KR20140090311A - Method for Dose reduction in full-field digital mammography using total variation based noise removal algorithms - Google Patents

Abstract

The present invention relates to a method for reducing a dose in a full-field digital mammography through a noise removal algorithm based on total variation. More particularly, the present invention is to minimize an average glandular dose (AGD) given to a patient by reducing an amount of radiation, which is generated when a digital mammography device is used, through the total variation noise removal algorithm. The present invention suggests the total variation algorithm to maintain breast information as much as possible and to remove the noise. According to the present invention, the AGD given to the patient at 35 kVp, W/Rh, or (AGD; 2.77 mGy), which represents the worst condition, is represented as 0.88 mGy, which is reduced by 2.10 mGy.

Description

[0001] The present invention relates to a method for reducing a dose in a mammography apparatus using a noise canceling algorithm based on a full-

The present invention relates to a dose reduction method in a full-field digital mammography (FFDM) using a noise reduction algorithm based on total variation. More particularly, the present invention relates to a method for minimizing the average glandular dose (AGD) received by a patient by reducing the amount of radiation generated when a digital mammography apparatus is used by applying a noise elimination algorithm based on a full-

Mammography is the best method of detecting untamaged breast cancer. Because the difference in the degree of X-ray absorption between normal and diseased tissues in breast is smaller than in other body parts examined by X-rays, an ancient illumination image is needed to maximize this difference. Therefore, the mammography camera has lower tube voltage (the amount of X-ray generated when the tube voltage is lowered) is lower than that of a general X-ray apparatus, and the target of the camera is a target It uses molybdenum instead of tungsten, and shoots with a pressing mechanism when shooting.

Full-field digital mammography (FFDM) has made it possible for women to perform breast cancer screening more efficiently and to screen for far more cases than conventional screen / film mammography (SFM) Could

It is the basic photography to take one breast in the mammogram and two times in the up and down direction and the inside and outside direction in order. When the breast is compressed enough to perform the test, the amount of radiation necessary for proper image quality is reduced and the contrast is improved

The breast is composed of mammary tissue, fibrous tissue, and adipose tissue. The breast is classified according to the relative amount and distribution of high density mammary tissue and low density lipid tissue. Young women have dense breasts with high density of abundance of fibrous mammary tissues and gradually become fat substitutes after menopause to become fatty breasts. In women with dense breasts, there is no distinction between normal and normal cancer tissues. When asymmetric shadows are seen, it is difficult to distinguish between normal breast tissue and unilateral breast cancer. In this case, breast cancer can be detected by breast ultrasound.

In mammography, the higher the exposure dose, the better the image quality. The lower the dose, the worse the image quality. Therefore, in mammography, the patient is exposed to a certain amount of radiation dose. In recent medical diagnosis process, since the frequency of exposing the patient to various radiography apparatuses such as CT is increased, Research is being attempted. However, there is a problem that the reduction of the radiation dose to be irradiated involves the deterioration of the image quality, and a method which can reduce the radiation dose while acquiring a good image is needed.

It is an object of the present invention to provide a method for minimizing the average amount of wire radiation exposed to a patient while preventing deterioration of image quality in mammography.

The present invention provides a method of reducing dose in a mammography apparatus using a noise-canceling algorithm based on a full-variable approach.

The present invention minimizes the average grandiose dose (AGD) that a patient receives by reducing the amount of radiation generated when using a digital mammography apparatus by applying a total variation noise elimination algorithm.

In the present invention, a total variation algorithm has been proposed to maintain maximum breast information and to remove noise.

According to the present invention, the AGD of the patient was 0.88 mGy at 35 kVp, W / Rh, (AGD; 2.77 mGy), which is the worst image quality. 2.10 mGy decreased

According to the present invention, it is possible to provide a method for minimizing the average wired dose amount exposed to a patient while preventing deterioration of image quality in mammography.

Particularly, through the present invention, the optimal exposure parameter is examined and the average wired dose (AGD) that the patient receives at diagnosis can be reduced by applying a noise elimination filter.

Fig. 1 Digital images of Phantom acquired by through appearance and FFDM used in the experiment
a) The phantom is 4.5 cm thick, simulates a 50% glandular tissue composition
b) Embedded details
c) Digital image of phantom to be used in the experiment
Figure 2 Example of extracted mass on a real phantom and denoising
a) Image with maximum noise value (W / Rh, 35 kVp)
b) Image with minimum noise value (Mo / Mo, 23 kVp)
c) Image of results after denoising processing (W / Rh, 35 kVp)
d) Image of results after denoising processing denoising (Mo / Mo, 23 kVp)
Figure 3. Example of extracted mass on a real phantom and denoising
a) Image with maximum noise value (w / rh, 35 kVp)
b) Image with minimum noise value (mo / mo 23 kVp)
c) Image of results after denoising processing (w / rh, 35 kVp)
d) Image of results after denoising processing (mo / mo, 23 kVp)
Figure 4. Example of extracted microcalcification from extracted mass on a real phantom
a) Image with maximum noise value (w / rh, 35 kVp)
b) Image with minimum noise value (mo / mo, 23 kVp)
c) Image of results after denoising processing (w / rh, 35 kVp)
d) Image of results after denoising processing (mo / mo, 23 kVp)
Figure 5. Suggested image of phantom is showing microcalcification and mass equivalent step wedged. Quadrangle box is used for the calculation of signal and noise.
Figure 6. Distribution of signal values obtained from the mass equivalent step wedge for each filtration-target material
Figure 7. Distribution of noise values obtained from the mass equivalent step wedge for each filtration-target material
Figure 8. Distribution of contrast values obtained from the mass equivalent step wedge for each filtration-target material
Figure 9. Distribution of SNR values obtained from the mass equivalent step wedge for each filtration-target material
Figure 10. FOM results to find an optimal exposure parameter
Figure 11. Distribution of AGD values according to filtration-target material and kV _p , when using FFDM system
12. Comparison of a denoising filter with median filter and wiener filter

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Although the configuration of the present invention is described in connection with the embodiment, such an embodiment describes an example for realizing the present invention concretely, and does not limit or limit the technical idea on which the constitution of the present invention is based.

Digital detectors used in full-field digital mammography (FFDM) have linear response characteristics over a wide range of radiation doses and can be used manually or automatically by setting the exposure parameters of digital mammography .

FFDM systems with automatic exposure control (AEC) function automatically calculate the mAs by the thickness of the breast, and in many cases select the tube voltage (kVp), filter-target material It is designed for use.

This function provides the possibility for the possibility of dose reduction. Decreasing dose by lowering mAs in digital mammography results in increased noise and poor detection of microcalcifications. However, it is very important to use the optimal exposure parameters because it is important to consider the average Grand Dose (AGD) of the patient exposed at the hospital, together with the effort to obtain optimal image quality, kVp). On the other hand, attempts to improve image quality through noise cancellation will be of great help in reducing AGD.

In the present invention, noise, contrast, signal to noise ratio (SNR), and figure of merit (FOM) are measured to understand the optimal filtration-target material and tube voltage (kVp) , And noise reduction methods, thereby improving the signal-to-noise ratio of the image.

<Examples>

1. Imaging Protocol (Imaging Protocol )

In the embodiment of the present invention, Siemens Mammomat Inspiration (Siemens, Erlangen, Germany) was used to acquire images. Embodiments of the present invention were performed in order to focus on the combination of filtration-target materials with Mo / Mo, W / Rh and Mo / Rh concentrations of about 0.3 and 3.

FFDM used in hospitals has passive mode and AEC mode, where kVp and filtration-target material are designed to be used selectively when AEC mode is used. The present embodiment examines the optimal exposure variable that can reduce the AGD when the AEC mode is used in hospitals, and examines the result of applying the noise reduction method.

Images used in this example were collected using the human model of Nuclear Associates Model 18-222 (CIRS, Virginia, USA) and the AEC mode of the FFDM system. At the same time, entrance surface air kerma (ESAK) and AGD were measured and are automatically indicated in the system. Then, the anatomical model was fixed at the same point and the image was obtained from several voltages (23 ~ 35 kVp). At the end of one embodiment, the experiment was performed by changing the filtration-target material and the tube voltage (kVp). After applying the noise reduction algorithm, we evaluated the obtained test image.

2. Human body model phantom )

A human model (CIRS, Virginia, USA) as a human model equivalent to the mammogram used in the examples is detailed in FIG. 1 and Table 1.

Specification of phantom Index Specification Line pair target    1: 20 lp / mm Grain you  ( mm ) 2: 0.130 3: 0.165 4: 0.196 5: 0.230 6: 0.275 7: 0.400 8: 0.230 9: 0.196 10: 0.165 11: 0.230 12: 0.196 13: 0.165 Step Wedge  (One cm thick )  14: 100% gland 15: 70% gland 16: 50% gland 17: 30% gland  18: 100% adipose Nylon Fibers
( diameter mm ) 19. 1.25 20: 0.83 21: 0.71 22: 0.53 23: 0.30 Hemispheric Masses
75% glandular / 25% Adipose
( thickness mm ) 24: 4.76 25: 3.16 26: 2.38 27: 1.98 28: 1.59 29: 1.19 30: 0.90 Dimensions
( Lengthwidthhight cm ) 12.518.54.5

The anatomical shape is hemispherical in shape with seven lumps inside and is filled with wax-embedded nylon fiber. The anatomy is designed to test the performance of the mammography system by a quantitative assessment of system performance so that a 4.5 cm thick, 50% linear organization is present and a small image similar to that found clinically.

3. Total Variation (TV)  Used noise  How to uninstall

The uncorrelated noise in the FFDM system affects the SNR of the mammogram. Uncorrelated noise includes x-ray quantum noise, electronic noise, and thermal noise, which includes the detector's fundamental and distributed noise. For optimum noise removal, a filter having non-linear characteristics is required, and the present invention proposes a total variation method.

는 함수의 2-norm을 나타낸다.An image can be interpreted as a real function defined as the bounded and open domains of OMEGA.

Represents the 2-norm of the function.

는 관찰된 이미지

는 오리지널 이미지이다.Our concern is image noise reduction. The noise reduction process is to recover the edges of the image.

The observed image

Is the original image.

   The model of degradation is shown in Equation 2. &lt;

는 노이즈(noise)

Noise,

=

는 부적절하게 정립된 문제이다. 이는 문제 선량이 데이터 z에 종속되지 않는 것을 의미한다. 이 문제에 대하여 수학식 3은 두 개의 제약 문제(constrained problem)를 보여준다. Generally inversion problem

=

Is an improperly established problem. This means that the problem dose is not dependent on the data z. Equation 3 for this problem shows two constrained problems.

Therefore, the noise removal problem can be created as shown in Equation (4).

와

를 실함수로 정의하고,

를 최소화하면,

Wow

Is defined as a real number,

If minimized,

는 수학식 6을 만족해야 한다.after that

Must satisfy Equation (6).

This equation is called the Euler-Lagrange equation. Then, Equation 7 can be reached.

에서

.

in

.

This solution procedure uses a parabolic equation for time as an evolution parameter.

　가 계산되어야 한다. 정상상태에 도달하면 왼쪽항은 사라진다. 결국 수학식 9을 얻을 수 있다. As t increases, the target can be reached. In other words,

Should be calculated. When the steady state is reached, the left term disappears. Consequently, Equation 9 can be obtained.

4. Image evaluation and dose reduction

The mammography needed to evaluate the image to evaluate the image quality in the acquired image. The image signal can be obtained by calculating the information of the regions of interest (ROI), which is shown in FIGS. 5 and 10-14 Lt; / RTI &gt;

To obtain the average signal intensity from the anatomy, 90 x 90 pixels were used in all ROIs. The position of ROI 1 is set in a step wedge that is present in the phantom present in the human body model, ROI 2 is set to the same size as the background over ROI 4 around ROI 2, And is set to 260 x 220 at the point where it is made. In general, the signal value is obtained by having a difference between the step wedge ROI 1 and the background ROI 2. However, regions such as ROI 1 and ROI 4 are added and calculated to correct the Heel Effect and background trends (see Equation 6). The noise value is obtained by taking the square root of 2 with the standard deviation value of the ROI 5 pixel value. The contrast is calculated for the difference between the step wedge ROI 1 and the background ROI 2. It is standardized by average background intensity. The contrast is always high, which is positive. The SNR is indicated by the ratio of the signal power to the noise power. When the SNR is higher than 1, it indicates that there is more signal than noise.

In the present invention, the performance index (FOM) proposed by Williams was employed to find a tradeoff between SNR and AGD. The FOM is an index used to determine the maximum SNR while minimizing the AGD. The higher the result, the better the value. While the SNR is used to evaluate the noise reduction or the parameters for the proposed algorithm, the FOM should be considered in minimizing the AGD.

5. Image quality evaluation

The signal is indicated by the signal strength of the pixel. If the signal value is large, a good quality image is obtained and the contrast and SNR value are increased.

FIG. 6 (a) shows the results of signal strength according to the combination of filtration and target material and selective kVp before noise removal is applied. Signal strength was best at Mo / Mo 23 kVp and decreased with increasing kVp. The Mo / Mo and Mo / Mo signals appear to decrease rapidly, while the Mo / Mo and Mo / Mo signals appear to be close to 26kVp. Fig. 1 (b) shows the result of applying noise removal, which is similar to the result before application. This shows that noise rejection does not affect signal strength.

Noise causes lumps or microcalcifications to be difficult to find. In Fig. 7 (a), the noise increases gradually as the voltage increases. Noise is lowest in W / Rh and 0.34 higher on average than Mo / Rh. In Fig. 7 (b), the results after noise removal show that the distribution of results is similar to that before application, and the noise value of the filtration and target material combination decreases significantly from 41.5 to 2.13 as a whole.

8 (a) and 8 (b) show the results before and after noise removal. The contrast is not changed, while the signal is sharply reduced by noise reduction (Figure 7), and signal strength is not increased when calculated using specific signal results (Figure 6).

The SNR is calculated using the signal value and the noise value result. This can be good information for the verifier because the SNR is clearly seen to be high. 9 (a), the SNR was highest at 6.2 at Mo / Mo 23 kVp and the SNR decreased with increasing kVp. Mo / Rh values remained highest at over 28 kVp and SNR was lowest at 3.9 fh at 35 kVp.

In FIG. 9 (b), noise rejection was applied and the SNR was lowest at 3.89 from 35 kVp. Which is higher than W / Rh and Mo / Rh by 31 kVp. W / Rh has a lower tube voltage than Mo / Mo and the lowest at 35 kVp to 6.58. This is similar to the result of 23 kVp of Mo / Mo, where the SNR is most noticeable before noise removal is applied.

Figures 2 and 4 illustrate the results of using extracted mass samples and microcalcifications as visual effects. We can see that the proposed noise reduction improves picture quality in W / Rh (35 kVp) with the maximum noise value and Mo / Mo (23 kVp) with the minimum noise value. FIG. 3 shows specific resolution information using profile information of extracted mass samples and demonstrates objectively the same result.

6. For optimal exposure parameters FOM  Calculation

SNR and AGD are considered simultaneously, and the FOM can obtain optimal exposure parameters to minimize the radiation dose of the patient.

According to the example, FOM was highest at 23 kVp of W / Rh, 23.95 before noise removal application. The current SNR has dropped from 6.22 to 4.59, and the AGD has decreased from 2.77 mGy to 0.88 mGy. The FOM results were then as high as 18.25 at Mo / Rh 26 kVp and 16.35 at 27 kVp for Mo / Mo, respectively.

Considering AGD after noise removal, FOM was the highest at 89.48 at W / Rh 29 kVp. At this time, the SNR improved from 6.22 to 9.57 and the AGD decreased from 2.77 to 0.88, showing the results for Mo / Rh and Mo / Mo, and the FOM of Mo / Rh was slightly larger than 24-26 kVp, but above 27 kVp It appears similar.

To initially consider minimization of radiation exposure dose to the patient, W / Rh 35 kVp can be commercialized (see Figure 11), with an AGD of 0.67. When the noise removal is applied (noise 2.37, SNR 9.73), better image quality is obtained than Mo / Mo 23 kVp (noise 3.77, SNR 6.23) which shows the best SNR before application and AGD is reduced to 2.10 mGy after noise removal process .

7. Denoising  Comparison of methods

The embodiment of the present invention is compared with the previously proposed median filter and Wiener filter by the TV noise reduction method. Noise was worst with an average value of 2.6 ± 0.13 for the Wiener filter, and was the best with an average value of 1.56 ± 0.07 for the method of the present invention. In the SNR results of FIG. 12, the Wiener filter was the worst at an average value of 7.69 +/- 7.47, and was the best at an average value of 12.83 +/- 0.1 in the method of the present invention. In FIG. 12 (c), the FOM with the ADG value was the best in the method of the present invention with an average value of 183.73 ± 22.37, followed by 80.21 ± 7.47 and median filter with 65.98 ± 6.02 wiener filters.

Comparison of results before denoising filter processing for image at 35kV _p of W / Rh with the smallest AGD Filter Signal Noise Contrast SNR Original 16.76 4.68 0.19 3.58 Wiener 16.78 2.90 2.89 5.80 Median 16.79 2.47 0.19 6.79 Proposed Method 16.72 1.72 0.19 9.73

AGD summary for the digital mammography system obtained at various voltages. Mo / Mo (mGy) W / Rh (mGy) Mo / Rh (mGy) ESD MGD ESD MGD ESD MGD 12.3 2.77 3.1 1.32 7.1 2.19 10.5 2.42 2.6 1.06 5.8 1.81 8.7 2.1 2.3 1.05 5.4 1.72 8 1.98 2.3 0.99 4.9 1.57 7.1 1.79 2.2 0.95 4.6 1.5 6.4 1.63 2.1 0.92 4.3 1.45 5.8 1.51 2 0.88 4.1 1.39 5.3 1.39 1.9 0.85 3.9 1.34 4.8 1.3 1.8 0.81 3.8 1.29 4.5 1.22 1.7 0.78 3.6 1.25 4.2 1.16 1.6 0.74 3.5 1.21 4 1.13 1.5 0.71 3.3 1.18 3.8 1.06 1.4 0.67 3.2 1.15

As described above, an optimal embodiment has been disclosed in the drawings and specification. Although specific terms have been employed herein, they are used for purposes of illustration only and are not intended to limit the scope of the invention as defined in the claims or the claims. Therefore, those skilled in the art will appreciate that various modifications and equivalent embodiments are possible without departing from the scope of the present invention. Accordingly, the true scope of the present invention should be determined by the technical idea of the appended claims.

Claims (1)

Wherein the noise reduction algorithm is applied to reduce the radiation dose during the photographing.

Images