BayesShrink
1
BayesShri nk Origin al WF Sigmo id Propos ed MTF Spatial frequency (mm -1 ) 0.2 1.0 0.8 0.6 0.4 0.0 1 2 3 4 5 0 Origin al Sigmo id Propose d BayesShr ink WF Figure 4 shows the GLC as a function of the radiation dose ratio. The sigmoid and the proposed methods could considerably improve the GLC at a 64/100 radiation dose ratio with respect to the standard dose, although the value of the GLC with the sigmoid and proposed methods is slightly lower than the original image at 100/100 and 80/100 radiation dose level. Furthermore, the proposed method shows the highest performance among all images at the 50/100 radiation dose ratio. The sigmoid filter was used to enhance image contrast. The coefficients with great values and the coefficients of high resolution levels are heavily weighted. Methods and Materials Methods and Materials Conclusion Conclusion IMPROVEMENT OF IMAGE QUALITY AND RADIATION DOSE REDUCTIONIN DIGITAL RADIOGRAPHY USING AN INTEGRATED WAVELET-TRANSFORM-BASED METHOD Haruyuki Watanabe 1 , Du-Yih Tsai 1 , Yongbum Lee 1 , Eri Matsuyama 1 , and Katsuyuki Kojima 2 1 Department of Radiological Technology, Graduate School of Health Sciences, Niigata University, Niigata, Japan 2 Department of Business Administration, Graduate School of Business Administration, Hamamatsu University, Hamamatsu, Japan Introduction Introduction In order to validate the superiority and effectiveness of the proposed method, we compared the proposed method with 3 conventional methods, namely, the Wiener filter (WF), BayesShrink method, and sigmoid-type method. In the wavelet-transform-based method, wavelet basis functions of Daubechies (order=4, level=4) were adopted for image processing. The proposed method and the above- described three methods were applied to the original images for performance comparison. Wavelet-Based Image Processing Wavelet-Based Image Processing The issue of radiation dose exposure to the patients with digital radiography is a major public health concern. It is known that a trade- off exists between noise level and radiation dose. On the one hand, high- dose radiation will lower the noise level, but may give excess radiation doses to the patient. On the other hand, low-dose radiation will lower the signal-to-noise ratio of the image and result in reducing the amount of image information. The balancing of dose and image quality should be performed explicitly to ensure that patient doses are kept as low as reasonable achievable, while maintaining a clinically acceptable image quality. To deal with this issue, much research, such as the development of new detectors and that of image processing methods has been carried out. In recent years, several investigators have reported that wavelet-based image processing techniques are effective in the reduction of radiation dose . In the BayesShrink scheme, the threshold is determined for each subband by assuming a generalized Gaussian distribution (GGD). ) , ( ) , ( exp / 1 1 1 ) , ( n m b c n m a n m w w w j input j input j output Results Results Fig. 1 Flow chart of the proposed method. As shown in Figure 1, the proposed method for denoising radiographic images starts by decomposition of the original image by use of the discrete wavelet transform, which results in obtaining different detail wavelet coefficients (horizontal, vertical, diagonal). The three detail coefficients are then processed by use of a sigmoid-type transfer curve for adjustment of wavelet coefficient, followed by BayesShrink thresholding. N j a ) 1 ( 2 MTFs MTFs were measured with an angled-edge method. A tungsten plate (1 mm thickness) was used as an edge device.The direction of the edge was oriented with a small angle approximately 3˚. NPSs NPSs were measured with a 2D Fourier transform method. For the calculation, the central portion of each obtained uniform image was divided into 4 non- overlapping regions, 256×256 in size (80 in total). GLCs GLCs were used to describe the relative contrast of the image, which were made using an acrylic disk (diameter : 8 mm, thickness : 8 mm) on the Burger’s phantom. C L : GLC value L acrylic : mean pixel value in acrylic disk L BG : mean pixel value in background Figure 2 shows the MTFs for the original image and the four processed images. The MTF for the sigmoid shows the highest, followed by that for the proposed method. Both the MTFs are considerably superior to the original image over the entire spatial frequency range. In contrast, the MTFs obtained from the BayesShrink and the WF methods are slightly lower than the original one. Figure 3 shows the NPS values. The NPS values for the sigmoid method had a pronounced increase as compared to that of the original image. The NPSs values for the proposed method are slightly higher and similar to those of the original image. In contrast, the NPS values for the BayesShrink and WF methods decreased as compared to those for the original images. Fig. 2 MTFs for the original image and the four processed images. Fig. 3 NPSs for the original image and the four processed images. Figure 5 illustrates visual evaluation results for the hip joint and lumbar spine at various radiation dose ratios by use of Scheffe’s method. In terms of diagnostic acceptability, the proposed method provides significantly better results compared to those for the original image up to a 64/100 radiation dose ratio in the hip joint. When the radiation dose ratio was 50/100, no significant difference was found between the image processed by the proposed method and the original image. In the lumbar radiographs, the result obtained from the proposed method was comparable to the original image up to a 64/100 radiation dose rate. Fig. 4 Gray-level contrasts as a function of the dose ratio for the original image and the four processed images obtained using a Burger phantom. Fig. 5 Visual evaluation results using Scheffe’s method of paired comparisons of the original image (Org) and various images processed by the WF, BayesShrink (Bay), sigmoid (Sig), and the proposed (Pro) methods at each radiation dose ratio with respect to the standard dose. There was a significant difference (p < 0.01) between the original and the processed image at various dose ratios if the * mark is shown. A visual evaluation A visual evaluation was conducted by five experienced radiological technologists. The images were displayed on a liquid crystal display. Statistical significance was tested using Scheffe’s method of paired comparisons Scheffe’s method of paired comparisons. Conditions for the visual evaluation, including window level, window width, and display image size were fixed. The method of paired comparisons calculates the score of each image by comparative assessment between all possible pairs of images. [%] w j input (m, n) : input value j : decomposition level w j output (m, n) : output value c=d+b×ln(a- 1.0) N : maximum decomposition level d : 25 b : 20 M: median value T B : threshold value σ 2 : noise variance Fig. 6 Original phantom images and the processed phantom images obtained using WF, BayesShrink, sigmoid, and the proposed methods for hip joint (a - e) and lumbar spine (f - j) at the standard dose. Physical Properties Measurement Physical Properties Measurement In this study, the experimental results demonstrated that the proposed method could improve the resolution and contrast characteristics while keeping the noise level within acceptable limits. Furthermore, our visual evaluation showed that an approximately 40 - 50% reduction in the exposure dose might be achieved with the proposed method. The proposed method has the potential to improve visibility in radiographs when a lower radiation dose is applied. In this paper, we propose an improved wavelet-transform-based method for offering the possibility to reduce the radiation dose while maintaining a clinically acceptable image quality. The proposed method integrates the advantages of our previously proposed wavelet-coefficient-weighted method and the existing BayesShrink thresholding method. For verifying the effectiveness of radiation dose reduction in digital radiography, the proposed method was assessed quantitatively and qualitatively. Visual Evaluation Visual Evaluation Physical Properties Measurement Physical Properties Measurement Sample image for measuring physical properties Hip Joint (Antero- Posterior) Lumbar (Lateral) Filte rs Parameters Modulation Transfer Function (MTF) Noise Power Spectrum (NPS) Gray Level Contrast (GLC) GLC GLC Visual Evaluation Visual Evaluation NPS NPS MTF MTF Sigmoid-type transfer curve for wavelet coefficient weighting adjustment (sigmoid) BayesShrink thresholding Input Image Wavelet decomposition approximat ion coefficien ts detail coefficients (horizontal) Sigmoid-Type Transfer Curves for Wavelet Coefficients Weighting Adjustment BayesShrink Output Image Wavelet reconstruction detail coefficients (vertical) detail coefficients (diagonal) Improvement & denoising stage D BG acrylic L L L L C 1 X B T 2 ) 0 , ( max 2 2 Y X 6745 . 0 M 2 2 2 X Y ・ ・ Sigmoid (Sig) Sigmoid (Sig) ・ ・ Wiener filter (WF) Wiener filter (WF) ・ ・ BayesShrink (Bay) BayesShrink (Bay) ・ ・ Proposed (Pro) Proposed (Pro) kernel = 5 × 5 matrices Daubechies, Order=4, Level=4 Daubechies, Order=4, Level=4 Daubechies, Order=4, Level=4 L D : gray level of CR system NPS (mm 2 ) Spatial frequency (mm -1 ) 1 2 3 4 5 0 10 -8 10 -4 10 -5 10 -6 10 -7 10 -9 Origin al Sigmoi d Propose d BayesShri nk WF BayesShri nk Origin al WF Sigmo id Propos ed 100/1 00 80/10 0 64/10 0 50/10 0 Radiation dose ratio 1.25×10 -2 Gray Level Contrast BayesShrin k Original WF Sigmoid Proposed × ＊ 1.23×10 -2 1.22×10 -2 1.21×10 -2 1.24×10 -2 1.20×10 -2 * * * WF 100/10 0 Hip Joint (AP) Lumbar (lateral) Radiatio n dose ratio 80/10 0 64/10 0 50/10 0 Org - 1.0 1.0 - 0.5 0 0.5 - 1.0 1.0 - 0.5 0 0.5 Pro Bay WF Si g Org Pro Bay WF Si g Org Pr o Bay WF Si g Org Si g Bay WF Pr o Org Pr o Bay WF Si g Org Pro Bay WF Si g Org Pro Bay Si g Org Pro Bay WF Si g * * * - 1.0 1.0 - 0.5 0 0.5 - 1.0 1.0 - 0.5 0 0.5 - 1.0 1.0 - 0.5 0 0.5 - 1.0 1.0 - 0.5 0 0.5 - 1.0 1.0 - 0.5 0 0.5 - 1.0 1.0 - 0.5 0 0.5 * * * * * * * * * * * * * * * * * * Original Sigmoid BayesShrin k WF Proposed ( a ) ( b ) ( c ) ( d ) ( e ) ( f ) ( g ) ( h ) ( i ) ( j ) Proposed method
description
IMPROVEMENT OF IMAGE QUALITY AND RADIATION DOSE REDUCTIONIN DIGITAL RADIOGRAPHY USING AN INTEGRATED WAVELET-TRANSFORM-BASED METHOD Haruyuki Watanabe 1 , Du-Yih Tsai 1 , Yongbum Lee 1 , Eri Matsuyama 1 , and Katsuyuki Kojima 2 - PowerPoint PPT Presentation
Transcript of BayesShrink
BayesShrink
Original WF
SigmoidProposed
MT
F
Spatial frequency (mm-1)
0.2
1.0
0.8
0.6
0.4
0.01 2 3 4 50
Original
Sigmoid
Proposed
BayesShrink WF
Figure 4 shows the GLC as a function of the radiation dose ratio. The sigmoid and the proposed methods could considerably improve the GLC at a 64/100 radiation dose ratio with respect to the standard dose, although the value of the GLC with the sigmoid and proposed methods is slightly lower than the original image at 100/100 and 80/100 radiation dose level. Furthermore, the proposed method shows the highest performance among all images at the 50/100 radiation dose ratio.
The sigmoid filter was used to enhance image contrast.The coefficients with great values and the coefficients of high resolution levels are heavily weighted.
Methods and Materials Methods and Materials
ConclusionConclusion
IMPROVEMENT OF IMAGE QUALITY AND RADIATION DOSE REDUCTIONIN DIGITAL RADIOGRAPHY
USING AN INTEGRATED WAVELET-TRANSFORM-BASED METHODHaruyuki Watanabe 1, Du-Yih Tsai 1, Yongbum Lee 1, Eri Matsuyama 1, and Katsuyuki Kojima 2
1 Department of Radiological Technology, Graduate School of Health Sciences, Niigata University, Niigata, Japan 2 Department of Business Administration, Graduate School of Business Administration, Hamamatsu University, Hamamatsu, JapanIntroductionIntroduction
In order to validate the superiority and effectiveness of the proposed method, we compared the proposed method with 3 conventional methods, namely, the Wiener filter (WF), BayesShrink method, and sigmoid-type method. In the wavelet-transform-based method, wavelet basis functions of Daubechies (order=4, level=4) were adopted for image processing. The proposed method and the above-described three methods were applied to the original images for performance comparison.
Wavelet-Based Image ProcessingWavelet-Based Image Processing
The issue of radiation dose exposure to the patients with digital radiography is a major public health concern. It is known that a trade-off exists between noise level and radiation dose. On the one hand, high-dose radiation will lower the noise level, but may give excess radiation doses to the patient. On the other hand, low-dose radiation will lower the signal-to-noise ratio of the image and result in reducing the amount of image information.
The balancing of dose and image quality should be performed explicitly to ensure that patient doses are kept as low as reasonable achievable, while maintaining a clinically acceptable image quality. To deal with this issue, much research, such as the development of new detectors and that of image processing methods has been carried out. In recent years, several investigators have reported that wavelet-based image processing techniques are effective in the reduction of radiation dose .
In the BayesShrink scheme, the threshold is determined for each subband by assuming a generalized Gaussian distribution (GGD).
),(),(
exp/11
1),( nm
b
cnmanmw w
wj
inputj
input
joutput
ResultsResults
Fig. 1 Flow chart of the proposed method.
As shown in Figure 1, the proposed method for denoising radiographic images starts by decomposition of the original image by use of the discrete wavelet transform, which results in obtaining different detail wavelet coefficients (horizontal, vertical, diagonal). The three detail coefficients are then processed by use of a sigmoid-type transfer curve for adjustment of wavelet coefficient, followed by BayesShrink thresholding.
N
ja
)1(2
MTFs MTFs were measured with an angled-edge method. A tungsten plate (1 mm thickness) was used as an edge device.The direction of the edge was oriented with a small angle approximately 3˚.
NPSs NPSs were measured with a 2D Fourier transform method. For the calculation, the central portion of each obtained uniform image was divided into 4 non-overlapping regions, 256×256 in size (80 in total).
GLCsGLCs were used to describe the relative contrast of the image, which were made using an acrylic disk (diameter : 8 mm, thickness : 8 mm) on the Burger’s phantom.
CL : GLC value
Lacrylic : mean pixel value in acrylic disk
LBG : mean pixel value in background
Figure 2 shows the MTFs for the original image and the four processed images. The MTF for the sigmoid shows the highest, followed by that for the proposed method. Both the MTFs are considerably superior to the original image over the entire spatial frequency range. In contrast, the MTFs obtained from the BayesShrink and the WF methods are slightly lower than the original one.
Figure 3 shows the NPS values. The NPS values for the sigmoid method had a pronounced increase as compared to that of the original image. The NPSs values for the proposed method are slightly higher and similar to those of the original image. In contrast, the NPS values for the BayesShrink and WF methods decreased as compared to those for the original images.
Fig. 2 MTFs for the original image and the four processed images. Fig. 3 NPSs for the original image and the four processed images.
Figure 5 illustrates visual evaluation results for the hip joint and lumbar spine at various radiation dose ratios by use of Scheffe’s method. In terms of diagnostic acceptability, the proposed method provides significantly better results compared to those for the original image up to a 64/100 radiation dose ratio in the hip joint. When the radiation dose ratio was 50/100, no significant difference was found between the image processed by the proposed method and the original image. In the lumbar radiographs, the result obtained from the proposed method was comparable to the original image up to a 64/100 radiation dose rate.
Fig. 4 Gray-level contrasts as a function of the dose ratio for the original image and the four processed images obtained using a Burger phantom.
Fig. 5 Visual evaluation results using Scheffe’s method of paired comparisons of the original image (Org) and various images processed by the WF, BayesShrink (Bay), sigmoid (Sig), and the proposed (Pro) methods at each radiation dose ratio with respect to the standard dose. There was a significant difference (p < 0.01) between the original and the processed image at various dose ratios if the * mark is shown.
A visual evaluation A visual evaluation was conducted by five experienced radiological technologists. The images were displayed on a liquid crystal display. Statistical significance was tested using Scheffe’s method of paired Scheffe’s method of paired comparisonscomparisons. Conditions for the visual evaluation, including window level, window width, and display image size were fixed. The method of paired comparisons calculates the score of each image by comparative assessment between all possible pairs of images.
[%]
wjinput (m, n) : input value
j : decomposition levelwj
output(m, n) : output value
c=d+b×ln(a-1.0)
N : maximum decomposition level
d : 25b : 20
M: median value
TB : threshold value
σ2 : noise variance
Fig. 6 Original phantom images and the processed phantom images obtained using WF, BayesShrink, sigmoid, and the proposed methods for hip joint (a - e) and lumbar spine (f - j) at the standard dose.
Physical Properties MeasurementPhysical Properties Measurement
In this study, the experimental results demonstrated that the proposed method could improve the resolution and contrast characteristics while keeping the noise level within acceptable limits. Furthermore, our visual evaluation showed that an approximately 40 - 50% reduction in the exposure dose might be achieved with the proposed method. The proposed method has the potential to improve visibility in radiographs when a lower radiation dose is applied.
In this paper, we propose an improved wavelet-transform-based method for offering the possibility to reduce the radiation dose while maintaining a clinically acceptable image quality. The proposed method integrates the advantages of our previously proposed wavelet-coefficient-weighted method and the existing BayesShrink thresholding method. For verifying the effectiveness of radiation dose reduction in digital radiography, the proposed method was assessed quantitatively and qualitatively.
Visual EvaluationVisual Evaluation
Physical Properties MeasurementPhysical Properties Measurement
Sample image for measuring physical properties
Hip Joint (Antero-Posterior)
Lumbar (Lateral)
Filters Parameters
Modulation Transfer Function (MTF)
Noise Power Spectrum (NPS)
Gray Level Contrast (GLC)
GLCGLC Visual EvaluationVisual Evaluation
NPSNPSMTFMTF
Sigmoid-type transfer curve for wavelet coefficient weighting adjustment (sigmoid)
BayesShrink thresholding
Input Image
Wavelet decomposition
approximation coefficients
detail coefficients(horizontal)
Sigmoid-Type Transfer Curves for Wavelet Coefficients Weighting Adjustment
BayesShrink
Output Image
Wavelet reconstruction
detail coefficients(vertical)
detail coefficients(diagonal)
Improvement & denoising stage
D
BGacrylicL L
LLC
1
XBT
2
)0,(max 22 YX
6745.0
M222 XY
・・ Sigmoid (Sig) Sigmoid (Sig)
・・ Wiener filter (WF) Wiener filter (WF)
・・ BayesShrink (Bay) BayesShrink (Bay)
・・ Proposed (Pro) Proposed (Pro)
kernel = 5 × 5 matrices
Daubechies, Order=4, Level=4
Daubechies, Order=4, Level=4
Daubechies, Order=4, Level=4
LD : gray level of CR system
NP
S (
mm
2 )
Spatial frequency (mm-1)
1 2 3 4 50
10-8
10-4
10-5
10-6
10-7
10-9
Original
SigmoidProposed
BayesShrink
WF
BayesShrink
Original WF
SigmoidProposed
100/100 80/100 64/100 50/100
Radiation dose ratio
1.25×10-2
Gra
y L
evel
Con
tras
t
BayesShrink
Original WF
SigmoidProposed
×
＊
1.23×10-2
1.22×10-2
1.21×10-2
1.24×10-2
1.20×10-2
*
*
*
WF
100/100
Hip Joint (AP) Lumbar (lateral)Radiation dose ratio
80/100
64/100
50/100 Org
- 1.0 1.0- 0.5 0 0.5 - 1.0 1.0- 0.5 0 0.5
Pro
BayWFSig
Org ProBayWF
Sig
Org Pro
BayWF
Sig
Org
Sig
BayWF Pro
Org
ProBay
WF Sig
Org
ProBay
WF
Sig
Org
Pro
BaySig
Org
Pro
BayWFSig
*
* *
- 1.0 1.0- 0.5 0 0.5 - 1.0 1.0- 0.5 0 0.5
- 1.0 1.0- 0.5 0 0.5 - 1.0 1.0- 0.5 0 0.5
- 1.0 1.0- 0.5 0 0.5 - 1.0 1.0- 0.5 0 0.5
*
* *
* *
*
* *
* *
* * *
*
* * * *
Original Sigmoid BayesShrink WF Proposed
( a ) ( b ) ( c ) ( d ) ( e )
( f ) ( g ) ( h ) ( i ) ( j )
Proposed method
