Customized Prediction of Respiratory Motion - Prediction and Classification of Respiratory Motion - page 88

Chapter 5
Customized Prediction of Respiratory
Motion
Accurate prediction of the respiratory motion would be beneficial to the treatment of
thoracic and abdominal tumors. However, a wide variety of breathing patterns can
make it difficult to predict the breathing motion with explicit models. We proposed a
respiratory motion predictor, i.e., customized prediction with multiple patient inter-
actions using neural network (CNN). For the preprocedure of prediction for individual
patient, we construct the clustering based on breathing patterns of multiple patients
using the feature selection metrics that are composed of a variety of breathing features.
In the intraprocedure, the proposed CNN used neural networks (NN) for a part of the
prediction and the Extended Kalman filter (EKF) for a part of the correction. The
prediction accuracy of the proposed method was investigated with a variety of pre-
diction time horizons using Normalized root mean squared error (NRMSE) values in
comparison with the alternate Recurrent neural network (RNN). We have also
evaluated the prediction accuracy using the marginal value that can be used as the
reference value to judge how many signals lie outside the confidence level. The
experimental results showed that the proposed CNN can outperform RNN with
respect to the prediction accuracy with an improvement of 50 %.
5.1 Introduction
Current developments in radiotherapy systems open a new era for treatment with
accurate dosimetry of thoracic and abdominal tumors [ 1 - 3 ]. Effective radiation
treatment requires motion compensation for uncertainty and irregularity originating
from systematic or random physiological phenomena [ 4 , 5 ]. Respiratory motion
severely affects precise radiation dose delivery because thoracic and abdominal
tumors may change locations by as much as three centimeters during radiation
treatment [ 6 ]. In patients with a wide range of respiratory motion, radiation treatment
can be delivered by dynamic gating, where radiation is activated only when the
respiratory motion is within a predefined amplitude or phase level [ 6 , 7 ].
In addition to the respiratory motion, system latency attributable to hardware
limitations and software processing time may affect the accurate radiation delivery