Multi-task Gaussian Processes for Multivariate Physiological Time-Series Analysis

Type of publication:  Artikel
Zeitschrift: IEEE Transactions on Biomedical Engineering
Band: 62
Nummer: 1
Jahr: 2014
Seiten: 314--322
Notiz: Motion Compensation in Radiosurgery
URL: http://www.robots.ox.ac.uk/~da...
DOI: 10.1109/TBME.2014.2351376
Abriss: Gaussian process (GP) models are a flexible means of performing non-parametric Bayesian regression. However, GP models in healthcare are often only used to model a single univariate output time-series, denoted as single-task GPs (STGP). Due to an increasing prevalence of sensors in healthcare settings, there is an urgent need for robust multivariate time-series tools. Here, we propose a new method using multi-task GPs (MTGPs) which can model multiple correlated multivariate physiological time-series simultaneously. The flexible MTGP framework can learn the correlation within multiple signals even though they might be sampled at different frequencies and have training sets available for different intervals. Furthermore, prior knowledge of any relationship between the time-series such as delays and temporal behaviour can be easily integrated. A novel normalisation is proposed to allow interpretation of the various hyperparameters used in the MTGP. We investigate MTGPs for physiological monitoring with synthetic datasets and two real-world problems from the field of patient monitoring and radiotherapy. The results are compared to standard Gaussian processes and other existing methods in the respective biomedical application areas. In both cases, we show that our proposed framework learned the correlation between physiological timeseries efficiently, outperforming the existing state-of-the-art.
Schlagworte:
Autoren: Dürichen, Robert
Pimentel, Marco A. F.
Clifton, Lei
Schweikard, Achim
Clifton, David A.
Anhänge
  • 2014 - TBME - Dürichen - Multi...
 [Bib|RIS]