N

D

C

Q

E

G

H

I

L

K

M

F

P

S

T

W

Y

V

1

1

1

1

1

1

2

2

1

1

2

1

1

1

2

1

1

e K-means cluster analysis algorithm

means cluster analysis algorithm as the simplest partitioning

g algorithm partitions data into clusters through estimating cluster

ased on a pre-defined cluster number [MacQueen, 1967]. Its

ture is the use of the hard membership function to define the

hip between a data point and a cluster. Suppose there are three

The hard membership function for indicating how a data point

o three clusters may be expressed by a vector which is either (0,

(0, 0, 1) or (1, 0, 0). A one entry means the belongingness

hip and a zero entry means the irrelevance relationship.

ting a K-means model is fast due to the use of this hard

hip function.

use of its simplicity, the K-means cluster analysis algorithm has

of the most popular ones for biological/medical pattern discovery

For instance, it has been used to improve quantitative

ility mapping, oxygen extraction fraction as well as cerebral

c rate of oxygen [Cho, et al., 2020]. It has been used to model

ution image data based on healthy subjects as well as ischemic

tients and the result shows that such a model can significantly

analysis robustness. In searching for MRI biomarkers of residue

for clinical impact assessment, it has been used to discover how

utation of cerebral blood volume in postprocessing steps based

ulation of 49 subjects with low- and high-grade gliomas [Bell,

20].

ey point of the K-means algorithm is to find the cluster centres,

means. Once the centres have been found, each data point is

with a membership vector, which is composed of either a zero or

ecause of the use of the least distance, only one entry of the

hip function vector is a one for a data point. The rest entries of