LAUSR

Extravasation is a common hazard in intravenous (IV) therapy. However, in clinical practice, the nursing staff is responsible for checking the injection status regularly, which leads to the risk of delayed treatment of the extravasation of IV therapy. In this study, we propose an open-source solution to detect the early signs of extravasation—including the wearable sensor to collect physical signals from the skin, the simulation platform to simulate extravasation on artificial skin, and the data server to collect and analyze data of occurring extravasation. Pressure, body temperature, and optical sensors were integrated into the wearable sensor and evaluated to understand their effectiveness in detection. We also propose the Light-ConvLSTM model that can predict extravasation with a comprehensive evaluation and show the advantages and feasibility of the proposed concept. The results show that the proposed design with Light-ConvLSTM can achieve significant performance: the extravasation detection rate can be raised to 83.7%, while the false alarm rate is only 6.2%.