The Annals of Applied Statistics

A location-mixture autoregressive model for online forecasting of lung tumor motion

Daniel Cervone, Natesh S. Pillai, Debdeep Pati, Ross Berbeco, and John Henry Lewis

Full-text: Open access


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 that admits multimodal conditional distributions, fast approximate inference using the EM 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.

Article information

Ann. Appl. Stat., Volume 8, Number 3 (2014), 1341-1371.

First available in Project Euclid: 23 October 2014

Permanent link to this document

Digital Object Identifier

Mathematical Reviews number (MathSciNet)

Zentralblatt MATH identifier

Lung tumor tracking external beam radiotherapy nonlinear time series mixture autoregressive process time series motifs likelihood approximation multiple-step prediction


Cervone, Daniel; Pillai, Natesh S.; Pati, Debdeep; Berbeco, Ross; Lewis, John Henry. A location-mixture autoregressive model for online forecasting of lung tumor motion. Ann. Appl. Stat. 8 (2014), no. 3, 1341--1371. doi:10.1214/14-AOAS744.

Export citation


  • Berbeco, R. I., Nishioka, S., Shirato, H., Chen, G. T. and Jiang, S. B. (2005). Residual motion of lung tumours in gated radiotherapy with external respiratory surrogates. Phys. Med. Biol. 50 3655–3667.
  • Boots, B. and Gordon, G. J. (2011). An online spectral learning algorithm for partially observable nonlinear dynamical systems. In Proceedings of the 25th International Conference on Artificial Intelligence 293–300. AAAI Press, San Francisco, CA.
  • Boshnakov, G. N. (2009). Analytic expressions for predictive distributions in mixture autoregressive models. Statist. Probab. Lett. 79 1704–1709.
  • Carvalho, A. X. and Tanner, M. A. (2005). Mixtures-of-experts of autoregressive time series: Asymptotic normality and model specification. IEEE Trans. Neural Netw. 16 39–56.
  • Cervone, D., Pillai, N., Pati, D., Berbeco, R. and Lewis, J. H. (2014). Supplement to “A location-mixture autoregressive model for online forecasting of lung tumor motion.” DOI:10.1214/14-AOAS744SUPP.
  • D’Souza, W. D., Naqvi, S. A. and Yu, C. X. (2005). Real-time intra-fraction-motion tracking using the treatment couch: A feasibility study. Phys. Med. Biol. 50 4021–4033.
  • Dempster, A. P., Laird, N. M. and Rubin, D. B. (1977). Maximum likelihood from incomplete data via the EM algorithm. J. R. Stat. Soc. Ser. B 39 1–38.
  • De Gooijer, J. G. and Kumar, K. (1992). Some recent developments in non-linear time series modelling, testing, and forecasting. Int. J. Forecast. 8 135–156.
  • Ernst, F., Schlaefer, A. and Schweikard, A. (2007). Prediction of respiratory motion with wavelet-based multiscale autoregression. In Medical Image Computing and Computer-Assisted Intervention (N. Ayache, S. Ourselin and A. Maeder, eds.) 668–675. Springer, New York.
  • Ernst, F. and Schweikard, A. (2009). Forecasting respiratory motion with accurate online support vector regression (SVRpred). Int. J. Comput. Assisted Radiol. Surg. 4 439–447.
  • Ernst, F., Dürichen, R., Schlaefer, A. and Schweikard, A. (2013). Evaluating and comparing algorithms for respiratory motion prediction. Phys. Med. Biol. 58 3911–3929.
  • Fu, T.-c. (2011). A review on time series data mining. Engineering Applications of Artificial Intelligence 24 164–181.
  • Gneiting, T., Balabdaoui, F. and Raftery, A. E. (2007). Probabilistic forecasts, calibration and sharpness. J. R. Stat. Soc. Ser. B Stat. Methodol. 69 243–268.
  • Goerg, G. M. (2013a). Forecastable component analysis. In Proceedings of the 30th International Conference on Machine Learning 64–72. JMLR W&CP.
  • Goerg, G. M. (2013b). LICORS: Light cone reconstruction of states—predictive state estimation from spatio-temporal data. R package version 0.2.0.
  • Goerg, G. M. and Shalizi, C. R. (2012). LICORS: Light cone reconstruction of states for non-parametric forecasting of spatio-temporal systems. Technical report, Statistics Dept., Carnegie Mellon Univ., Pittsburgh, PA.
  • Goerg, G. M. and Shalizi, C. R. (2013). Mixed LICORS: A nonparametric algorithm for predictive state reconstruction. In Proceedings of the 16th International Conference on Artificial Intelligence and Statistics 289–297. JMLR W&CP.
  • Hamilton, J. D. (1989). A new approach to the economic analysis of nonstationary time series and the business cycle. Econometrica 57 357–384.
  • Hoerl, A. E. and Kennard, R. W. (1970). Ridge regression: Biased estimation for nonorthogonal problems. Technometrics 12 55–67.
  • Hornik, K., Stinchcombe, M. and White, H. (1989). Multilayer feedforward networks are universal approximators. Neural Netw. 2 359–366.
  • Kalbfleisch, J. D. and Sprott, D. A. (1970). Application of likelihood methods to models involving large numbers of parameters. J. Roy. Statist. Soc. Ser. B 32 175–208.
  • Kalet, A., Sandison, G., Wu, H. and Schmitz, R. (2010). A state-based probabilistic model for tumor respiratory motion prediction. Phys. Med. Biol. 55 7615–7631.
  • Krauss, A., Nill, S. and Oelfke, U. (2011). The comparative performance of four respiratory motion predictors for real-time tumour tracking. Phys. Med. Biol. 56 5303–5317.
  • Krolzig, H.-M. (2000). Predicting Markov-switching vector autoregressive processes. Technical report, Economics Dept., Oxford Univ.
  • Lau, J. W. and So, M. K. P. (2008). Bayesian mixture of autoregressive models. Comput. Statist. Data Anal. 53 38–60.
  • Le, N. D., Martin, R. D. and Raftery, A. E. (1996). Modeling flat stretches, bursts, and outliers in time series using mixture transition distribution models. J. Amer. Statist. Assoc. 91 1504–1515.
  • Lin, J., Lonardi, S., Keogh, E. and Patel, P. (2002). Finding motifs in time series. In Proceedings of the 2nd Workshop on Temporal Data Mining 53–68. ACM, New York.
  • Littman, M. L., Sutton, R. S. and Singh, S. P. (2002). Predictive representations of state. In Proceedings of Advances in Neural Information Processing Systems 14 1555–1561. MIT Press, Cambridge, MA.
  • McCulloch, R. E. and Tsay, R. S. (1994). Statistical analysis of economic time series via Markov switching models. J. Time Series Anal. 15 523–539.
  • Murphy, M. J. and Dieterich, S. (2006). Comparative performance of linear and nonlinear neural networks to predict irregular breathing. Phys. Med. Biol. 51 5903–5914.
  • Murphy, M. J., Isaakson, M. and Jalden, J. (2002). Adaptive filtering to predict lung tumor motion during free breathing. In Computer Assisted Radiology and Surgery (H. U. Lemke, M. W. Vannier, K. Inamura, A. G. Farman, K. Doi and J. H. C. Reiber, eds.) 539–544. Springer, New York.
  • Riaz, N., Shanker, P., Wiersma, R., Gudmundsson, O., Mao, W., Widrow, B. and Xing, L. (2009). Predicting respiratory tumor motion with multi-dimensional adaptive filters and support vector regression. Phys. Med. Biol. 54 5735–5748.
  • Rottmann, J., Keall, P. and Berbeco, R. (2013). Markerless EPID image guided dynamic multi-leaf collimator tracking for lung tumors. Phys. Med. Biol. 58 4195–4204.
  • Ruan, D. and Keall, P. (2010). Online prediction of respiratory motion: Multidimensional processing with low-dimensional feature learning. Phys. Med. Biol. 55 3011–3025.
  • Schweikard, A., Glosser, G., Bodduluri, M., Murphy, M. J. and Adler, J. R. (2000). Robotic motion compensation for respiratory movement during radiosurgery. Comput. Aided Surg. 5 263–277.
  • Shalizi, C. R. (2003). Optimal nonlinear prediction of random fields on networks. In Discrete Mathematics and Theoretical Computer Science Proceedings (volume AB) 11–30. DMTCS, Nancy, France.
  • Sharp, G. C., Jiang, S. B., Shimizu, S. and Shirato, H. (2004). Prediction of respiratory tumour motion for real-time image-guided radiotherapy. Phys. Med. Biol. 49 425–440.
  • Tanaka, Y., Iwamoto, K. and Uehara, K. (2005). Discovery of time-series motif from multi-dimensional data based on MDL principle. Mach. Learn. 58 269–300.
  • Tibshirani, R. (1996). A comparison of some error estimates for neural network models. Neural Comput. 8 152–163.
  • Tong, H. (1978). On a threshold model. In Pattern Recognition and Signal Processing (C. H. Chen, ed.) 575–586. Sijthoff & Noordhoff, Amsterdam.
  • Tong, H. (1990). Nonlinear Time Series: A Dynamical System Approach. Oxford Univ. Press, New York.
  • Tong, H. and Lim, K. S. (1980). Threshold autoregression, limit cycles and cyclical data. J. R. Stat. Soc. Ser. B Stat. Methodol. 42 245–292.
  • Tong, H. and Moeanaddin, R. (1988). On multi-step non-linear least squares prediction. J. R. Stat. Soc. Ser. D (Statistician) 37 101–110.
  • Venables, W. N. and Ripley, B. D. (2002). Modern Applied Statistics with $S$, 4th ed. Springer, New York.
  • Wong, C. S. and Li, W. K. (2000). On a mixture autoregressive model. J. R. Stat. Soc. Ser. B Stat. Methodol. 62 95–115.
  • Ye, L. and Keogh, E. (2009). Time series shapelets: A new primitive for data mining. In Proceedings of the 15th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining 947–956. ACM, New York.

Supplemental materials