Appendix A - Prediction and Classification of Respiratory Motion - page 138

x Estimated feature metrics

K Total number of feature combination vector

z Element number of feature extraction metrics

C(10, z) Combination function for the number of selecting z objects from

ten feature metrics

G Total number of class in the given datasets

px ; H ð Þ

g m ¼ 1

Joint probability density with H a m ; l m ; R m f

a m Prior probability

M Number of finite mixture model (Cluster number)

K Number of patient datasets

L(•) Objective function to maximize the log-likelihood function

e m

Classified feature vectors of class m

o m Reconstructed feature vectors with NN

d m Reconstruction error

b m Probability of class m

m m Means of class m

R m Covariance of class m

Mean value of the classified feature vectors (x m ) in class m M m

Imx i

ð Þ

Generalized function depending on x i , where I(m|x i )=1ifx i is
classified into class m; otherwise I(m|x i )=0

j

m Averaged class mean with the probability for each class

R

Averaged covariance with the probability for each class

n m Threshold value to detect the irregular breathing pattern

L m Total number of breathing data in class m

P m Subset of the patient whose score is within n m in class m

T i Observation period of the patient i

BC i Breathing cycle range of the patient i

w i Number of irregular breathing pattern region of the patient i

R i TP True positive range within the observation period (T i )

R i TN True negative range within the observation period (T i )

Ratio of the R i TN c i to the T i for the patient i ð 0 c i 1 Þ

W th Regular threshold to decide whether patient i is regular or not