JP4648658B2 - Medical image processing device - Google Patents

Description

  The present invention relates to a medical image processing apparatus that processes a medical image such as an X-ray image so as to be suitable for diagnosis.

In the field of medical image processing, frequency emphasis that is performed after performing a plurality of frequency processes with different frequency characteristics has been conventionally performed as disclosed in Patent Document 1.
A plurality of pieces of image data having different frequency components for the same image are generated from the original image. Each of the plurality of image data is amplified with a variable amplification factor, and the amplified output is added. Each gain of the image data is controlled by the operator by the control unit. In order to perform frequency processing for emphasizing a high spatial frequency, non-sharp mask processing is generally performed in which a smoothed image is once obtained and subtracted from the original image. Non-sharp mask processing can be performed at high speed when a rectangular averaging mask is used as a smoothing mask, and is suitable for medical images intended for X-ray images and the like that require a large matrix size. In the non-sharp mask process, the frequency characteristic can be changed by changing the mask size of the smoothing process.

As shown in Patent Document 2, synthesis of frequency-processed images using a plurality of non-sharp mask processing methods with different frequency characteristics can also be performed by subtracting a smoothed image obtained by combining smoothed images with different frequency characteristics from the original image. Is possible.
A plurality of smoothed images are obtained from the target radiographic image. Further, one smoothed image is obtained based on the plurality of smoothed images.
The synthesized smoothed image is subtracted from the target image to obtain a difference image. In order to prevent false contours and overflow, which are side effects of frequency emphasis, the difference component is converted into a non-linearity. The converted difference image is added to the target image.
JP-A-5-242243 JP 2002-358517 A

When emphasizing the frequency component of a medical image so as to be useful for diagnosis, the optimum emphasis frequency band and the degree of emphasis differ depending on the location in one image.
For example, in the case of a chest X-ray image, enhancement of the medium frequency band is suitable on the mediastinum or ribs so as to make it easy to see a shadow of a tumor without enhancing X-ray noise or bone ridges.
On the other hand, in the lung field, since there are few X-ray noises and bone obstacle shadows, high frequency band enhancement is suitable for making it easy to see the edges of fine blood vessels and tumor shadows.
In the prior art, the enhancement frequency band and the degree of enhancement cannot be changed in accordance with the object in one image, and thus cannot meet the above requirements.

  In [Patent Document 1], the ratio of frequency-processed image synthesis is variable, but in this case as well, the same ratio is applied to the entire image. A frequency-processed image suitable for both the ribs and the lung field cannot be obtained.

  Further, since the nonlinear transformation of the difference component described in [Patent Document 2] is a function of only the difference component, in order to prevent false contours and overflow, which are side effects of frequency enhancement, the absolute difference component is used. Since it is necessary to suppress uniformly when the value is large, sufficient emphasis may not be obtained, and overflow cannot be prevented completely.

  An object of the present invention is to provide a medical image processing apparatus capable of changing an emphasis frequency band and an emphasis degree in a single image according to an object.

In the present invention, by changing the composition ratio for synthesizing a plurality of frequency processed images according to the image density, the emphasis frequency band and the degree of emphasis are changed according to the object in one image.
The relationship between the composition ratio and the image density is different for each facility that performs interpretation. Means for setting the relationship between the image density and the composition ratio. The relationship between the composition ratio and the image density varies from site to site even within the same facility. For example, a chest X-ray image and a bone image required in the shaping region are greatly different. The relationship between the image density and the composition ratio is set according to the part.

The relationship between the composition ratio and the image density varies depending on the image even in the same facility and the same part. For example, in the case of chest X-ray images, the density contrast between the mediastinum and ribs and the lung field changes depending on the body thickness and radiation quality. It is more desirable to change the relationship between the processed image composition ratio and the image density.
The relationship between the image density and the composition ratio is transformed according to the feature amount of the image. As the image feature amount, feature amounts of a plurality of ROIs corresponding to the mediastinum and lung field and position are used.
In order to apply the above to the non-sharp mask process, a plurality of smoothed images are combined at a rate that changes in accordance with the image density and subtracted from the original image to obtain a difference image.
Further, an enhanced image obtained by multiplying the difference image by a coefficient that changes in accordance with the image density is added to the original image.

Here, as the image density for changing the composition ratio, an image density which is less influenced by noise and has a global characteristic to some extent is suitable. The image density of the smoothed image is used as the image density for changing the composition ratio.
Furthermore, when an enhanced image is added to an original image, depending on the image density of the added original image and the value of the enhanced image, the change to lower density is suppressed when the image density is low, and the image density is reduced. By performing asymmetric / non-linear conversion, which suppresses the change to a higher density when it is higher. Overflow can be completely prevented without unnecessary suppression of emphasis.
Here, the ratio of the smoothed image to be synthesized is set so that the DC component of the difference image becomes zero, thereby preventing a change in the DC component due to frequency enhancement.

By changing the composition ratio for synthesizing a plurality of frequency processed images according to the image density, the emphasis frequency band and the degree of emphasis can be changed in one image according to the object.
A means for setting the relationship between the image density and the composition ratio is provided, and the relationship between the composition ratio and the image density can be set for each facility that interprets the image according to the site. By changing the relationship between the image density and the composition ratio for each image using the features of multiple ROIs corresponding to the mediastinum and lung field and position, it is possible to cope with changes in the target density to be seen for each image. Since it is realized as a non-sharp mask process, the process is faster than the conventional process.

By using the image density of the smoothed image as the image density for changing the composition ratio, the influence of noise can be reduced, and the change in the composition ratio can have some global properties.
When adding an enhanced image to an original image, depending on the image density of the added original image and the value of the enhanced image, the change to lower density is suppressed when the image density is low, and the image density is high By asymmetric / non-linear conversion, which suppresses changes to higher concentrations. Completely prevent overflow without suppressing unnecessary emphasis.
By changing the ratio of the smoothed image so that the DC component of the difference image becomes zero, a change in the DC component due to frequency enhancement can be prevented.

FIG. 1 is an overall configuration diagram of an X-ray imaging system using a medical image processing apparatus 1 of the present invention.
X-rays generated by the X-ray generator 2 pass through the subject, enter the medical image processing apparatus 1, and are converted into digital image data by the image input unit 11 including the X-ray plane sensor 111 and the image data collection circuit 112. The image is processed by the unit 10 and output as a display image by the image output unit 12 including the display memory 121 and the liquid crystal display 122.

FIG. 2 is a block diagram showing the configuration of the image processing unit 10 of the medical image processing apparatus 1 of the present invention.
First, image data input from the image input unit 11 is stored in the original image memory 101.
The process starts when the operator selects a part with the operation / control means 102.

The operation / processing unit 102 sets a larger mask size (mask size 30) to the smoothing processing unit 103 by the mask size setting 1021, and the output of the switch 104 is set to the smoothed image memory (1) 105 side by the memory switching 1022. After that, the smoothing processing means 103 is operated to store the smoothed image M1 having a larger mask size in the smoothed image memory (1) 105.
Here, the smoothing processing means 103 performs smoothing processing by a known moving average method.

  The operation / processing unit 102 sets a smaller mask size (mask size 10) to the smoothing processing unit 103 by the mask size setting 1021, and the output of the switch 104 is set to the smoothed image memory (2) 106 side by the memory switching 1022. Thereafter, the smoothing processing means 103 is operated to store the smoothed image M2 having a smaller mask size in the smoothed image memory (2) 106.

  The operation / processing means 102 inputs the image data 1061 of the smoothed image M2 and the part selection 1024 to the composite coefficient creating means 107, and the images of the coefficients f1 and f2 used in the enhanced image creating means 108 by the method described later. Two types of LUTs that define data dependence are created in the synthesis coefficient creation means 107.

The operation / processing unit 102 simultaneously converts the original image data O from the original image memory 101, the smoothed image M1 from the smoothed image memory (1) 105, and the smoothed image M2 from the smoothed image memory (2) 106 for each pixel. By inputting the image data 1061 of the read and smoothed image M2 into the two types of LUTs created earlier in the synthesis coefficient creation means 107, the coefficients f1 and f2 1071 are obtained, and from these, the enhanced image creation means The enhanced image data V is obtained by performing the calculation according to Equation 1 at 108:
V = (f1 + f2) O-f1 * M1-f2 * M2 (Formula 1)
Here, by setting the coefficient of O to (f1 + f2), the DC component of the enhanced image is set to zero to prevent a change in DC component due to frequency enhancement.

The enhanced image data V is input to the asymmetric / nonlinear conversion means 109 together with the original image data O, and the equation 2
Get converted by:
V ′ = γ × h (V / γ) (Formula 2)
Here, V ′ is the output of the asymmetric / nonlinear conversion means 109.
γ is obtained according to Equation 3 from the original image data O, the upper limit Omax of the output image data, and the lower limit Omin of the output image data according to the sign of the enhanced image data V:
γ = Omax−O (when V ≧ 0); (Formula 3)
γ = O−Omin (when V <0);

Further, h is a function defined as shown in Equation 4 based on the arctangent function.
h (x) ≡ (2 / π) arctan (πx / 2) (Formula 4)
This function has a value in the interval (−1, 1), and has the property of Equation 5 near 0.
h (x) ≒ x (Formula 5)

The output V ′ of the asymmetric / nonlinear conversion means 109 is added to the original image data O by the adder 110 to be frequency-enhanced repair image data (O + V ′) and output to the image output unit 12.
From the range of (Expression 2), (Expression 3), and h (x), O + V ′ is a value in the range of (Omin, Omax), and thus no overflow occurs.
It can be seen that when γ is larger than the properties of (Equation 2) and (Equation 5), Equation 6 is obtained.
V '≒ V (Formula 6)
This means that the emphasis for increasing the density is hardly suppressed in the region where the image density is low, and the emphasis for decreasing the density is hardly suppressed when the image density is high. That is, unnecessary suppression of emphasis does not occur.

FIG. 3 is a block diagram showing a configuration of the synthesis coefficient creating means 107 in the image processing unit 10 of the medical image processing apparatus 1 of the present invention.
The composite coefficient creating means 107 includes an ROI (1) setting table 1701, an ROI (2) setting table 1702, a constant A and B setting table 1703, a coefficient F1 LUT group 1705, and a coefficient F2 LUT group 1706. The ROI (1) setting data, the ROI (2) setting data, the constants A and B values, the coefficient F1 LUT, and the coefficient F2 LUT are stored for each part.
The setting for each part and the LUT are written in advance for each facility by the table setting 1023 from the operation / control means 102.

  Prior to using the coefficients f1 and f2 1071 in the enhanced image creating unit 108, the operation / processing unit 102 inputs the image data 1061 of the smoothed image M2 and the region selection 1024 to the synthesis coefficient creating unit 107 to enhance the image. Two types of LUTs that define the image data dependency of the coefficients f1 and f2 used by the image creation means 108 are created.

ROI (1) setting table 1701 , ROI (2) setting table 1702, constant A, B setting table 1703, coefficient F1 LUT group 1705, coefficient F2 LUT group 1706 ROI (1) setting data by site selection 1024, ROI (2) setting data, constant A and B values, LUT with coefficient F1, and LUT with coefficient F2 are selected according to the part.
The ROI (1) maximum value detecting means 1707 is operated according to the selected ROI (1) setting data, and the maximum value P of ROI (1) in the image data 1061 of the smoothed image M2 is obtained.

ROI (1) is set in the lung field in the case of a chest image, for example.
The smoothed image M2 is used here because it is image data that is less affected by noise and that leaves some variation in image data.
The ROI (2) minimum value detecting means 1708 is operated according to the selected ROI (2) setting data, and the minimum value Q of ROI (2) in the image data 1061 of the smoothed image M2 is obtained.

ROI (2) is set to the mediastinum in the case of a chest image, for example.
The constants A and B are set to standard values of the maximum value P and the minimum value Q, and this is different for each image. From the P and Q, the LUT conversion parameter calculation means 1709 performs LUT conversion using the following formula 7. Parameters α and Δ are determined:
α = (PQ) / (AB) (Formula 7)
Δ = Q / α-B

  The LUT conversion parameters α and Δ are input to the LUT conversion means 1710, the input side of the LUT of the coefficient F1 and the LUT of the coefficient F2 selected from the LUT group 1705 of the coefficient F1 and the LUT group 1706 of the coefficient F2 is multiplied by α, and Δ By moving only in parallel, an LUT 1712 having a coefficient f1 and an LUT 1713 having a coefficient f2 are obtained.

  By doing this, the output corresponding to the input A of the LUT with the coefficient F1 and the LUT with the coefficient F2 and the output with respect to the input B are the output with respect to the input P and the output with respect to the input Q of the LUT 1712 with the coefficient f1 and the LUT 1713 with the coefficient f2, respectively. It will be obtained.

  Original image data O is read from the original image memory 101, the smoothed image M1 is read from the smoothed image memory (1) 105, and the smoothed image M2 is read from the smoothed image memory (2) 106 for each pixel at the same time to create an enhanced image. When the enhanced image data V is obtained by the means 108, the image data 1061 of the smoothed image M2 is input to the LUT 1712 of the coefficient f1 and the LUT 1713 of the coefficient f2 in the synthesis coefficient creating means 107, whereby the coefficients f1 and f2 1071 are obtained. obtain.

FIG. 4 shows an example of the LUT with the coefficient F 1 and the LUT with the coefficient F 2 designed for the chest image as a graph representing the input / output relationship.
By doing this, in the vicinity of the mediastinum corresponding to the input B, the middle band is emphasized by the relatively large mask M1 by the positive F1 value, and the high band is emphasized by the negative F2 value. F1 is reduced because it is suppressed and is a relatively noisy area.

  In the vicinity of the lung field corresponding to the input A, the F1 value is set to 0 and the F2 value is set to 1, so that the enhancement band is shifted to a high frequency range so that a finer structure is emphasized. In addition, F2 is increased because it is an area with relatively little noise.

1 is an overall configuration diagram of an X-ray imaging system using a medical image processing apparatus of the present invention. The block diagram which shows the structure of the image process part of the medical image processing apparatus of this invention. The block diagram which shows the structure of the synthetic | combination coefficient preparation means in the image processing part of the medical image processing apparatus of this invention. 6 is an example of a graph showing an input / output relationship of a coefficient G LUT, a coefficient F1 LUT, and a coefficient F2 LUT set in a synthesis coefficient creating unit in the image processing unit of the medical image processing apparatus of the present invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 Original image memory 102 Operation / control means 103 Smoothing processing means 104 Switching device 105 Smoothed image memory (1)
106 Smoothed image memory (2)
107 Synthesis coefficient creation means 108 Enhanced image creation means 109 Asymmetric / nonlinear transformation means 110 Adder 1021 Mask size setting 1022 Memory switching 1023 Table selection 1024 Region selection 1061 Image data of smoothed image M2 1071 Coefficients f1 and f2

Claims (4)

Frequency enhancement processing for generating a single composite image using a medical image generated by the medical image generation means and a plurality of smoothed images having different frequency characteristics obtained by performing frequency processing with different frequency characteristics on the medical image In a medical image processing apparatus comprising means,
A medical image processing apparatus, wherein a ratio of synthesis of the plurality of smoothed images having different frequency characteristics is changed according to a density value of the medical image.   Emphasized image using a medical image generated by the medical image generation unit, the plurality of smoothed images having different frequency characteristics, and a coefficient indicating the ratio of the plurality of smoothed images having different frequency characteristics. An enhanced image creating means for computing data, the enhanced image data computed by the enhanced image creating means, and the medical image are converted by an asymmetric and non-linear converting means to generate one composite image The medical image processing apparatus according to claim 1, further comprising a frequency enhancement processing unit. Two region of interest setting data selected in accordance with more sites to the site selection means, and a maximum value detecting means and the minimum value detection means for respectively operating on the basis of the two regions of interest setting data said selected, the maximum value detecting means And a maximum value and a minimum value of the image data in one smoothed image among the plurality of smoothed images having different frequency characteristics based on the minimum value detecting means, the maximum value and the minimum value, and the maximum value The medical image processing apparatus according to claim 1, wherein a value of a plurality of coefficients indicating the ratio is obtained in the ratio of the synthesis using the standard value of each of the value and the minimum value.   Of the two coefficients indicating the ratio, one coefficient is determined using a coefficient that decreases in value as the density value of the medical image increases in the ratio of the smoothed images having different frequency characteristics. 2. The medical image processing apparatus according to claim 1, wherein the other coefficient is determined by using a coefficient that increases as the density value of the medical image increases.

Images