Health status detection for patients in physiological monitoring

A primary difficulty in physiological monitoring is detecting changes of health status for patients. In order to address this difficulty, we propose a new framework in patient-specific physiological monitoring by defining a density ratio using the training density and testing density to denote the changes of patient status, such as health, sub-health and abnormalities. We use a Least Square-based algorithm to estimate density ratio parameters without involving density estimation. For verifying the availability and efficacy of the proposed framework, we apply our approach to physiological monitoring data (11901 beats) from the Physionet database to do the pilot experiments. Results demonstrate that the approach is effective in detecting the patient status.