Advertisement

Predicting respiratory motion signals for image-guided radiotherapy using multi-step linear methods (MULIN)

  • Floris Ernst
  • Achim Schweikard
Original Article

Abstract

Objective

Forecasting of respiration motion in image-guided radiotherapy requires algorithms that can accurately and efficiently predict target location. Improved methods for respiratory motion forecasting were developed and tested.

Materials and methods

MULIN, a new family of prediction algorithms based on linear expansions of the prediction error, was developed and tested. Computer-generated data with a prediction horizon of 150 ms was used for testing in simulation experiments. MULIN was compared to Least Mean Squares-based predictors (LMS; normalized LMS, nLMS; wavelet-based multiscale autoregression, wLMS) and a multi-frequency Extended Kalman Filter (EKF) approach. The in vivo performance of the algorithms was tested on data sets of patients who underwent radiotherapy.

Results

The new MULIN methods are highly competitive, outperforming the LMS and the EKF prediction algorithms in real-world settings and performing similarly to optimized nLMS and wLMS prediction algorithms. On simulated, periodic data the MULIN algorithms are outperformed only by the EKF approach due to its inherent advantage in predicting periodic signals. In the presence of noise, the MULIN methods significantly outperform all other algorithms.

Conclusion

The MULIN family of algorithms is a feasible tool for the prediction of respiratory motion, performing as well as or better than conventional algorithms while requiring significantly lower computational complexity. The MULIN algorithms are of special importance wherever high-speed prediction is required.

Keywords

Respiration Forecasting Algorithms Radiosurgery 

References

  1. 1.
    Ernst F, Schlaefer A, Schweikard A (2007) Prediction of respiratory motion with wavelet-based multiscale autoregression, MICCAI 2007, Part II. In: Ayache N, Ourselin S, Maeder A(eds) Lecture Notes in Computer Science, vol. 4792 MICCAI. Springer, Berlin, pp 668–675Google Scholar
  2. 2.
    Haykin S (2002) Adaptive filter theory. Prentice-Hall, Englewood CliffsGoogle Scholar
  3. 3.
    Kalman RE (1960) A new approach to linear filtering and prediction problems. Trans ASME J Basic Eng 82(Series D): 35–45Google Scholar
  4. 4.
    Kalman RE, Bucy RS (1961) New results in linear filtering and prediction theory. Trans ASME J Basic Eng 83(Series D): 95–108Google Scholar
  5. 5.
    Ramrath L, Schlaefer A, Ernst F, Dieterich S, Schweikard A (2007) Prediction of respiratory motion with a multi-frequency based Extended Kalman Filter. In: Proceedings of the 21st international conference and exhibition on computer assisted radiology and surgery (CARS’07) (Berlin, Germany), vol 21, CARSGoogle Scholar
  6. 6.
    Schweikard A, Glosser G, Bodduluri M, Murphy MJ, Adler JR (2000) Robotic motion compensation for respiratory motion during radiosurgery. J Comput Aided Surg 5(4): 263–277CrossRefGoogle Scholar
  7. 7.
    Schweikard A, Murphy MJ, Hancock SL (1998) Image-guided stereotactic radiosurgery: the CyberKnife. Image Guid Neurosurg Clin Appl Interact Surg Navig 16: 193–204Google Scholar
  8. 8.
    Douglas SC (1994) A family of normalized LMS algorithms. IEEE Signal Process Lett 1(3): 49–55CrossRefGoogle Scholar
  9. 9.
    Sharp GC, Jiang SB, Shimizu S, Shirato H (2004) Prediction of respiratory tumour motion for real-time image-guided radiotherapy. Phys Med Biol 49: 425–440PubMedCrossRefGoogle Scholar
  10. 10.
    Urschel HC, Kresl JJ, Luketich JD, Timmermann RD (2007) Robotic radiosurgery. Treating tumors that move with respiration, 1st edn. Springer, BerlinCrossRefGoogle Scholar

Copyright information

© CARS 2008

Authors and Affiliations

  1. 1.University of Lübeck, Institute for Robotics and Cognitive SystemsLübeckGermany

Personalised recommendations