View/Open

munchsistersΤεχνίτη Νοημοσύνη και Ρομποτική

17 Οκτ 2013 (πριν από 3 χρόνια και 7 μήνες)

60 εμφανίσεις

v


CYST AND
TUMOR
CLASSIFICATION O
N HUMAN DENTAL
PANORAMIC IMAGES



Name

: Ingrid Nurtanio




NRP : 2209 301 016



Supervisor

: Prof. Ir. Mauridhi Hery Purnomo, M.Eng.,Ph.D.



Co
-
Supervisor

: Mochamad Hariadi, ST., MSc., Ph.D.


Dr. I Ketut Eddy Purnama, ST., MT.




ABSTRA
CT



Diffe
rentiating cyst and
tumor

lesi
on on human dental panoramic image based
solely on
visual
appearance

is
a
hard task and
this differentiation

real
ly depends
on the knowledge and
experience of the dentist and the analysis result from the
radiologis
t. Wrong determination of les
ion can cause mistreatment of the disease
and increase the risk
to the patient. Therefore, a tool is needed for dentist to help
determine the type of l
es
ion on human jaw.



This research is conducted to classify
cyst
from tumor le
sion on
human dental
panoramic image using texture feature based Support Vector Machin
e (SVM).

A total of 133 dental panoramic images were segmented with active contour
models (Snake) to get the cyst and tumor lesions images.

Thirty
three
texture
based features were extracted from first order statistic method, second order (Gray
Level Co
-
o
ccurence Matrix, GLCM) and
high order method (Gray Level Run
Leng
th Matrix, GLRLM). Six features
were extracted from first ord
er statistic
method;

they are mean,
standard deviation
, smoothness, third moment, uniformity
and entropy. Twenty features were ext
racted from GLCM method, they are
contrast, correlation, energy, homogeneity, and entropy

each with 1 pixel interval
and orien
tation angle of

0

, 45

, 90


and 135

Seven features were extracted
using GLRLM method, they are
Short Runs Emphasis

(SRE),
Long
Runs
Emphasis

(LRE),
Gray Level Non
-
uniformity

(GLN),
Run Percentage

(RP),
Run
Length Non
-
uniformity

(RLN),
Low Gray Level Run Emphasis

(LGRE) and
High
Gray Level Run Emphasis

(HGRE).


The result of this research shows that
SVM classification can be u
sed to
differenti
ate cyst
from tumo
r les
ion on human dental panoramic image with the
highest accuracy rate of 87.18%. Mean accuracy value is 79.85%. Evaluation on
SVM work resulted in AUC value of 0.9444 which shows SVM works as
‘excellent’.


Keywords : cyst, tumo
r, panoramic image,
feature
extraction, classification



vi














Halaman ini sengaja dikosongkan