RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación
Artículo científico (JournalArticle)
Detection of correct and incorrect measurements in real-time continuous glucose monitoring systems applying a postprocessing support vector machine
[EN] Support vector machines (SVMs) are an attractive option for detecting correct and incorrect measurements in real-time continuous glucose monitoring systems (RTCGMSs), because their learning mechanism can introduce a postprocessing strategy for imbalanced datasets. The proposed SVM considers the geometric mean to obtain a more balanced performance between sensitivity and specificity. To test this approach, 23 critically ill patients receiving insulin therapy were monitored over 72 h using an RTCGMS, and a dataset of 537 samples, classified according to International Standards Organization (ISO) criteria (372 correct and 165 incorrect measurements), was obtained. The results obtained were promising for patients with septic shock or with sepsis, for which the proposed system can be considered as reliable. However, this approach cannot be considered suitable for patients without sepsis., This work was supported by the Spanish Ministry of Science and Innovation under Grant DPI-2010-20764-C02, and by the Autonomous Government of Catalonia under Grant 2009 SGR 523. In addition, this work was partially supported by the Spanish Ministry of Economy and Competitiveness through ARTEMISA (TIN2009-14378-C02-01), and by the Andalusia Regional Ministry of Economy, Innovation and Science through Simon (TIC-8052). Y. Leal was supported by BR Grants of the University of Girona.