Patient physiological monitoring with machine learning

The task of discovering novel medical knowledge from complex, large-scale and high-dimensional patient data, collected during care episodes, is central to innovation in medicine. The recognition of complex trajectories in multivariate time-series data requires effective models and representations for the analysis and matching of functional data. In this chapter, we describe a method based on Gaussian processes for exploratory data analysis using the observational physiological time-series data. The method focuses on a representation of unevenly sampled trajectories that allows for revealing physiological recovery patterns in a database of vital signs acquired from post-operative patients. While our primary motivation comes from clinical data, this approach may be applicable to other time-series domains.We first describe methods that have been proposed in the literature for the same purpose. We then provide a brief summary of Gaussian processes, and describe our proposed approach for performing "clustering" of patients' trajectories.