Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
RADIOGRAPH COMPRESSION USING NEURAL NETWORKS
AND HAAR WAVELET
1Dr.F.Emerson Soloman , A karthik
2
1Assistant Professor
2UG Students,
Department of Biomedical Engineering
BIHER, BIST, Bharath University
Chennai- 6000073.
Abstract
Efficient storage and transmission of medical images in telemedicine is of utmost importance
however, this efficiency can be hindered due to storage capacity and constraints on
bandwidth. Thus, a medical image may require compression before transmission or storage.
Ideal image compression systems must yield high quality compressed images with high
compression ratio; this can be achieved using wavelet transform based compression,
however, the choice of an optimum compression ratio is difficult as it varies depending
on the content of the image. In this paper, a neural network is trained to relate
radiograph image contents to their optimum image compression ratio. Once trained,
the neural network chooses the ideal Haar wavelet compression ratio of the x-ray
images upon their presentation to the network. Experimental results suggest that our proposed
system, can be efficiently used to compress radiographs while maintaining high image
quality.
I. INTRODUCTION
Radiographs are images produced on a radiosensitive surface, such as a photographic
film, by radiation other than visible light, especially by x-rays passed through an object or by
photographing a fluoroscopic. These images, commonly referred to as x-rays, are usually
used in medical diagnosis, particularly to investigate bones, dental structures, and foreign
objects within the body. X-rays are the second most commonly used medical tests, after
laboratory tests. Recently, teleradiology, which is one of the most used clinical aspects of
telemedicine, has received much attention. [1-6]Teleradiology attempts to transfer medical
images of various modalities, like computerized tomography (CT) scans, magnetic imaging
(MRI), ultrasonography (US), and x-rays from one location to another such as in hospitals,
imaging centers or a physician’s desk. The radiological images are needed to be
International Journal of Pure and Applied MathematicsVolume 119 No. 12 2018, 11897-11911ISSN: 1314-3395 (on-line version)url: http://www.ijpam.euSpecial Issue ijpam.eu
11897
compressed before transmission to a distant location or due to the bandwidth or storage
limitations [1].
There has been a rapid developmentin compression methods to compress large data
files such as images where data compression in various applications has lately became more
vital [2]. With the improvements of technology efficient methods of compression are needed
to compress and store or transfer image data files while retaining high image quality and
marginal reduction in size [3]. Wavelets are a mathematical tool for hierarchically
decomposing functions. There is a general preference to use wavelet transforms in image
compression because the compressed images and be obtained with higher compression
ratios and higher PSNR values [4]. Unlike the discrete cosine transform, the wavelet
transforms are not fourier based and therefore discontinuities in image data can be handled
with better results using wavelets [5]. Haar wavelet transform is one of the wavelet methods
applied in compressing the digital images. Previous works using haar image compression
include an application that is applied to adaptive data hiding for the images dividing the
original image into 8x8 sub-blocks and reconstructing the images after compression with
good quality [6], and the use of Parametric Haar-like transform that is based on a fast
orthogonal parametrically adaptive transform such that it may be computed with a fast
algorithm in structure similar to classical haar transform [7].
The use of compression methods in general, and wavelet compression in particular, with
medical images has been previously investigated. For example, in [1] an adaptive
approach for the classification of blocks on the basis of an adaptive threshold value of
variance, was presented and applied to different medical images such as CT, X- ray
and ultrasound images. In [8] a compression algorithm for far infrared medical images was
proposed based on lifting wavelet transform and histogram shifting. In [9] an algorithm to
compress and to reconstruct Digital Imaging and Communications in Medicine
(DICOM) images. Their algorithm consisted of two stages: DICOM images were first
decomposed using generalized Cohen-Daubechies-Feauveau biorthoganal wavelet and
the wavelet coefficients were encoded using Set Partitioning In Hierarchical Trees (SPIHT).
In [10] an image compression scheme, using the discrete wavelet transformation
(DWT) and the k-means clustering technique, was suggested and applied to medical images.
In [11] a method based on topology- preserving neural networks was used to implement
vector quantization for medical image compression.
International Journal of Pure and Applied Mathematics Special Issue
11898
Original Image 10% 20% 30% 40%
50% 60% 70% 80% 90%
Fig.1 – An original radiograph and its Haar compression at nine ratios.
In [12] A lossless wavelet-based image compression method with adaptive prediction was
proposed, and applied to achieve higher compression rates on CT and MRI images. In
[13] a combining technique for image compression based on the Hierarchical Finite State
Vector Quantization (HFSVQ) and a neural network, was proposed and applied to medical
images. Artificial neural networks implementations in image processing applications
has marginally increased in recent years. Image compression using wavelet transform and
a neural network was suggested previously [14]. Moreover, different image
compression techniques were combined with neural network classifier for various
applications [15]. A neural network model called direct classification was also suggested;
this is a hybrid between a subset of the self-organising Kohonen model and the adaptive
resonance theory model to compress the image data [6]. Periodic Vector Quantization
algorithm based image compression was suggested previously based on competitive neural
networks quantizer and neural networks predictor [7]. More works using neural networks
emerged lately. In [8] a Multi-Resolution Neural Network (MRNN) filter bank was used as a
transform for coding. In [9] an image compression algorithm based on image blocks
International Journal of Pure and Applied Mathematics Special Issue
11899
complexity measure methods and a neural network was proposed. In [10] a direct solution
method applied to image compression using neural networks was suggested. In [11] a
Principal Component Analysis based neural network was used for image compression. In
[12] a neural network quantizer was used to yield a high compression ratio while
maintaining high quality images.
Recently, a neural network based DCT compression system that finds the optimum
compression ratios for a variety of images was also suggested [14]. The evaluation method of
the neural network-obtained optimum compression results was based on the comparison
criteria; which was suggested in [14]. More recently, a neural network model was used to
obtain the optimum compression ratio when using Haar wavelet image compression which
was applied to a variety of images [5]. The aim of the work presented within this paper is to
develop a medical radiographs compression system using Haar wavelet transform and a
neural network.
The proposed method suggests that a trained neural network can learn the non-linear
relationship between the intensity (pixel values) of a radiograph, or x-ray, image and its
optimum compression ratio. Once the highest compression ratio is obtained,
while maintaining good image quality, the result reduction in radiograph image size,
should make the storage and transmission of radiographs more efficient. The paper is
organized as follows: Section II describes the radiograph database which is used for the
implementation of our proposed system. Section III presents the radiograph
compression system; describing the x-ray image pre-processing and the neural network
design and implementation. Section IV introduces the method used to evaluate the results
and provides an analysis of the system implementation. Finally,
Section V concludes the work that is presented within this paper and suggests further
work.
II. RADIOGRAPH DATABASE
The development and implementation of the proposed medical radiograph compression
system uses 60 x-ray images from our medical image database [15], which contains
radiographs of fractured, dislocated, broken, and healthy bones in different parts of the
body. Haar wavelet compression has been applied to 48 radiographs using nine
compression ratios (10%, 20%, ..., 90%) as shown in an example in Fig. 1.
International Journal of Pure and Applied Mathematics Special Issue
11900
a- Training set examples, b- Testing Set 1 examples, c- Testing Set 2 examples
Fig.2 – Radiograph database examples.
Radiograph 1 70 % Compressed
Radiograph 2 80 % Compressed
Radiograph 3 90 % Compressed
International Journal of Pure and Applied Mathematics Special Issue
11901
Fig.3 – Examples of Training set radiographs and their optimum Haar compression.
The optimum Haar compression ratios for the 48 radiographs were determined using the
optimum compression criteria based on visual inspection of the compressed images as
suggested in [7], thus providing 48 images with known optimum compression
ratios and the remaining 12 images with unknown optimum compression ratios. The image
database is then organized into three sets:
Training Image Set: contains 25 images with known optimum compression ratios which
are used for training the neural network within the radiograph compression system.
Examples of training images are shown in Fig. 2a.
Testing Image Set 1: contains 23 images with known optimum compression ratios
which are used to test and validate the efficiency of the trained neural network.
Examples of these testing images are shown in Fig. 2b.
Testing Image Set 2: contains 12 images with unknown optimum compression ratios
which are used to further test the trained neural network. Examples of these testing
images are shown in Fig. 2c.[11-36]
Examples of original radiographs and their compressed versions using their
ptimum compression ratios while training the neural network are shown in Fig. 3.
III. RADIOGRAPH COMPRESSION SYSTEM
The optimum radiograph compression system uses a supervised neural network based on
the back propagation learning algorithm, due to its implementation
simplicity, and the availability of sufficient “input/target” database for training this
supervised learner. The neural network relates the x- ray image intensity (pixel values) to
the image optimum compression ratio having been trained using images with predetermined
optimum compression ratios. The ratios vary according to the variations in pixel values
within the images. Once trained, the neural network would choose the optimum
compression ratio of a radiograph upon presenting it to the neural network by using its
intensity values. The radiographs require pre-processing prior to presenting them to the
neural network. Pre-processing the x-rays aims to reduce the amount of necessary data
from images while maintaining meaningful representation of the contents of the
radiographs. Adobe Photoshop was used to resize the original radiographs of size (256x256)
pixels into (64x64) pixels. Further reduction to the size of the images was attempted in
International Journal of Pure and Applied Mathematics Special Issue
11902
order to reduce the number of input layer neurons and consequently the training time,
however, meaningful neural network training could not be achieved thus, the use of
whole images of the reduced size of 64x64 pixels. The size of the input x-ray images affects
the choice of the number of neurons in the neural network’s input layer, which has three
layers; input, hidden and output layers. Using one-pixel-per-neuron approach, the neural
network’s input layer has 4096 neurons, its hidden layer has 50 neurons, which assures
meaningful training while keeping the time cost to a minimum, and its output layer has nine
neurons according to the number of the considered compression ratios (10% - 90%).
During the learning phase, the learning coefficient and the momentum rate were adjusted
during various experiments in order to achieve the required minimum compression ratio Si
as determined by the trained neural network, and is defined as follows:
Error value of 0.005; which was considered as sufficient for this application. Fig. 4
shows the topology of this neural network, within the radiograph compression system.
OCD = (| Sp – Si |)*10.
IV. RESULTS AND DISCUSSIONS
The evaluation of the training and testing results was performed using two
measurements: the recognition rate and the accuracy rate. The recognition rate is defined as
follows:
The OCD is used to indicate the accuracy of the system, and depending on its value the
recognition rates vary. Table 1 shows the three considered values of OCD and their
corresponding accuracy rates and recognition rates. The evaluation of the system
implementation results uses (OCD = 1) as it provides a minimum accuracy rate of 89%
which is considered sufficient for this application.
The neural network learnt and converged after 1831 iterations or epochs, and within
543.33 seconds,
RR OHC = I OHC *100,
IT
where RROHC is the recognition rate for the neural network within the radiograph
compression system, IOHC is the number of optimally compressed x-ray images, and
IT is the total number of x-ray images in the database set.
The accuracy rate RAOHC for the neural network output results is defined as follows:
International Journal of Pure and Applied Mathematics Special Issue
11903
whereas the running time for the generalized neural networks after training and using
one forward pass was 0.013 seconds. These results were obtained using a 2.0 GHz PC
with 2 GB of RAM, Windows XP OS and Matlab R2009a software. Table 2 lists the final
parameters of the successfully trained neural network, whereas Fig. 5 shows the error
minimization curve of the neural network during learning. The trained neural network
recognized correctly the optimum compression ratios for all 25 training images as would
be expected, thus yielding 100% recognition of the training set. Testing the trained
neural network using the 23 images from Test Set 1
(| Sp – Si|)*10
RAOHC = 1- *100
ST
that were not presented to the network before yielded 95.65% recognition rate, where 22
out of the 23 images with known optimum compression ratios. where SP represents the
pre-determined (expected) optimum compression ratio in percentage, Si
represents the optimum compression ratio as determined by the trained
neural network in percentage and ST represents the total number of compression
ratios.
The Optimum Compression Deviation (OCD) is another term that is used in our
evaluation. OCD is the difference between the pre-determined or expected optimum
compression ratio SP and the optimum The trained neural network was also implemented
using the remaining 12 images with unknown optimum compression ratios from the testing
set. The results of this application are demonstrated Fig. 6 which shows examples of the
optimally compressed radiographs as determined by the trained neural network.[37-45
Table 1. Optimum Compression Deviation and corresponding Rates
Input neurons 4096
Hidden neurons 50
Output neurons 9
Learning coefficient 0.005
Momentum rate 0.40
Minimum error 0.005
Iterations 1831
Training time (seconds) 543.33
Run time (seconds) 0.013
International Journal of Pure and Applied Mathematics Special Issue
11904
Table 2. Trained neural network final parameters
OCD
Accuracy Rate
(RAOHC)
Recognition
Rate
(RROHC)
0 100 % 13/23 (56.52 %)
1 89 % 22/23 (95.65 %)
2 78 % 23/23 (100 %)
Fig. 5 – Neural Network Learning Curve
Fig. 7 – Performance Curve
Fig. 8 – Training State
International Journal of Pure and Applied Mathematics Special Issue
11905
V. CONCLUSIONS
A novel method to medical radiograph compression with nine compression ratios
and a supervised neural network that learns to associate the grey x-ray image intensity
(pixel values) with a single optimum compression ratio. The implementation of the
proposed method uses haar image compression where the quality of the compressed images
degrades at higher compression ratios due to the nature of the lossy wavelet compression.
The aim of an optimum ratio is to combine high compression ratio with good quality
compressed radiograph, thus making the storage and transmission of radiographs
more efficient. The proposed system was developed and implemented using
60 radiographs or x-ray images of fractured, dislocated, broken, and healthy bones in
different parts of the body. The neural network within the radiograph compression system
learnt to associate the 25 training x-ray images with their predetermined optimum
Radiograph 1
Radiograph 2
80% Compression
80% Compression
International Journal of Pure and Applied Mathematics Special Issue
11906
compression ratios within 543.33 seconds. Once trained, the neural network could
recognize the optimum compression ratio of an x-ray image within 0.013 seconds. In this
work, a minimum accuracy level of 89% was considered as acceptable. Using this accuracy
level, the neural network yielded 95.65% correct recognition rate of optimum compression
ratios. The successful implementation of our proposed method using neural networks was
shown throughout the high recognition rates and the minimal time costs when running the
trained neural networks. Future work will include the implementation of this method using
biorthogonal wavelet transform compression and comparing its performance with Haar-
based radiograph compression.
REFERENCES
1. Khanaa, V., Thooyamani, K.P., Udayakumar, R., Multi resigncryption protocol
scheme for an identification of malicious mobile agent host, World Applied Sciences
Journal, V-29, I-14, PP-299-303, 2014
2. Khanaa, V., Thooyamani, K.P., Udayakumar, R., A novel approach towards
prevention of spam, phising, controlling hierarchical access using Domain Keys,
World Applied Sciences Journal, V-29, I-14, PP-181-185, 2014
3. Khanaa, V., Thooyamani, K.P., Udayakumar, R., Dual tree complex wavelet
transform for adaptive interferogram residual reduction, Middle - East Journal of
Scientific Research, V-20, I-9, PP-1059-1064, 2014
4. Khanaa, V., Thooyamani, K.P., Udayakumar, R., An improved rao-blackwellized
particle filter for GPS-INS integrated navigation, Middle - East Journal of Scientific
Research, V-20, I-9, PP-1111-1115, 2014
5. Khanaa, V., Thooyamani, K.P., Udayakumar, R., WSN based gas leakage detecting
and locating system for petrochemical industry, Middle - East Journal of Scientific
Research, V-20, I-9, PP-1107-1110, 2014
6. Khanaa, V., Thooyamani, K.P., Udayakumar, R., Information sharing time based
controlled system in emergency situations, Middle - East Journal of Scientific
Research, V-18, I-12, PP-1845-1850, 2013
7. Kumar, J.R., Vikram, C.J., Naveenchandran, P., Investigation for the performance and
emission test using citronella oil in single cylinder diesel engine, Middle - East
Journal of Scientific Research, V-12, I-12, PP-1754-1757, 2012
8. Manavalan, S., Golden RenjithNimal, R.J., Efficient and evaluation properties of
medicinal products and flowering plant conventional diesel, International Journal of
Pure and Applied Mathematics, V-116, I-19 Special Issue, PP-407-411, 2017
9. Manavalan, S., Jose Ananth Vino, V., Extraction and optimization of Biodiesel from
neem oil, International Journal of Pure and Applied Mathematics, V-116, I-19 Special
Issue, PP-401-404, 2017
10. Manavalan, S., Rakesh, N.L., Heat transfer enhancement in shell and tube heat
exchangers using twisted tapes, International Journal of Pure and Applied
Mathematics, V-116, I-17 Special Issue, PP-405-407, 2017
International Journal of Pure and Applied Mathematics Special Issue
11907
11. Manikandan, J., Hameed Hussain, J., Design on blind shoe using ATMEGA328
arduino, International Journal of Pure and Applied Mathematics, V-116, I-19 Special
Issue, PP-413-415, 2017
12. Manikandan, J., Hameed Hussain, J., Design and fabrication of vibrating table for
seperating of nuts using different sizes mesh, International Journal of Pure and
Applied Mathematics, V-116, I-19 Special Issue, PP-417-419, 2017
13. Manikandan, J., Hussain, J.H., Design on blind shoe using ATMEGA328 micro
controller, International Journal of Mechanical Engineering and Technology, V-8, I-8,
PP-1575-1579, 2017
14. Manikandan, J., Hussain, J.H., Design and fabrication of blind shoe using
ATMEGA328 micro controller and vibration motor, International Journal of Pure and
Applied Mathematics, V-116, I-17 Special Issue, PP-287-289, 2017
15. Manikandan, J., Hussain, J.H., Design and fabrication of blind shoe using
ATMEGA328 micro controller and vibration motor, International Journal of
Mechanical Engineering and Technology, V-8, I-8, PP-1588-1593, 2017
16. Manikandan, J., Sabarish, R., Waste heat recovery power generation system using
thermoelectric generators, International Journal of Pure and Applied Mathematics, V-
116, I-17 Special Issue, PP-85-88, 2017
17. Meenakshi, D., Udayakumar, R., Protocol with hybridized bluetoothscatternet
formation for wireless network, International Journal of Applied Engineering
Research, V-9, I-22, PP-7299-7307, 2014
18. Meikandaan, T.P., Hemapriya, M., Use of glass FRP sheets as external flexural
reinforcement in RCC Beam, International Journal of Civil Engineering and
Technology, V-8, I-8, PP-1485-1501, 2017
19. Meikandaan, T.P., Hemapriya, M., Use of glass FRP sheets as external flexural
reinforcement in RCC beam, International Journal of Pure and Applied Mathematics,
V-116, I-13 Special Issue, PP-481-485, 2017
20. Meikandaan, T.P., Hemapriya, M., Study on properties of concrete with partial
replacement of cement by rice husk ash, International Journal of Pure and Applied
Mathematics, V-116, I-13 Special Issue, PP-503-507, 2017
21. Meikandaan, T.P., Hemapriya, M., Experimental behaviour of retrofitting of
prestressed concrete beam with FRP laminates, International Journal of Pure and
Applied Mathematics, V-116, I-13 Special Issue, PP-509-513, 2017
22. Meikandaan, T.P., Hemapriya, M., Study of damaged RC beams repaired by bonding
of CFRP laminates, International Journal of Pure and Applied Mathematics, V-116, I-
13 Special Issue, PP-495-501, 2017
23. Meikandaan, T.P., Hemapriya, M., Experimental study on strengthening of shear
deficient RC beam with externally bonded GFRP sheets, International Journal of Pure
and Applied Mathematics, V-116, I-13 Special Issue, PP-487-493, 2017
24. Michael, G., Arunachalam, A.R., Srigowthem, S., Ecommerce transaction security
challenges and prevention methods-new approach, International Journal of Pure and
Applied Mathematics, V-116, I-13 Special Issue, PP-285-289, 2017
25. Michael, G., Karthikeyan, R., Studies on malicious software, International Journal of
Pure and Applied Mathematics, V-116, I-8 Special Issue, PP-315-319, 2017
International Journal of Pure and Applied Mathematics Special Issue
11908
26. Michael, G., Srigowthem, S., Vehicular cloud computing security issues and
solutions, International Journal of Pure and Applied Mathematics, V-116, I-8 Special
Issue, PP-17-21, 2017
27. Micheal, G., Arunachalam, A.R., EAACK: Enhanced adaptive acknowledgment for
MANET, Middle - East Journal of Scientific Research, V-19, I-9, PP-1205-1208,
2014
28. Murugesh, Kaliyamurthie, Thooyamani, K.P., ICI self-cancellation schee for OFDM
systems, Middle - East Journal of Scientific Research, V-18, I-12, PP-1775-1779,
2013
29. Murugesh, Kaliyamurthie, Thooyamani, K.P., Preprocessing and postprocessing
decision tree, Middle - East Journal of Scientific Research, V-13, I-12, PP-1599-1603,
2013
30. Nakkeeran, S., Hussain, J.H., Innovative technique for running a petrol engine with
diesel as a fuel, International Journal of Pure and Applied Mathematics, V-116, I-18
Special Issue, PP-41-44, 2017
31. Nakkeeran, S., Nimal, R.J.G.R., Anticipation possessions for seismic use carbon black
on rubbers, International Journal of Pure and Applied Mathematics, V-116, I-17
Special Issue, PP-63-67, 2017
32. Nakkeeran, S., Sabarish, R., New method of running a petrol engine with diesel diesel
as fuel, International Journal of Pure and Applied Mathematics, V-116, I-18 Special
Issue, PP-35-38, 2017
33. Nakkeeran, S., Vino, J.A.V., Prevention effects for earthquakes using carbon black on
rubbers, International Journal of Pure and Applied Mathematics, V-116, I-17 Special
Issue, PP-301-305, 2017
34. Nakkeeran, S., Vino, J.A.V., Performance development by using mesh plates in
parallel and counter flows of heat exchangers, International Journal of Pure and
Applied Mathematics, V-116, I-17 Special Issue, PP-291-295, 2017
35. Nalini, C., Arunachalam, A.R., A study on privacy preserving techniques in big data
analytics, International Journal of Pure and Applied Mathematics, V-116, I-10 Special
Issue, PP-281-285, 2017
36. Nalini, C., Ayyappan, G., Academic social network dataset applying various metrics
for measuring author's contribution, International Journal of Pure and Applied
Mathematics, V-116, I-8 Special Issue, PP-341-345, 2017
37. Nalini, C., Brintha Rajakumari, S., Iron toxicity of polluted river water based on data
mining prediction analysis, International Journal of Pure and Applied Mathematics,
V-116, I-8 Special Issue, PP-353-356, 2017
38. Nandhini, P., Arunachalam, A.R., Fault-tolerant quality using distributed cluster based
in mobile ADHOC networks, International Journal of Pure and Applied Mathematics,
V-116, I-8 Special Issue, PP-365-368, 2017
39. Nandhini, P., Arunachalam, A.R., Security for computer organize database assaults
from threats and hackers, International Journal of Pure and Applied Mathematics, V-
116, I-8 Special Issue, PP-359-363, 2017
40. Nandhini, P., Arunachalam, A.R., Mobile ADHOC networks: Security and quality of
services, International Journal of Pure and Applied Mathematics, V-116, I-8 Special
Issue, PP-371-374, 2017
International Journal of Pure and Applied Mathematics Special Issue
11909
41. Naveenchandran, P., Vijayaragavan, S.P., A high performance inverter fed energy
efficient cum compact microcontroller based power conditioned distributed photo
voltaic system, International Journal of Pure and Applied Mathematics, V-116, I-13
Special Issue, PP-165-169, 2017
42. Naveenchandran, P., Vijayaragavan, S.P., Combination of wind energy-solar energy
power generation, International Journal of Pure and Applied Mathematics, V-116, I-
13 Special Issue, PP-117-121, 2017
43. Naveenchandran, P., Vijayaragavan, S.P., A sensor less control of SPM using fuzzy
and ANFIS technique, International Journal of Pure and Applied Mathematics, V-116,
I-13 Special Issue, PP-43-50, 2017
44. Nima, R.J.G.R., Hussain, J.H., Design and fabrication of an indexing fixture in a
shaper machine, International Journal of Pure and Applied Mathematics, V-116, I-18
Special Issue, PP-441-445, 2017
45. Nimal, R.J.G.R., Hussain, J.H., Deflection and stress analysis of spur gear tooth using
steel, International Journal of Pure and Applied Mathematics, V-116, I-17 Special
Issue, PP-119-124, 2017
International Journal of Pure and Applied Mathematics Special Issue
11910
11911
11912