Irregular Breathing Classification from Multiple Patient Datasets - Prediction and Classification of Respiratory Motion - page 110

Table 6.1 shows the feature extraction metrics for the breathing pattern clas-
sification. We create Table 6.1 based on previous entities for breathing features, so
that the table can be variable. The feature extraction metrics can be derived from
multiple patient datasets with the corresponding formula. To establish feature
metrics for breathing pattern classification, we define the candidate feature com-
bination vector (x) from the combination of feature extraction metrics in Table 6.1 .
We defined 10 feature extraction metrics in Table 6.1 . The objective of this section
is to find out the estimated feature metrics (x) from the candidate feature combi-
nation vector (x) using discriminant criterion based on clustered degree. We can
define the candidate feature combination vector as x ¼ x 1 ; ... ; x ð Þ , where variable
z is the element number of feature combination vector, and each element corre-
sponds to each of the feature extraction metrics depicted in Table 6.1 .
For example, the feature combination vector may be defined as x ¼ x 1 ; x 2 ; x ð Þ if
the feature combination vector has feature extraction of BRF, PCA, and MLR.
The total number (K) of feature combination vector using feature extraction
metrics can be expressed as follows:

K ¼ X

10

C 10 ; z ð Þ; ð 6 : 3 Þ

z ¼ 2

where, the combination function C(10, z) is the number of ways of choosing
z objects from ten feature metrics. For the intermediate step, we may select which
features to use for breathing pattern classification with the feature combination
vectors, i.e., the estimated feature metrics (x). For the efficient and accurate
classification of breathing patterns, selection of relevant features is important [ 24 ].
In this study, the discriminant criterion based on clustered degree can be used to

Table 6.1 Feature extraction metrics including the formula and notation

Name Formula

R xx ðÞ¼ 2T R T T x ðÞ xt þ s

AMV

max R xx ½ ; ð Þ dt (T: period of observation)

ADT arg max

s

R xx ðÞ arg min
s

R xx ðÞ

var Dx Dt 2

(x : observed breathing data)

ACC

VEL var Dx = Dt ð Þ

BRF mean 1 = BC i ð Þ BC i : ith breathing cycle range

FTP

X ð k Þ¼ P

N

ðÞ N : vectors of length N ; 1 k N

x ð n Þ e j2p ð k 1 Þ n 1 maxX ; ð Þ

n ¼ 1

PCA Y = PrinComp(X) (PrinComp ðÞ :PCA function,

ð

N M ; M ¼ 3

Þ , Y: coefficient matrix ð

M M

Þ

X: data matrix

1 Z T y (Z: predictor, y: observed response)

MLR

Z T Z

s

1

STD

N P

N

Þ 2 s N ¼ ð x i x

i ¼ 1

MLE

' ¼ N P

N

h mle ¼ arg max
h 2 H

' h j x 1 ; ... ; x N

ð Þ; ln fx i jðÞ f h jð ): Normal distribution

i ¼ 1