17-10-2016, 10:08 AM
Segmentation of Preprocessed MR and CT Images Containing Tumors using Edge Detection and Watershed Segmentation
1459327724-segmentationwatershed.pdf (Size: 562.72 KB / Downloads: 14)
ABSTRACT
Segmentation of images aims at dividing areas corresponding
to different objects. There are two approaches for image
segmentation, one is based on discontinuities and other is based
on similarities. These approaches can be used for enhancing
and extracting the tumor area in MRI/CT images. In this
paper Sobel and Extended Sobel edge operators are applied
on the MRI / CT images containing tumors. It is noticed that
the MR/CT images contain unwanted portions that make
segmentation difficult. If such images are segmented without
any preprocessing for removal of the unwanted portions, it
results into over segmentation. In this paper, we propose to use
Preprocessed MRI/CT image for the segmentation by using
Sobel and extended Sobel operators. Results of both the
methods on original and preprocessed images are displayed.
The results of Watershed segmentation algorithm on original
and preprocessed images are also displayed. It is observed that,
the appropriate preprocessing of MR/CT images helps to
significantly reduce the problem of over segmentation of these
images still retaining the tumors.
INTRODUCTION
Medical image processing is used as an important tool in
computer-aided diagnosis for assisting doctors in evaluating
medical imagery or in recognizing abnormal findings in a medical
image. Structures of interest include organs or parts thereof, such
as cardiac ventricles or kidneys, abnormalities such as tumors and
cysts, as well as other structures such as bones, vessels, brain
structures etc.
CONCLUSIONS
From the displayed results it is seen that edge detection on original
image leads to detection of edges which are not of relevance from
the point of enhancement of tumor. In this case the tumor may or
may not be retained as seen from images b and c of fig.3. The edge
detection operators give far better results in terms of segmentation
and retention of tumors. The watershed segmentation is known to
produce over segmentation. But, if watershed segmentation is
carried on preprocessed image, as proposed, the over segmentation problem significantly subsides. It is also seen that the Extended
Edge operator gives better results considering prominent edges.