A Location-Mixture Autoregressive Model for Online Forecasting of Lung Tumor Motion

• Activity Number: 506
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Daniel Cervone*+ and Natesh S. Pillai and Debdeep Pati and John Henry Lewis and Ross Berbeco
• Companies: Harvard and Harvard and Florida State University and Dana-Farber Cancer Institute/Harvard Medical School and Brigham & Women's Hospital/Harvard Medical School
• Keywords: lung tumor tracking ; nonlinear time series ; mixture autoregressive process ; time series motifs

Abstract

Lung tumor tracking for radiotherapy requires real-time, multiple-step ahead forecasting of a quasi-periodic time series recording instantaneous tumor locations. We introduce a location-mixture autoregressive (LMAR) process, a novel time series model that admits multimodal conditional distributions, fast approximate inference using the Expectation-Maximization algorithm, and accurate multiple-step ahead predictive distributions. LMAR outperforms several commonly used methods in terms of out-of-sample prediction accuracy using clinical data from lung tumor patients. With its superior predictive performance and real-time computation, the LMAR model could be effectively implemented for use in current tumor tracking systems.