Customized Prediction of Respiratory Motion - Prediction and Classification of Respiratory Motion - page 102

breathing features. After this clustering, appropriate parameter selections with
respect to each class—e.g., optimal neuron number for the prediction process of
the neural network and/or interactive (coupling) degree for the multiple breathing
information and so forth—can improve the prediction accuracy in comparison to
the previous prediction method, because the multiple respiratory information does
not have identical relationships, but relationships that closely resemble one
another. That means that when the system for respiratory prediction considers the
breathing patterns of multiple patients, it can yield a more accurate prediction
performance than when it does not.
For the evaluation criteria of prediction, we showed NRMSE (which is a nor-
malized error value between the predicted and actual signal over all the samples),
and prediction overshoot as the reference value to judge how many signals lie
outside the confidence level. Our experimental results reveal that the proposed
CNN needs more computational time to process due to the abundant breathing
information and the additional signal processing and correction process for each
RMLP. The proposed CNN, however, can improve NRMSE values by 50 % in
contrast to the RNN. Moreover, the proposed CNN decreases the number of
average prediction overshoot values by 8.37 %, whereas the RNN generates pre-
diction overshoot values in more than 40 % over all the patients.
References
1. P.J. Keall, G.S. Mageras, J.M. Balter, R.S. Emery, K.M. Forster, S.B. Jiang, J.M. Kapatoes,
D.A. Low, M.J. Murphy, B.R. Murray, C.R. Ramsey, M.B. Van Herk, S.S. Vedam, J.W.
Wong, E. Yorke, The management of respiratory motion in radiation oncology report of
AAPM Task Group 76. Med. Phys. 33(10), 3874-3900 (2006)
2. K. Bush, I.M. Gagne, S. Zavgorodni, W. Ansbacher, W. Beckham, Dosimetric validation of
Acuros XB with Monte Carlo methods for photon dose calculations. Med. Phys. 38(4),
2208-2221 (2011)
3. L.I. Cervino, J. Du, S.B. Jiang, MRI-guided tumor tracking in lung cancer radiotherapy. Phy.
Med. Biol. 56(13), 3773-3785 (2011)
4. K. Nakagawa, K. Yoda, Y. Masutani, K. Sasaki, K. Ohtomo, A rod matrix compensator for
small-field intensity modulated radiation therapy: a preliminary phantom study. IEEE Trans.
Biomed. Eng. 54(5), 943-946 (2007)
5. T. Mu, T.C. Pataky, A.H. Findlow, M.S.H. Aung, J.Y. Goulermas, Automated nonlinear
feature generation and classification of foot pressure lesions. IEEE Trans. Inf Technol.
Biomed. 14(2), 418-424 (2010)
6. R.I. Berbeco, S. Nishioka, H. Shirato, G.T.Y. Chen, S.B. Jiang, Residual motion of lung
tumours in gated radiotherapy with external respiratory surrogates. Phys. Med. Biol. 50(16),
3655-3667 (2005)
7. E.W. Pepina, H. Wu, H. Shirato, Dynamic gating window for compensation of baseline shift
in respiratory-gated radiation therapy. Med. Phys. 38(4), 1912-1918 (2011)
8. P.R. Poulsenb, B. Cho, A. Sawant, D. Ruan, P.J. Keall, Detailed analysis of latencies in
image-based dynamic MLC tracking. Med. Phys. 37(9), 4998-5005 (2010)
9. T. Roland, P. Mavroidis, C. Shi, N. Papanikolaou, Incorporating system latency associated
with real-time target tracking radiotherapy in the dose prediction step. Phys. Med. Biol.
55(9), 2651-2668 (2010)