Patient Prognosis from Vital Sign Time Series: Combining Convolutional Neural Networks with a Dynamical Systems Approach

• Type of Material: Conference Paper
• Authors: Li-wei Lehman, Massachusetts Institute of Technology; Mohammad Ghassemi, Massachusetts Institute of Technology; Jasper Snoek, Harvard University; Shamim Nemati, Emory University
• Language: English
• Date: 2015-01-01
• Publisher: Emory University Libraries
• Publication Version: Author Accepted Manuscript (After Peer Review)
• Copyright Statement: © 2015, IEEE
• Final Published Version (URL): http://dx.doi.org/10.1109/CIC.2015.7411099
• Title of Journal or Parent Work: Computing in Cardiology
• Conference or Event Name: Computing in Cardiology Conference (CinC), 2015
• Volume: 42
• Start Page: 1069
• End Page: 1072
• Grant/Funding Information: This work was supported by the National Institutes of Health (NIH) grant R01-EB001659 and R01GM104987 from the National Institute of Biomedical Imaging and Bioengineering (NIBIB), the James S. McDonnell Foundation Postdoctoral grant.
• Keywords: Engineering, Multidisciplinary; Technology; Computer Science, Interdisciplinary Applications; Engineering, Biomedical; Computer Science; Science & Technology; Engineering
• Research Categories: Computer Science; Engineering, Biomedical

Abstract

In this work, we propose a stacked switching vector-autoregressive (SVAR)-CNN architecture to model the changing dynamics in physiological time series for patient prognosis. The SVAR-layer extracts dynamical features (or modes) from the time-series, which are then fed into the CNN-layer to extract higher-level features representative of transition patterns among the dynamical modes. We evaluate our approach using 8-hours of minute-by-minute mean arterial blood pressure (BP) from over 450 patients in the MIMIC-II database. We modeled the time-series using a third-order SVAR process with 20 modes, resulting in first-level dynamical features of size 20×480 per patient. A fully connected CNN is then used to learn hierarchical features from these inputs, and to predict hospital mortality. The combined CNN/SVAR approach using BP time-series achieved a median and interquartile-range AUC of 0.74 [0.69, 0.75], significantly outperforming CNN-alone (0.54 [0.46, 0.59]), and SVAR-alone with logistic regression (0.69 [0.65, 0.72]). Our results indicate that including an SVAR layer improves the ability of CNNs to classify nonlinear and nonstationary time-series.