Many previous studies claim to have developed machine learning models that diagnose COVID-19 from blood tests. However, we hypothesize that changes in the underlying distribution of the data, so called domain shifts, affect the predictive performance and reliability and are a reason for the failure of such machine learning models in clinical application. Domain shifts can be caused, e.g., by changes in the disease prevalence (spreading or tested population), by refined RT-PCR testing procedures (way of taking samples, laboratory procedures), or by virus mutations. Therefore, machine learning models for diagnosing COVID-19 or other diseases may not be reliable and degrade in performance over time. We investigate whether domain shifts are present in COVID-19 datasets and how they affect machine learning methods. We further set out to estimate the mortality risk based on routinely acquired blood tests in a hospital setting throughout pandemics and under domain shifts. We reveal domain shifts by evaluating the models on a large-scale dataset with different assessment strategies, such as temporal validation. We present the novel finding that domain shifts strongly affect machine learning models for COVID-19 diagnosis and deteriorate their predictive performance and credibility. Therefore, frequent re-training and re-assessment are indispensable for robust models enabling clinical utility.

Electronic health records (EHRs) have been successfully used in data science and machine learning projects. However, most of these data are collected for clinical use rather than for retrospective analysis. This means that researchers typically face many different issues when attempting to access and prepare the data for secondary use. We aimed to investigate how raw EHRs can be accessed and prepared in retrospective data science projects in a disciplined, effective, and efficient way. We report our experience and findings from a large-scale data science project analyzing routinely acquired retrospective data from the Kepler University Hospital in Linz, Austria. The project involved data collection from more than 150,000 patients over a period of 10 years. It included diverse data modalities, such as static demographic data, irregularly acquired laboratory test results, regularly sampled vital signs, and high-frequency physiological waveform signals. Raw medical data can be corrupted in many unexpected ways that demand thorough manual inspection and highly individualized data cleaning solutions. We present a general data preparation workflow, which was shaped in the course of our project and consists of the following 7 steps: obtain a rough overview of the available EHR data, define clinically meaningful labels for supervised learning, extract relevant data from the hospital’s data warehouses, match data extracted from different sources, deidentify them, detect errors and inconsistencies therein through a careful exploratory analysis, and implement a suitable data processing pipeline in actual code. Only few of the data preparation issues encountered in our project were addressed by generic medical data preprocessing tools that have been proposed recently. Instead, highly individualized solutions for the specific data used in one’s own research seem inevitable. We believe that the proposed workflow can serve as a guidance for practitioners, helping them to identify and address potential problems early and avoid some common pitfalls.

The Theorema project aims at the development of a computer assistant for the working mathematician. Support should be given throughout all phases of mathematical activity, from introducing new mathematical concepts by definitions or axioms, through first (computational) experiments, the formulation of theorems, their justification by an exact proof, the application of a theorem as an algorithm, until to the dissemination of the results in form of a mathematical publication, the build up of bigger libraries of certified mathematical content and the like. This ambitious project is exactly along the lines of the QED manifesto issued in 1994 (see e.g. http://www.cs.ru.nl/~freek/qed/qed.html) and it was initiated in the mid-1990s by Bruno Buchberger. The Theorema system is a computer implementation of the ideas behind the Theorema project. One focus lies on the natural style of system input (in form of definitions, theorems, algorithms, etc.), system output (mainly in form of mathematical proofs) and user interaction. Another focus is theory exploration, i.e. the development of large consistent mathematical theories in a formal frame, in contrast to just proving single isolated theorems. When using the Theorema system, a user should not have to follow a certain style of mathematics enforced by the system (e.g. basing all of mathematics on set theory or certain variants of type theory), rather should the system support the user in her preferred flavour of doing math. The new implementation of the system, which we refer to as Theorema 2.0, is open-source and available through GitHub.

Advanced Driver Assistance Systems rely on automated traffic sign recognition. Today, Deep Learning methods outperform other approaches in terms of accuracy and processing time; however, they require vast and well-curated data sets for training. In this paper, we present the Austrian Highway Traffic Sign Data Set (ATSD), a comprehensive annotated data set of images of almost all traffic signs on Austrian highways in 2014, and corresponding images of full traffic scenes they are contained in. Altogether, the data set consists of almost 7500 scene images with more than 28,000 detailed annotations of more than 100 distinct traffic sign classes. It covers diverse environments, ranging from urban to rural and mountainous areas, and includes many images recorded in tunnels. We further evaluate state-of-the-art traffic sign detectors and classifiers on ATSD to establish baselines for future experiments. The data set and our baseline models are freely available online.

We investigate machine learning models that identify COVID-19 positive patients and estimate the mortality risk based on routinely acquired blood tests in a hospital setting. However, during pandemics or new outbreaks, disease and testing characteristics change, thus we face domain shifts. Domain shifts can be caused, e.g., by changes in the disease prevalence (spreading or tested population), by refined RT-PCR testing procedures (taking samples, laboratory), or by virus mutations. Therefore, machine learning models for diagnosing COVID-19 or other diseases may not be reliable and degrade in performance over time. To countermand this effect, we propose methods that first identify domain shifts and then reverse their negative effects on the model performance. Frequent re-training and reassessment, as well as stronger weighting of more recent samples, keeps model performance and credibility at a high level over time. Our diagnosis models are constructed and tested on large-scale data sets, steadily adapt to observed domain shifts, and maintain high ROC AUC values along pandemics.

Quantum machine learning has gained significant momentum in recent years, particularly with the use of parameterized quantum circuits that exhibit a certain level of resilience to noise in current NISQ devices. However, the real-world applicability of QML remains an active field of research as transitioning from idealized simulations to real quantum hardware introduces challenges related to data size and noise. Our study empirically analyzes the performance of popular quantum neural networks in ideal, noise-free conditions, but also in the presence of realistic noise, which was modeled to mimic the noisy behavior of current IBM quantum devices. We carefully selected three medical classification tasks that cover common data modalities with varying numbers of features-from tabular cell data to signal sampling points and image pixels. Our findings statistically show that while complex parameterized circuits perform well in noise-free simulations, their performance significantly deteriorates with increasing qubit numbers under noisy conditions. Our study suggests that parameterized quantum circuits of moderate complexity are a favorable choice for future experiments on noisy quantum devices.