Documente Academic
Documente Profesional
Documente Cultură
PG Scholar
Department of Computer Science and Engineering
G. H. Raisoni College of Engineering,
Nagpur (MS), India
prachikawalkar5@gmail.com
Abstract- In the last few decades, a dynamic growth within the
range of analysis works conducted within the space of organ
structure designation. This paper gives short reviews computer
assisted histopathology image analysis for gland detection,
segmentation and classification. The term Histopathology refers
to the study of changes in biopsy sample taken by a pathologist
under microscope. Main task of pathologist is to analyzing,
locating and classifying most of the diseases, similarly appear at
the tissue structure, distribution of cells in tissue, regularities of
cell shapes and determine benign and syndrome in image. It is
very important because the gland in human tissues is the area
where cancer can be experiential. But this process is too time
consuming and lead to intra and inter observer variability. To
remove this drawback automatic detection of images is needed
for quantitative diagnosis of Tissue. In this paper we have
consolidated such recent techniques and its unique features. The
survey done provides different approaches for detecting glands
and the parameters considered for same. Discussion is also made
on various Databases used for detection of glandular structure.
Keywords Cancer, Gleason Grading System, Graph Theory,
Medical Diagnosis, Texture Analysis
I. INTRODUCTION
Image processing [1] is a very big area in engineering
branch and successfully implemented on biomedical as well as
fundamentally can be viewed as a solo block in which image
or video as input and gives output either an image or a video
or set of parameters related with the images. Image processing
used in various field and one of the important field where
researchers focus now-a-days is medical image processing. It
saves time of pathologist by providing an automatic computer
aided diagnosis (CAD) to diagnosis problems. There are
hundreds of cases came to pathologist every day, with the help
of CAD its really time saving and also gives accurate
judgment than classical eye screening evaluation. CADS
include much basic functionality like image or picture
acquisition, image pre-processing, image segmentation, image
detection etc.
The recent investigate prove that the cancer is next reason
of fatality after cardiovascular diseases. And it caused just
978-1-4799-6272-3/15/$31.00(c)2015 IEEE
Girish Talmale
Assistant Professor
Department of Computer Science and Engineering
G. H. Raisoni College of Engineering,
Nagpur (MS), India
girishtalmale@gmail.com
because of the abnormal growth of glands in human tissues. If
early detection of malignancy glands found then there is a key
to survival. As a result, the development of a totally automatic
cancer designation scheme sustaining someone professional
has become associate convoluted duty. An analysis completed
by a diagnostician is commonly supported the subjective
judgment on some factors like the scale of nuclei, nucleoli and
chromatin granule clumps, roughness of chromatin granule
facade, form and efficiency of nuclear structures, and so on.
Numerous institution of higher education centers and
analysis laboratories are targeted on this issue as a result of
this remains a difficult drawback because the discrepancy in
fixation, coloring and sectioning measures cause a substantial
quantity of artifacts and variances in tissue sections, which can
outcome in massive variances in appearances. Gland is a
complex and closed structure present in the tissue. The major
quality of the segmentation method owes to the wholeness and
uncertainty of the knowledge restricted within the image.
While these days several biological sciences comes assume
full automation and time period method with a high degree of
competency, a way freed from drawbacks of the currently
legendary approaches has got to be created. In initial stage, if
malady is suspected, tissue samples are in use from the patient
and viewed beneath the magnifier by a specialist. In case, the
standard architecture of tissue pattern seems to be different
from sample taken patient-tissue then disease recognized by
pathologist. Sample of tissue sectioned in different ways in
pathology lab so the detection of accurate pixel structure of
gland in tissue seems to be very difficult. In tissue, cells
having complex anatomical units called gland. Various
protective and physiological function performed by tubular
structure of gland. In several sickness states the glands are
interrupted, usually during a characteristic style. If machinecontrolled image analysis is to be accustomed create simple
tissue diagnosing, then recognition of glands is crucial. Up till
now several detection ways are projected.
There are many automatic gland segmentation and detection
systems which are facing problems of irregular gland shape,
size and different way of tissue sectioned during sampling.
The creation of an absolutely automatic detection of glandular
b.
Though this two techniques more popular but up till now most
of the research done on H & E stained images [3, 4, 5, 6, 7, 8].
Very few research on H-DAB images [2]. This tissue samples
is made with wax and dyed with one or more stains for
observing their architecture and various components [9].
2)
D. Classification
The distinctive classifiers are used to categorize images for
analyze irregularity in image after segmentation and feature
extraction process. They classify cells or tissues according to
some classes and then also detect the malignancy level in
particular samples. A lot of work is carry out on the
classification of medical images using machine learning
algorithms like neural network[10], logistic regression
method, k-nearest neighborhood algorithm, fuzzy systems etc.
Muthu Rama et al. [18] used SVM technique to classify oral
mucosa images. Identical SVM technique along with
multilayer perception and k-nearest neighbor classifier [15]
used as well as compare all tree classifier. To detect prostatic
cancer [17] author refer bayes classifier to classify each
regions to suitable grade. In [6] linear classifier applies to
determine final output. Linear classifier trained such that they
can classify malignant and benign samples.
III. COMPARISON
Histological image analysis is very recent and interested
area in image processing research, along with very difficult to
process as its having complicated and intricate structure. But
in each cytological image different methodologies applied to
obtain output. Different paper show that images goes through
diverse techniques like pre-processing, feature abstraction,
segmentation and classification.
Sr
No
1
Name
C.Smochin
a,V.Manta,
andW.
Kropatsch[
3]
R. Farjam
et al[6]
Colon
gland
Prostate
gland
Hitesh
garg[10]
Thyroid
gland
D.
Altunbay[1
2]
Colon
gland
K Nguyen,
Anil
K.
Jain
et.al[15]
Prostate
gland
Limitation
1.Memory use
decrease due
to
Morphological
pyramid
2.
Computation
time decrease
1.Give
false
negative result for
big
gaps
in
epithelial layer
2.Fails in Elongate
lumen
1.Automatic
approach for
detection
gland
2.Calculate
index
proportional
to malignancy
MRI and CT
images
segmented
and
obtain
result
Extract new
features from
color
graph
and use it for
classification
of tissues
extract glands
along
with
grading
information
1.Changes energy
and
entropy
features
when
images in low
magnification
2.Too
many
Produced features
Cannot calculate
estimation
of
segmented thyroid
area
Not possible to
use color graph
features
for
unsupervised
segmentation
Problem to detect
boundary of grade
3 and grade 4
Extracting Glandular
Feature
Compute Normal &
Abnormal Glands
Generate Report
End
Fig. 4. Flow of Proposed Work
[5]
[6]
[7]
[8]
[9]