Open Access

Interpretability and uncertainty modeling are important key factors for medical applications. Moreover, data in medicine are often available as a combination of unstructured data like images and structured predictors like patient’s metadata. While deep learning models are state-of-the-art for image classification, the models are often referred to as ’black-box’, caused by the lack of interpretability. Moreover, DL models are often yielding point predictions and are too confident about the parameter estimation and outcome predictions. On the other side with statistical regression models, it is possible to obtain interpretable predictor effects and capture parameter and model uncertainty based on the Bayesian approach. In this thesis, a publicly available melanoma dataset, consisting of skin lesions and patient’s age, is used to predict the melanoma types by using a semi-structured model, while interpretable components and model uncertainty is quantified. For Bayesian models, transformation model-based variational inference (TM-VI) method is used to determine the posterior distribution of the parameter. Several model constellations consisting of patient’s age and/or skin lesion were implemented and evaluated. Predictive performance was shown to be best by using a combined model of image and patient’s age, while providing the interpretable posterior distribution of the regression coefficient is possible. In addition, integrating uncertainty in image and tabular parts results in larger variability of the outputs corresponding to high uncertainty of the single model components.

Background: One of the most promising health care development areas is introducing telemedicine services and creating solutions based on blockchain technology. The study of systems combining both these domains indicates the ongoing expansion of digital technologies in this market segment. Objective: This paper aims to review the feasibility of blockchain technology for telemedicine. Methods: The authors identified relevant studies via systematic searches of databases including PubMed, Scopus, Web of Science, IEEE Xplore, and Google Scholar. The suitability of each for inclusion in this review was assessed independently. Owing to the lack of publications, available blockchain-based tokens were discovered via conventional web search engines (Google, Yahoo, and Yandex). Results: Of the 40 discovered projects, only 18 met the selection criteria. The 5 most prevalent features of the available solutions (N=18) were medical data access (14/18, 78%), medical service processing (14/18, 78%), diagnostic support (10/18, 56%), payment transactions (10/18, 56%), and fundraising for telemedical instrument development (5/18, 28%). Conclusions: These different features (eg, medical data access, medical service processing, epidemiology reporting, diagnostic support, and treatment support) allow us to discuss the possibilities for integration of blockchain technology into telemedicine and health care on different levels. In this area, a wide range of tasks can be identified that could be accomplished based on digital technologies using blockchains.

Normal breathing during sleep is essential for people’s health and well-being. Therefore, it is crucial to diagnose apnoea events at an early stage and apply appropriate therapy. Detection of sleep apnoea is a central goal of the system design described in this article. To develop a correctly functioning system, it is first necessary to define the requirements outlined in this manuscript clearly. Furthermore, the selection of appropriate technology for the measurement of respiration is of great importance. Therefore, after performing initial literature research, we have analysed in detail three different methods and made a selection of a proper one according to determined requirements. After considering all the advantages and disadvantages of the three approaches, we decided to use the impedance measurement-based one. As a next step, an initial conceptual design of the algorithm for detecting apnoea events was created. As a result, we developed an activity diagram on which the main system components and data flows are visually represented.

A residual neural network was adapted and applied to the Physionet/Computing data in Cardiology Challenge 2020 to detect 24 different classes of cardiac abnormalities from 12-lead. Additive Gaussian noise, signal shifting, and the classification of signal sections of different lengths were applied to prevent the network from overfitting and facilitating generalization. Due to the use of a global pooling layer after the feature extractor, the network is independent of the signal’s length. On the hidden test set of the challenge, the model achieved a validation score of 0.656 and a full test score of 0.27, placing us 15th out of 41 officially ranked teams (Team name: UC_Lab_Kn). These results show the potential of deep neural networks for ap- plication to raw data and a complex multi-class multi-label classification problem, even if the training data is from di- verse datasets and of differing lengths.

The main aim of presented in this manuscript research is to compare the results of objective and subjective measurement of sleep quality for older adults (65+) in the home environment. A total amount of 73 nights was evaluated in this study. Placing under the mattress device was used to obtain objective measurement data, and a common question on perceived sleep quality was asked to collect the subjective sleep quality level. The achieved results confirm the correlation between objective and subjective measurement of sleep quality with the average standard deviation equal to 2 of 10 possible quality points.

We propose and apply a requirements engineering approach that focuses on security and privacy properties and takes into account various stakeholder interests. The proposed methodology facilitates the integration of security and privacy by design into the requirements engineering process. Thus, specific, detailed security and privacy requirements can be implemented from the very beginning of a software project. The method is applied to an exemplary application scenario in the logistics industry. The approach includes the application of threat and risk rating methodologies, a technique to derive technical requirements from legal texts, as well as a matching process to avoid duplication and accumulate all essential requirements.

The evaluation of the effectiveness of different machine learning algorithms on a publicly available database of signals derived from wearable devices is presented with the goal of optimizing human activity recognition and classification. Among the wide number of body signals we choose a couple of signals, namely photoplethysmographic (optically detected subcutaneous blood volume) and tri-axis acceleration signals that are easy to be simultaneously acquired using commercial widespread devices (e.g. smartwatches) as well as custom wearable wireless devices designed for sport, healthcare, or clinical purposes. To this end, two widely used algorithms (decision tree and k-nearest neighbor) were tested, and their performance were compared to two new recent algorithms (particle Bernstein and a Monte Carlo-based regression) both in terms of accuracy and processing time. A data preprocessing phase was also considered to improve the performance of the machine learning procedures, in order to reduce the problem size and a detailed analysis of the compression strategy and results is also presented.

Polysomnography is a gold standard for a sleep study, and it provides very accurate results, but its cost (both personnel and material) are quite high. Therefore, the development of a low-cost system for overnight breathing and heartbeat monitoring, which provides more comfort while recording the data, is a well-motivated challenge. The system proposed in this manuscript is based on the usage of resistive pressure sensors installed under the mattress. These sensors can measure slight pressure changes provoked during breathing and heartbeat. The captured signal requires advanced processing, like applying filters and amplifiers before the analog signal is ready for the next step. Then, the output signal is digitalized and further processed by an algorithm that performs a custom filtering before it can recognize breathing and heart rate in real-time. The result can be directly visualized. Furthermore, a CSV file is created containing the raw data, timestamps, and unique IDs to facilitate further processing. The achieved results are promising, and the average deviation from a reference device is about 4bpm.