding the segmentation and tracking of minimally invasive surgical instruments in endoscopic images. The paper focuses on methods that work purely visually without attached markers of any kind on the instruments, taking into account both single-frame segmentation approaches as well as those involving temporal information. A discussion of the reviewed literature is provided, highlighting existing shortcomings and emphasizing available potential for future developments. The publications considered were identified through the platforms Google Scholar, Web of Science, and PubMed. The search terms used were "instrument segmentation", "instrument tracking", "surgical tool segmentation", and "surgical tool tracking" and result in 408 articles published between 2015 and 2022 from which 109 were included using systematic selection criteria.}, language = {en} } @article{SafiBandicNiedermeieretal., author = {Safi, Hila and Bandic, Medina and Niedermeier, Christoph and Almudever, Carmen G. and Feld, Sebastian and Mauerer, Wolfgang}, title = {Stacking the odds: full-stack quantum system design space exploration}, series = {EPJ Quantum Technology}, volume = {12}, journal = {EPJ Quantum Technology}, publisher = {Springer}, address = {Heidelberg}, doi = {10.1140/epjqt/s40507-025-00413-7}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:898-opus4-85157}, pages = {31}, abstract = {Design space exploration (DSE) plays an important role in optimising quantum circuit execution by systematically evaluating different configurations of compilation strategies and hardware settings. In this paper, we conduct a comprehensive investigation into the impact of various layout methods, qubit routing techniques, and optimisation levels, as well as device-specific properties such as different variants and strengths of noise and imperfections, the topological structure of qubits, connectivity densities, and back-end sizes. By spanning through these dimensions, we aim to understand the interplay between compilation choices and hardware characteristics. A key question driving our exploration is whether the optimal selection of device parameters, mapping techniques, comprising of initial layout strategies and routing heuristics can mitigate device induced errors beyond standard error mitigation approaches. Our results show that carefully selecting software strategies (e.g., mapping and routing algorithms) and tailoring hardware characteristics (such as minimising noise and leveraging topology and connectivity density) significantly improve the fidelity of circuit execution outcomes, and thus the expected correctness or success probability of the computational result. We provide estimates based on key metrics such as circuit depth, gate count and expected fidelity. Our results highlight the importance of hardware-software co-design, particularly as quantum systems scale to larger dimensions, and along the way towards fully error corrected quantum systems: Our study is based on computationally noisy simulations, but considers various implementations of quantum error correction (QEC) using the same approach as for other algorithms. The observed sensitivity of circuit fidelity to noise and connectivity suggests that co-design principles will be equally critical when integrating QEC in future systems. Our exploration provides practical guidelines for co-optimising physical mapping, qubit routing, and hardware configurations in realistic quantum computing scenarios.}, language = {en} } @inproceedings{GrafHachaniFischeretal., author = {Graf, Julian and Hachani, Murad and Fischer, Sebastian and Hackenberg, Rudolf}, title = {A heuristic packet processing model for improved encrypted network analysis}, series = {CSCS '25: Proceedings of the 2nd Cyber Security in CarS Workshop}, booktitle = {CSCS '25: Proceedings of the 2nd Cyber Security in CarS Workshop}, publisher = {ACM}, address = {New York, USA}, doi = {10.1145/3736130.3764510}, pages = {12}, abstract = {Modern networked systems, such as those in the automotive sector, face increasing complexity and growing attack surfaces due to the rise of interconnected and data-driven technologies. Detecting malicious behavior in these environments requires efficient and scalable methods that can operate reliably despite limited resources and high communication volumes. This paper proposes a heuristic packet processing model designed to support intrusion detection based on structural and temporal characteristics of encrypted network traffic. The model follows a modular architecture consisting of four key phases: recording, sorting, prioritizing, and analyzing. At the core of the approach is the Polymetric Queueing Topology Space, a feature space that combines statistical and time series attributes derived from model structure and flow data. These features serve as input for machine learning models, which can effectively distinguish between benign and intrusion traffic patterns without relying on packet data beyond the transport layer. The approach was evaluated using the publicly available ToN_IoT dataset and demonstrated that reliable classification is achievable using a subset of the developed feature space that contains model-derived traffic features. We used Random Forest for supervised binary and multi-class classification achieving high accuracy scores of 99\% for binary and 98\% for multi-class classification. Additionally, for unsupervised anomaly detection, we created an Isolation Forest model accomplishing F1-scores of 0.92 for the benign and 0.96 for intrusion class. The architecture is designed to enable dynamic traffic prioritization and to offer a flexible foundation that can observe diverse network domains while maintaining efficient performance under constrained computational conditions.}, language = {en} } @misc{HauserGrafFischer, author = {Hauser, Dominic and Graf, Julian and Fischer, Sebastian}, title = {SEPP - Security Education and Penetration-Testing Platform for IoT}, series = {Conference programme \& abstract book}, journal = {Conference programme \& abstract book}, publisher = {IAFOR}, issn = {2433-7544}, pages = {101}, abstract = {The Internet of Things (IoT) is becoming a major part of our everyday lives, offering convenience and smarter solutions, but also bringing significant security challenges. While theoretical knowledge in IoT security is essential, studies have shown that practical content can be an essential part of internalizing understanding. To address this, we developed the Security Education and Penetration-Testing Platform (SEPP) as the practical component of an existing IoT security course at the OTH Regensburg. SEPP uses real IoT devices like smart locks, cameras, and plugs, simulating a smart home environment to make learning interactive and engaging. Students can explore vulnerabilities, conduct penetration tests, and document their findings through structured exercises. By working on tasks like network scanning, analyzing data traffic, and simulating attacks, students gain a deeper understanding of IoT security risks. Initial tests show that this approach helps students apply their theoretical knowledge and significantly improve their practical skills. This paper explains how SEPP was built, the exercises it offers, and why it's an important step forward in teaching IoT security effectively. Furthermore, we aim to share the findings and tasks from this paper with other universities, providing them with a solid foundation to teach practical IoT security knowledge in their own courses.}, language = {en} } @misc{Fischer, author = {Fischer, Sebastian}, title = {The persistent problems with cybersecurity : a negative example and an outlook on the Cyber Resilience Act}, publisher = {IARIA}, pages = {44}, language = {en} } @inproceedings{BauerFrikel, author = {Bauer, Patrick and Frikel, J{\"u}rgen}, title = {BPConvNet: a deep learning based ρ-Filtered layergram reconstruction method for computed tomography}, series = {AIP Conference Proceedings}, volume = {3315}, booktitle = {AIP Conference Proceedings}, number = {1}, publisher = {AIP Publishing}, issn = {0094-243X}, doi = {10.1063/5.0286063}, abstract = {In this article, we address the reconstruction problem in computed tomography (CT) when dealing with sparse view data. Traditional approaches like filtered backprojection (FBP) often fail under these conditions, leading to streaking artifacts. We propose BPConvNet, a deep learning based version of the ρ-filtered layergram or backprojection filtration (BPF) technique (cf. [1]). Unlike FBP, the BPF method applies filtering (F) after backprojection (BP), hence the name. The proposed BPConvNet adapts the BPF workflow by substituting the filtering step with a residual convolutional neural network. Our numerical experiments demonstrate that BPConvNet is competitive to similar deep learning methods. Moreover, we explain that BPConvNet can be easily adapted to to different CT acquisition geometries, such as fan beam and 3D configurations}, language = {en} } @misc{Schaffer, type = {Master Thesis}, author = {Schaffer, Josefa}, title = {Precision in eye tracking: evaluation and improvement of accuracy}, address = {Regensburg}, doi = {10.35096/othr/pub-8529}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:898-opus4-85296}, school = {Ostbayerische Technische Hochschule Regensburg}, pages = {IX, 74}, abstract = {This thesis will evaluate the accuracy of the Tobii Pro Glasses 3, which are wearable eye tracking glasses. A static and dynamic accuracy study will be conducted, utilising a poster and imitating realistic environments. The results will be evaluated using a self-developed sample-wise approach, trying to determine possible differences between static and dynamic conditions while wearing eye tracking glasses. In addition to that, a tool named glassesValidator, used to calculate data quality metrics of eye tracking glasses, will be evaluated on its accuracy and validity itself. In the static accuracy study, participants will focus, in reading order, on nine fixation targets on the poster for at least two seconds while using a chin rest to prevent head movement. In the dynamic accuracy study participants will be walking in a semicircular movement around the poster while focusing only on the central fixation target on the poster. The results for the static and dynamic study will be compared to each other, providing possible influence of different lighting conditions and the usage of contact lenses. The results provided by glassesValidator and the self-developed accuracy approach will also be compared to each other to evaluate and provide possible improvements to increase the accuracy of the tool.}, language = {en} } @article{NeubauerFischerHackenberg, author = {Neubauer, Katrin and Fischer, Sebastian and Hackenberg, Rudolf}, title = {Security risk analysis of the cloud infrastructure of Smart Grid and IoT - 4-Level-Trust-Model as a security solution}, series = {International Journal on Advances in Internet Technology}, volume = {13}, journal = {International Journal on Advances in Internet Technology}, number = {1\&2}, pages = {11 -- 20}, abstract = {The digital transformation has found its way into business and private life. It consists of digitization and digitaliza- tion. Digitization means the technical process and digitalization is the socio-technological process. Technologies of digitization are Cloud Computing (CC), Internet of Things (IoT) and Smart Grid (SG), which are separate technologies. The increasing digitalization in the private sector and of the energy industry connect these technologies. Actually, there is no connection between the CC infrastructure and the SG infrastructure at the moment, because in Germany the SG is currently under construction. If one looks at the CC and IoT, it must be stated there is an connection between the IoT infrastructure and the CC infrastructure as a service provider. To connect the technologies CC, IoT and SG and also build an SG cloud for innovative services, the new laws for privacy must be implemented. For privacy and security analyses it is important to know which data can be stored and distributed on a cloud. To illustrate this analysis, we connect the SG infrastructure with the IoT. An IoT device (car charging station) should be able to transfer data to and from the SG. SG is a critical infrastructure and the IoT device a potential insecure device and network. We show the communication between the smart meter switching box and the IoT device and the data transferred between their clouds. The charging station is connected to the SG to get the current amount of renewable energy in the grid. This is necessary to create a new smart service. But this service also generates private data (e.g., name, address, payment details). The private data should not be transferred to the IoT cloud. For the connection of SG and IoT, availability, confidentiality and integrity must be ensured. A risk analysis over all the cloud connections, including the vulnerability and the ability of an attacker, the resulting risk and the 4-Level- Trust-Model for security assessment are developed. Furthermore, we show the application of the 4-Level-Trust-Model in this paper.}, language = {en} } @inproceedings{SchwaegerlBuchmannWestfechtel, author = {Schw{\"a}gerl, Felix and Buchmann, Thomas and Westfechtel, Bernhard}, title = {Multi-variant model transformations - a problem statement}, series = {Proceedings of the 11th International Conference on Evaluation of Novel Software Approaches to Software Engineering}, booktitle = {Proceedings of the 11th International Conference on Evaluation of Novel Software Approaches to Software Engineering}, publisher = {SCITEPRESS}, isbn = {978-989-758-189-2}, issn = {2184-4895}, doi = {10.5220/0005878702030209}, pages = {203 -- 209}, abstract = {Model Transformations are a key element of Model-Driven Software Engineering. As soon as variability is involved, transformations become increasingly complicated. The lack of support for variability in model transformations impairs the acceptance of approaches to organized reuse such as software product lines. In this position paper, the general problem of multi-variant model transformations is formulated for MOF-based, XMI-serialized models. A simplistic case study is presented to specify the input and the expected output of such a transformation. Furthermore, requirements for tool support are defined, including a standardized representation of both multi-variant model instances and variability information, as well as an execution specification for multi-variant transformations. A literature review reveals that the problem is weakly identified and often solved using ad-hoc solutions; there exists no tool providing a general solution to the proposed problem statement. The observation s presented here may serve for the future development of standards and tools.}, language = {en} }