Irregular Breathing Classification from Multiple Patient Datasets - Prediction and Classification of Respiratory Motion - page 114

X

K

;

1
b m K

Þ T R m ¼ Im j x i ð Þ x i M m ð Þ x i M m ð ð 6 : 15 Þ

i ¼ 1

where Imx i jð Þ¼ 0, M m is the

mean value of the classified feature vectors x ðÞ in class m, and K is the total
number of the patient datasets. To decide the reference value to detect the irregular
breathing pattern, we combine the class means ( 6.14 ) and the covariance ( 6.15 )
with the probability ( 6.13 ) for each class as follows:

ð j Þ¼ 1ifx i is classified into class m; otherwise Imx i

X

b m X
m

M

m ¼ M P

P ¼ 1
M

M

: ð 6 : 16 Þ

b m m m ;

m ¼ 1

m ¼ 1

With Eq. ( 6.16 ), we can make the threshold value (n m ) to detect the irregular
breathing pattern in Eq. ( 6.17 ), as follows:

Þ

p

R

n m ¼ m m m

ð

; ð 6 : 17 Þ

L m

where L m is the total number of breathing data in class m. For each patient i in

class m, we define P m as a subset of the patient whose score d i is within the
threshold value (n m ) in class m and 1 - P m as a subset of the patient whose score

d i is greater than the threshold value (n m ) in class m, as shown in Fig. 6.3 .
The digit ''1'' represents the entire patient set for class m in Fig. 6.3 . With
Fig. 6.3 we can detect the irregular breathing patterns in the given class m with the

P 1

Regular
(Yellow Area)

P M

P m

Irregular
(Gray Area)

< Class 1 > < Class m > < Class M >

P 1 P m P M

ξ m ξ M

ξ 1

1− P 1 1− P m 1− P M

Fig. 6.3 Detection of regular/irregular patterns using the threshold value (n m )