Content uploaded by Bhavani Krishna
Author content
All content in this area was uploaded by Bhavani Krishna on Oct 01, 2024
Content may be subject to copyright.
© 2019, IJCSE All Rights Reserved 95
International Journal of Computer Sciences and Engineering Open Access
Research Paper Vol.-7, Issue-6, June 2019 E-ISSN: 2347-2693
Analysis of Pre-processing Techniques on CT DICOM Images
1*Bhavani K, 2M T Gopalakrishna
1 Department of Information Science and Engineering, Dayananda Sagar College of Engineering, Bengaluru, India
2Department of Computer Science and Engineering, K S School of Engineering and Management, Bengaluru, India
Corresponding Author: bhavanik7@gmail.com
DOI: https://doi.org/10.26438/ijcse/v7i6.9598 | Available online at: www.ijcseonline.org
Accepted: 14/Jun/2019, Published: 30/Jun/2019
Abstract— In the present days, cancer has become a menacing disease. Lung cancer is the foremost cancer affecting both men
and women throughout the world. In this regard, biomedical imaging is a technology that aids fundamental medical
investigations. Some of the widely applied biomedical imaging techniques are Computed Tomography (CT), Magnetic
Resonance Imaging (MRI), etc. Among the imaging techniques, CT images are generally used for detecting life frightening
pathologies. CT images present high spatial resolution including contrast deviation in tissue. However, CT images are prone to
Gaussian noise due to thermal energy fluctuations. Also CT images get affected by artifact and structural noise which hamper
correct diagnosis. To overcome this problem, different de-noising filters like Median filter, Gaussian filter, Box filter, Average
filter, X-filter are applied on CT images before further processing. In order to identify the superlative filter metrics like SNR
(Signal to Noise Ratio) and PSNR (Peak Signal to Noise Ratio) are used. The CT image dataset in (Digital Imaging and
Communications in Medicine) DICOM format provided by the (Lung Image Database Consortium) LIDC has been utilized to
perform the analysis in the present work.
Keywords— CT, SNR, PSNR , Filter, DICOM
I. INTRODUCTION
Medical imaging is a method and way of visual depiction of
internal parts of human body [3]. Medical imaging is a
blooming technology used for diagnosing the patient’s
abnormalities. There are many modalities [3] like CT, PET,
and MRI etc. Majorly, the CT is the initially preferred
modality used for cancer detection. CT scan, a superior,
influential x-ray facilitates imaging of inner organs. In CT, a
gleam of X-rays rolls across the person being examined and
is picked up via amenable energy detectors subsequent to
penetration of the inner organs from multiple angles. A
computer subsequently examines the details acknowledged
from the CT detectors, as well as builds a complete image of
the inner organs. Compared to other imaging techniques CT
scan provides better detail of the images.
Cancer is a disorder which occurs when there is uncontrolled
increase of cells in human body. The growth can invade or
stretch to further parts of the body. Symptoms of Cancer are
varied and it depends on the volume and the type of organ
affected. Cancerous cells hamper the normal functioning of
the human body. Few of the common symptoms of cancer
are increase in body temperature, weariness, extreme
sweating, anemia, and mysterious mass reduction.
Treatments given for cancer in general are hormonal therapy,
targeted therapy, immune therapy, radiation therapy, surgery,
gene therapy and chemotherapy. Surgery is the principal
means of handling for last stage cancers. Cancer detection is
done using any of these methods like objective inspection
(biopsy), blood or urine tests, or medical imaging. Medical
imaging plays an extremely crucial role in diagnosing cancer
without the need to cut open the body of the patient. Lung
cancer in general is the uncontrolled development of
irregular cells in one or both lungs that can be better detected
using CT images.
During image acquisition and transmission of CT images,
noises get added on to the images. So to remove these noises,
filters are used. The pre processing of CT medical images
consists of removing noises before further processing. There
are many pre processing filters like Median filter, Gaussian
filter, Box filter etc. that can be used for removing the noises.
In the subsequent section, various filters used in the pre
processing stage of lung cancer detection are conversed.
II. RELATED WORK
Lung cancer is the foremost cancer affecting both men and
women throughout the world. Many studies have been
conducted and are still being conducted to detect lung cancer at
International Journal of Computer Sciences and Engineering Vol. 7(6), Jun 2019, E-ISSN: 2347-2693
© 2019, IJCSE All Rights Reserved 96
an early stage. In this section, an outline of research works
carried out to detect lung cancer is discussed.
Suren Makaju et al. [1] propose a model consisting of initial
processing of images to eliminate random variations followed
by image partitioning techniques. During the initial processing
stage, the filters like median and Gaussian are employed to
eliminate the random variations in the image. Kamil et al [2]
proposes to use median filter to lessen the random variations
in CT images while maintaining the fine points. The median
filter concurrently minimized noise and preserved edges.
Aarthi et al. [3] explain the different image modalities like
ultrasound, CT, MRI stressing more upon the ultrasound
images. Hasan Koyuncu et al. [4] propose a denoising
process using Block Matching and 3D Filtering (BM3D)
method intended for removal of noise. Brij Bhan et al. [5]
discuss contrast enhancement methods using the blend of
contrast limited adaptive histogram equalization and discrete
wavelet transform methods to apply on medical images.
M.Jayanthi [6] converts the standard DICOM images into
gray level images before further processing. Median filter is
then applied to get rid of unnecessary variations in images.
Khobragade et al. [7] propose segmentation of lungs, its
feature extraction plus classification using artificial neural
network method designed in support of finding lung ailments
like TB, lung cancer of the lungs as well as pneumonia.
Carlos Ciompi et al. [8] presents a new method for
categorizing pulmonary nodules. The method uses nodule
morphology like bigotry of nodules, vascular constitution
along with depiction of conjecture. Md. Badrul Alam et al.
[9] have used binarization method for preprocessing. The
system uses stages such as acquiring of image, pre
processing, binarization, thresholding, segmentation, feature
extraction, and neural network detection [9]. Ritika Ankit
Rajet al. [10] have reviewed the literature on finding lung
malignancy via medical content based image retrieval. They
have provided content support based image reclamation
Computer Aided Diagnosis System (CAD) intended for lung
cancer detection at the early stage using the CT images.
III. PRE PROCESSING
CT images widely supports planning and follow up actions
necessary during the course of cancer stages. Further CT
scans can expose the presence of tumours and lesion, along
with the spread, position and deepness of tumour. Noise in
CT is an unnecessary change in pixel values in an otherwise
homogenous image. The noise in CT images essentially
condenses the visibility of small distinct objects. CT images
are prone to Gaussian noise owing to the appearance of
electrical signals. The presence of random noise also
confines the capability of the medical practitioners to
discriminate between tissues of varying density.
Pre processing is a technique adapted to remove noise
present in the image and sharpen the edges. An image
filtering algorithm produces an output pixel by examining
the neighbourhood of each of the input pixel in an
image. Various filters like Median filter, Gaussian Filter,
Box filter are applied to eliminate the random variations
present within the CT image.
A. Median filter
The median filter operates by affecting the image pixel by
pixel, substituting every value by the median of adjoining
pixels in the window w. The median is calculated by
G(x,y) ={ median{f(i,j), (i,j)Ɛ w} (i)
where w characterizes a neighborhood centered around the
locality [x,y] in the image.
B. Gaussian filter
A Gaussian kernel provides more weight to pixels in close
proximity of the current pixel and smaller weight to far
pixels when calculating sum. The nature of the Gaussian
function is dogged based on the standard deviation.
(ii)
where σ is the standard deviation of the allocation
encompassing the input image.
C. Box filter
Box filtering is applied based on the average of adjoining
pixel encompassing the image. In filtering process, image
model and the filter core are multiplied to obtain the output
result.
(iii)
where G(x,y) represents the filtered image.
D. Average filter
The average filter working is by affecting the image pixel by
pixel, restoring every value by the average value of adjacent
pixels. It is calculated by
(iv)
E. X-filter
The X-filter are non-linear spatial filters where the retort is
dependent on the organization of the pixels confined in the
image region contained within the filter and subsequently
restoring the value in the middle pixel with the value dogged
by means of the organization result.
Based on the application of the above filters, results are
tabulated based on the SNR and PSNR values. SNR is given
as the relation of the mean value of the signal and the
standard deviation of the noise. SNR is assessed independent
of the nature of noise being analyzed. But the consequence
International Journal of Computer Sciences and Engineering Vol. 7(6), Jun 2019, E-ISSN: 2347-2693
© 2019, IJCSE All Rights Reserved 97
and usability of the parameter is especially dependent of the
kind of noise.
PNSR (Peak Signal to Noise Ratio) is a feature measure
linking the original image and a image being tested. The high
PSNR value indicates the healthier value of the image. PSNR
basically characterizes a measure of the peak error. To
determine PSNR value, MSE (Mean Squared Error) values
are computed first using the equation:
(v)
The PSNR values are then determined by the formula:
(vi)
A range of filters are applied on images and the SNR and
PSNR values are found out. The filters like median Filter,
Gaussian Filter, Box filter, Average Filter, X Filter are
applied on Lung CT DICOM images. The Table I are
tabulated with SNR results and Table II are tabulated with
PSNR results respectively.
TABLE I
SNR OF DIFFERENT FILTERING TECHNIQUES
FILTERING
TECHNIQUES
CT SCAN IMAGES
1
2
3
4
5
Gaussian Filter
19.51
20.29
21.07
22.09
20.53
Box Filter
19.62
20.45
21.04
22.5
20.68
Average Filter
19.5
20.26
21.04
22.03
20.49
Median Filter
21.47
21.69
23.01
23.87
21.8
X Filter
24.52
25.02
24.12
26.28
24.84
TABLE II
PSNR OF DIFFERENT FILTERING TECHNIQUES
FILTERING
TECHNIQUES
CT SCAN IMAGES
1
2
3
4
5
Gaussian Filter
35.21
32.68
37.58
34.14
32.24
Box Filter
35.18
32.73
37.45
34.46
32.28
Average Filter
35.06
32.54
37.45
33.99
32.10
Median Filter
37.03
33.97
39.42
35.83
33.40
X Filter
40.08
37.30
40.53
38.23
36.45
IV. RESULTS AND DISCUSSION
Pre processing is basically required to eliminate random
variations in the image and sharpen the edges. The filters are
applied to eliminate noise. The results of the application of
filters on lung CT DICOM images are graphically
represented in figure 1.
Figure 1: PSNR of different filtering techniques
Table 2 shows the PSNR values of every tested filters
specifically Median filter, Gaussian filter, Box filter,
Average filter and X filter. The application of the
corresponding filters on lung CT images can be viewed in
figure 2. Figure 2 shows the original images and the
corresponding images after the application of filters. The
results confirm that X filter produces the highest PSNR
compared to other filters. Thus, it can be concluded that X-
filter offers the improved PSNR value with better clarity
Figure 2: Original images and the corresponding images after
filtering
V. CONCLUSION AND FUTURE SCOPE
The current systems for cancerous nodule dection consist of
segmentation process followed by classification. But, before
International Journal of Computer Sciences and Engineering Vol. 7(6), Jun 2019, E-ISSN: 2347-2693
© 2019, IJCSE All Rights Reserved 98
performing segmentation, image preprocessing is very much
necessary as this improves the results of further stages. So in
this paper, the application of various filters and its
corresponding results are analyzed . Various de-noising
filters applied are Median filter, Gaussian filter, Box filter,
Average filter and X-filter. From the results, it can be
analyzed that X filter bestows the superior PSNR value
compared to any other filter. The work can be further
validated by testing on a larger dataset.
REFERENCES
[1] Suren Makaju, P.W.C. Prasad, Abeer Alsadoon, A. K. Singh, A
Elchouemi, ―Lung Cancer Detection using CT Scan Images ‖, 6th
International Conference on Smart Computing and
ommunications, ICSCC 2017, 7-8 December 2017, Kurukshetra,
India.
[2] Kamil Dimililer, Buse Ugur, Yoney K. Ever , ―Tumor Detection On
CT Lung Images Using Image Enhancement‖, The Online Journal
of Science and Technology,Volume 7, Issue 1, January 2017.
[3] Aarthi poornima Elangovan1, Jeyaseelan.T, ―Medical Imaging
Modalities: A Survey‖, IEEE, 2017.
[4] Hasan Koyuncu, Rahime Ceylan, ―A Hybrid Tool on Denoising and
Enhancement of Abdominal CT Images before Organ & Tumour
segmentation‖, IEEE 37th International Conference on Electronics
and Nanotechnology,2017.
[5] Brij Bhan Singh, Shailendra Patel, ―Efficient Medical Image
Enhancement using CLAHE Enhancement and Wavelet
Fusion‖,International Journal of Computer Applications (0975 –
8887) Volume 167 – No.5, June 2017.
[6] M.Jayanthi, ―Comparative Study of Different Techniques Used for
medical image Segmentation of Liver from Abdominal CT Scan‖,
IEEE WiSPNET 2016 conference.
[7] Khobragade, S., Tiwari, A., Patil, C., Narke, V., ―Automatic
detection of major lung diseases using Chest Radiographs and
classification by feed-forward artificial neural network.‖, IEEE
International Conference on Power Electronics, Intelligent Control
and Energy Systems (ICPEICES), 2016.
[8] Ciompi F, Jacobs C, Scholten E.T, Wille M.M.W de Jong, P.A.,
Prokop, M., van Ginneken, B, ―Bag-of-Frequencies: A Descriptor
of Pulmonary Nodules in Computed Tomography Images‖,
Medical Imaging, IEEE Transactions on , vol.34, no.4,
pp.962,973, April 2015.
[9] Md. Badrul Alam Miah, Mohammad Abu Yousuf, ―Detection of
lung cancer from CT image using image processing and neural
network‖, International Conference on Electrical Engineering and
Information Communication Technology (ICEEICT), 2015.
[10] Ritika Agarwal, Ankit Shankhadhar, Raj Kumar Sagar, ―Detection
of Lung Cancer Using Content Based Medical Image Retrieval‖,
Fifth International Conference on Advanced Computing &
Communication Technologies, IEEE, 2015.
[11] V. Vijaya Kishore, R.V.S.Satyanarayana, ―Performance Evaluation
of Edge Detectors -Morphology Based ROI Segmentation and
Nodule Detection from DICOM Lung Images in the Noisy
Environment‖, IEEE,2012.
[12] Yang Yu, Hong Zhao, ―A Texture-based Morphologic
Enhancement Filter in Two-dimensional Thoracic CT scans‖,
IEEE,2006.
[13] Kshipra Singh, Jijo S Nair, ―A Literature Review On Satellite
Image Data Enhancement Using Digital Image Processing‖,
International Journal of Computer Sciences and Engineering,
Vol.6, Issue.7, pp.1114-1119, 2018.
[14] M. Rathika, R. Shenbagavalli, "Image Restoration of Damaged
Mural images based on Image Decomposition", International
Journal of Computer Sciences and Engineering, Vol.07, Special
Issue.08, pp.32-37, 2019.
Authors Profile
Mrs Bhavani K received Bachelor of
Engineering in Computer Science &
Engineering in 2006, Master Degree M.Tech
in Computer Science and Engineering in
2009. She is currently pursuing Ph.D
from Visvesvaraya Technological University,
Karnataka. Her area of research include Medical image
processing Computer Vision,AI.
Dr. M T Gopalakrishna received Bachelor of
Engineering in Computer Science &
Engineering in 1999 from Bangalore
University of Karnataka, India, the Master
Degree M.Tech in Computer Science and
Engineering from Visvesvaraya
Technological University, Karnataka, India in 2004 and PhD
in Computer Science and Engineering from Visvesvaraya
Technological University, Karnataka, India in 2014. He has
published research papers in reputed international journals
and conferences. His area of research include Computer
Vision, Pattern Recognition , Image
Processing,Visual Surveillance-Object Detection/
Classifications/ Tracking/ Identifications.
