Читать книгу Big Data Analytics and Machine Intelligence in Biomedical and Health Informatics - Группа авторов - Страница 20

Medical signals like medical images present volume and velocity challenges, most notably during continuous, high-resolution acquisition and storage from a plethora of monitors connected to each patient. Additionally, the problem of size is posed by physiological signals in that they possess a size/physical dimension in time and space. In order to derive the most useable and appropriate responses from physiological data, an individual must be aware of the circumstances that are affecting the measurements and have continual monitoring to be established in place to assure effective use and robustness, rigorous monitoring of those variables is required.

Currently, healthcare systems rely on a patchwork of disparate and continuous monitoring devices that use single physiological waveform data or discretized vital information to generate alerts in the event of over events [7]. However, such uncomplicated approaches to developing and implementing alarm systems are inherently unreliable, and their sheer volume may result in “alarm fatigue” for care givers and patients alike [8, 9]. In this context, the capacity for new medical knowledge discovery is constrained by prior knowledge that has frequently fallen short of fully exploiting high-dimensional time series data. In [10] Jphan et al. suggested the reason these alarm mechanisms frequently fail is that they rely on isolated sources of information and lack context regarding the patients’ true physiological conditions from a broader and more comprehensive perspective. As a result, improved and more comprehensive approaches to studying interactions and correlations between multimodal clinical time series data are required. This is critical because research consistently demonstrates that humans are unable to reason about changes affecting more than two signals [11, 12].