As today’s society ages, age-related diseases become more frequent. One very common but yet preventable disease is the development of pressure ulcers (PUs). PUs can occur if tissue is exposed to a long-lasting pressure load, e.g., lying on tissue without turning. The cure of PUs requires intensive care, especially for the elderly or people with preexisting conditions whose tissue needs longer healing times. The consequences are heavy suffering for the patient and extreme costs for the health care system. To avoid these consequences, our objective is to develop a pressure ulcer prophylaxis device. For that, we built a new sensor system able to monitor the pressure load and tissue vital signs in immediate local proximity at patient’s predilection sites. In the clinical study, we found several indicators showing correlations between tissue perfusion and the risk of PU development, including strongly reduced SpO${}_2$ levels in body tissue prior to a diagnosed PU. Finally, we propose a prophylaxis system that allows for the prediction of PU developments in early stages before they become visible. This work is the first step in generating an effective system to warn patients or caregivers about developing PUs and taking appropriate preventative measures. Widespread application could reduce patient suffering and lead to substantial cost savings.

The causes of pressure ulcers are well known: A high skin surface pressure over a long period of time due to insufficient relief of the tissue and a lack of alarm signals from the nervous system. A healthy person would move and relieve the skin after a while. A medical wearable can warn a patient without natural alarm signals if a high risk of a pressure wound is detected. Clinical data to find a pressure injury predictor with machine learning is not available, yet. This work presents a medical wearable that is able to measure pressure load at wound risk areas as well as the skin temperature and the blood oxygen saturation in close proximity of these areas. The functionality of the medical wearable is demonstrated by acquiring raw data of a healthy test person during seventeen nights. The necessary steps to transform this raw data into a machine learning data set are described. A multivariate subsequence clustering algorithm that is able to cluster data with the same temporal evolution, is presented. The medical wearable combined with the data analysis pipeline are well suited for a larger clinical study with patients that have a risk of pressure injuries.

Time series analysis has become increasingly important in various domains, and developing effective models relies heavily on high-quality benchmark datasets. Inspired by the success of Natural Language Processing (NLP) benchmark datasets in advancing pre-trained models, we propose a new approach to create a comprehensive benchmark dataset for time series analysis. This paper explores the methodologies used in NLP benchmark dataset creation and adapts them to the unique challenges of time series data. We discuss the process of curating diverse, representative, and challenging time series datasets, highlighting the importance of domain relevance and data complexity. Additionally, we investigate multi-task learning strategies that leverage the benchmark dataset to enhance the performance of time series models. This research contributes to the broader goal of advancing the state-of-the-art in time series modeling by adopting successful strategies from the NLP domain.