Fakultät Informatik
Invariant concept classes form the backbone of classification algorithms immune to specific data transformations, ensuring consistent predictions regardless of these alterations. However, this robustness can come at the cost of limited access to the original sample information, potentially impacting generalization performance. This study introduces an addition to these classes—the permutation-invariant linear classifiers. Distinguished by their structural characteristics, permutation-invariant linear classifiers are unaffected by permutations on feature vectors, a property not guaranteed by other non-constant linear classifiers. The study characterizes this new concept class, highlighting its constant capacity, independent of input dimensionality. In practical assessments using linear support vector machines, the permutation-invariant classifiers exhibit superior performance in permutation experiments on artificial datasets and real mutation profiles. Interestingly, they outperform general linear classifiers not only in permutation experiments but also in permutation-free settings, surpassing unconstrained counterparts. Additionally, findings from real mutation profiles support the significance of tumor mutational burden as a biomarker.
The integration of deep learning-based tools into diagnostic workflows is increasingly prevalent due to their efficiency and reproducibility in various settings. We investigated the utility of automated nuclear morphometry for assessing nuclear pleomorphism (NP), a criterion of malignancy in the current grading system in canine pulmonary carcinoma (cPC), and its prognostic implications. We developed a deep learning-based algorithm for evaluating NP (variation in size, i.e., anisokaryosis and/or shape) using a segmentation model. Its performance was evaluated on 46 cPC cases with comprehensive follow-up data regarding its accuracy in nuclear segmentation and its prognostic ability. Its assessment of NP was compared to manual morphometry and established prognostic tests (pathologists’ NP estimates (n = 11), mitotic count, histological grading, and TNM-stage). The standard deviation (SD) of the nuclear area, indicative of anisokaryosis, exhibited good discriminatory ability for tumor-specific survival, with an area under the curve (AUC) of 0.80 and a hazard ratio (HR) of 3.38. The algorithm achieved values comparable to manual morphometry. In contrast, the pathologists’ estimates of anisokaryosis resulted in HR values ranging from 0.86 to 34.8, with slight inter-observer reproducibility (k = 0.204). Other conventional tests had no significant prognostic value in our study cohort. Fully automated morphometry promises a time-efficient and reproducible assessment of NP with a high prognostic value. Further refinement of the algorithm, particularly to address undersegmentation, and application to a larger study population are required.
ÖPNV für Kinder
(2024)
Diese Bachelorarbeit befasst sich mit der nutzerzentrierten Konzeption hochautomatisierter Shuttlebusse im öffentlichen Personennahverkehr (ÖPNV) mit einem besonderen Fokus auf die Bedürfnisse von Kindern und deren Eltern. Ziel der Arbeit war es, herauszufinden, wie Shuttlebusse gestaltet werden müssen, um sie für Kinder optimal nutzbar zu machen und gleichzeitig das Vertrauen der Eltern in die Technologie zu gewinnen. Um die Perspektiven der zwei Zielgruppen besser nachvollziehen zu können, wurden Interviews mit Schulkindern und eine Onlineumfrage mit den Eltern durchgeführt. Basierend auf den Erkenntnissen wurden dann Prototypen entwickelt und in Nutzertests evaluiert. Die Ergebnisse der Studie mit Kindern zeigen, dass diese positiv und offen gegenüber der neuen Technologie eingestellt sind. Zudem lässt sich erkennen, dass den Kindern vor allem ein großes Medienangebot während der Fahrt wichtig ist. Die Studienergebnisse der Eltern haben gezeigt, dass durch erhöhte Level an Echtzeitinformationen, während der Fahrt der Kinder, die Zufriedenheit gesteigert wird. Zusätzlich dazu, wird der Trend erkennbar, dass durch die Bereitstellung vermehrter Informationen auch das Vertrauen in das System ansteigen kann. Die Arbeit liefert einige Empfehlungen dazu, wie die Bedürfnisse beider Zielgruppen miteinander vereint werden können.
Background
Anti-vascular endothelial growth factor (VEGF) monoclonal antibodies (mAbs) are widely used for tumor treatment, including metastatic colorectal cancer (mCRC). So far, there are no biomarkers that reliably predict resistance to anti-VEGF mAbs like bevacizumab. A biomarker-guided strategy for early and accurate assessment of resistance could avoid the use of non-effective treatment and improve patient outcomes. We hypothesized that repeated analysis of multiple cytokines and angiogenic growth factors (CAFs) before and during treatment using machine learning could provide an accurate and earlier, i.e., 100 days before conventional radiologic staging, prediction of resistance to first-line mCRC treatment with FOLFOX plus bevacizumab.
Patients and methods
15 German and Austrian centers prospectively recruited 50 mCRC patients receiving FOLFOX plus bevacizumab as first-line treatment. Plasma samples were collected every two weeks until radiologic progression (RECIST 1.1) as determined by CT scans performed every 2 months. 102 pre-selected CAFs were centrally analyzed using a cytokine multiplex assay (Luminex, Myriad RBM).
Results
Using random forests, we developed a predictive machine learning model that discriminated between the situations of “no progress within 100 days before radiological progress” and “progress within 100 days before radiological progress”. We could further identify a combination of ten out of the 102 CAF markers, which fulfilled this task with 78.2% accuracy, 71.8% sensitivity, and 82.5% specificity.
Conclusions
We identified a CAF marker combination that indicates treatment resistance to FOLFOX plus bevacizumab in patients with mCRC within 100 days prior to radiologic progress.
Cooperative intelligent transport systems (C-ITSs) are mass-produced and sold in Europe, promising enhanced safety and comfort. Direct vehicle communication, known as vehicle-to-everything (V2X) communication, is crucial in this context. Drivers receive warnings about potential hazards by exchanging vehicle status and environmental data with other communication-enabled vehicles. However, the impact of these warnings on drivers and their inclusion in accident reconstruction remains uncertain. Unlike sensor-based warnings, V2X warnings may not provide a visible reason for the alert, potentially affecting reaction times and behavior. In this work, a simulator study on V2X warnings was conducted with 32 participants to generate findings on reaction times and behavior for accident reconstruction in connection with these systems. Two scenarios from the Car-2-Car Communication Consortium were implemented: “Stationary Vehicle Warning—Broken-Down Vehicle” and “Dangerous Situation—Electronic Emergency Brake Lights”. Volkswagen’s warning concept was utilized, as they are the sole provider of cooperative vehicles in Europe. Results show that V2X warnings without visible reasons did not negatively impact reaction times or behavior, with average reaction times between 0.58 s (steering) and 0.69 s (braking). No significant distraction or search for warning reasons was observed. However, additional information in the warnings caused confusion and was seldom noticed by subjects. In this study, participants responded correctly and appropriately to the shown false-positive warnings. A wrong reaction triggering an accident is possible but unlikely. Overall, V2X warnings showed no negative impacts compared with sensor-based systems. This means that there are no differences in accident reconstruction regarding the source of the warning (sensors or communication). However, it is important that it is known that there was a warning, which is why the occurrence of V2X warnings should also be saved in the EDR in the future.
Markov reward models are commonly used in the analysis of systems by integrating a reward rate to each system state. Typically, rewards are defined based on system states and reflect the system’s perspective. From a user’s point of view, it is important to consider the changing system conditions and dynamics while the user consumes a service. The key contributions of this paper are proper definitions for (i) system-centric reward and (ii) user-centric reward of the Erlang loss model M/M/n-0 and M/M(x)/n with state-dependent service rates, as well as (iii) the analysis of the relationships between those metrics. Our key result allows a simple computation of the user-centric rewards. The differences between the system-centric and the user-centric rewards are demonstrated for a real-world cloud gaming use case. To the best of our knowledge, this is the first analysis showing the relationship between user-centric rewards and system-centric rewards. This work gives relevant and important insights in how to integrate the user’s perspective in the analysis of Markov reward models and is a blueprint for the analysis of other services beyond cloud gaming while also considering user engagement.
The QUILT-1M dataset is the first openly available dataset containing images harvested from various online sources. While it provides a huge data variety, the image quality and composition is highly heterogeneous, impacting its utility for text-conditional image synthesis. We propose an automatic pipeline that provides predictions of the most common impurities within the images, e.g., visibility of narrators, desktop environment and pathology software, or text within the image. Additionally, we propose to use semantic alignment filtering of the image-text pairs. Our findings demonstrate that by rigorously filtering the dataset, there is a substantial enhancement of image fidelity in text-to-image tasks.
Among the possible traffic members on a Vehicle-to-Everything network, the term Vulnerable Road User (VRU) is assigned e.g. to pedestrians and cyclists. The VRU Awareness Message (VAM) is used by VRUs to inform other users of their presence and ensure they are perceived in a traffic system. Since the number of VRUs in crowded areas might be very high, the over-the-air traffic might be overloaded. To reduce channel overload, VAMs offer a clustering feature in which VRUs with similar kinematics and positions can group themselves so that only one device transmits messages. The VRU Basic Service specification describes the cluster as a bounding box that must cover all its members using a geometric shape so that other vehicles in the vicinity can avoid colliding with the contained VRUs. This paper contributes to the standardization effort by introducing a data structure, the Cluster Map, for the clustering in the VRU Basic Service. Furthermore, this work is the first to suggest strategies for forming bounding box shapes. Simulation results show that each of the geometry types is useful in different situations, thus further research on the topic is advised.