Conclusions and Contributions - Prediction and Classification of Respiratory Motion - page 130

Chapter 7
Conclusions and Contributions
7.1 Conclusions
The following conclusions can be made from the results obtained from Chap. 4 :
7.1.1 Hybrid Implementation of Extended Kalman Filter
• RNN executes in the supervised-training part of the prediction, whereas EKF
executes in the part of the correction with predicted or filtered estimation.
• The coupling technique using multiple sensory channel inputs can be used to
compensate the computational accuracy.
• Fisher linear discriminant on the discriminant analysis can decide the optimized
neuron number for RMLP in the given samples.
• The average percentage of prediction overshoot for HEKF is 3.72 %, whereas
the average percentage of prediction overshoot for DEKF is 18.61 %.
• The proposed HEKF showed the better NRMSE performance across all variable
prediction interval times.
• HEKF method needs more time comparing to DEKF because of the calculation
of the coupling matrix and the separate neural network for channel number.
The following conclusions can be made from the results obtained from Chap. 5 :
7.1.2 Customized Prediction of Respiratory Motion
with Clustering
• For the preprocedure of prediction for individual patient, we construct the
clustering (five classes) based on breathing patterns of multiple patients using