CN105701777B - A kind of spiral-fault radiotherapy image quality improving method - Google Patents

Abstract

A method for improving the image quality of helical tomotherapy, which relates to a method for improving the quality of helical tomotherapy images based on retina-cerebral cortex theory, it has four major steps: Step 1: computer reads the image of helical tomotherapy under MATLAB environment ; Step 2: Perform bilateral filtering and denoising on the image; Step 3: Enhance the contrast of the image based on the retina-cerebral cortex theory; Step 4: Use the Gauss-Segel iterative method to enhance the edge of the image. The invention solves the problems of high noise, poor contrast and unclear edges of the original spiral tomographic radiotherapy image, achieves better quality enhancement results, and has broad application prospects in the field of spiral tomographic therapy.

Description

A method for improving the image quality of helical tomotherapy

(1) Technical field:

The invention provides a method for improving the image quality of helical tomotherapy, which relates to a method for improving the quality of helical tomotherapy images based on the retina-cerebral cortex theory, and belongs to the field of medical image processing.

(two) background technology:

With the development of information technology and medical imaging technology, medical image processing plays an increasingly important role in medical clinical and scientific research, and strongly promotes the progress of medical scientific research and clinical treatment. Among them, image processing technology for CT (Computed Tomography, ie computerized tomography) images plays an important role. However, research on image processing for mega-voltage computed tomography (MVCT, Mega-Voltage Computed Tomography) images for tomotherapy is still at an early stage. Helical tomotherapy is a cancer radiotherapy method based on helical tomotherapy equipment, and it is the most advanced tumor radiotherapy technology today. Helical tomotherapy is image-mediated three-dimensional intensity-modulated radiation therapy, which integrates linear accelerator and helix, integrates treatment planning, patient positioning and treatment process, and can treat different target areas, from stereotactic treatment to small The treatment from the tumor to the whole body is completed by a single helical ray beam. Through the MVCT images obtained from each treatment, the tumor dose distribution and the changes of the tumor during the treatment can be observed, and the treatment plan of the target volume can be adjusted in time. It has the incomparable advantages of conventional accelerator radiation therapy, opens up a new treatment platform for radiation therapists, and occupies an important position in the development history of intensity-modulated radiation therapy. However, the imaging of MVCT images obtained with therapeutic rays has some limitations: Compared with common diagnostic CT scanners, MVCT requires a compromise between administered dose and imaging effect. That is, the noise, brightness balance, contrast and resolution of the image will all be affected by the dose, thus affecting the doctor's accurate judgment of the patient's condition.

At present, there are limited methods for processing helical tomotherapy images, and most of the methods in CT image processing are directly transplanted to the processing of helical tomotherapy images. Common ones include pixel-level linear and nonlinear function operations, histogram equalization algorithms, mean filtering and median filtering methods with denoising effects, relatively recent adaptive filtering, wavelet shrinkage and other methods, and methods based on histogram equalization. Improved algorithms and center/surround methods based on background variance and more. These methods are universal methods for CT images and even ordinary images, and it is difficult to perform image processing for the characteristics of helical tomotherapy images. The present invention uses a bilateral filtering method, and enhances the helical tomographic radiotherapy image according to the retina-cerebral cortex (Retinex) theory. Some helical tomotherapy images are enhanced to achieve the desired effect.

(3) Contents of the invention:

1. Purpose: The purpose of the present invention is to provide a method for enhancing the image quality of helical tomotherapy based on the retina-cerebral cortex theory. The image quality is improved to solve the problems of more noise, low contrast and unclear edges in the helical tomotherapy image.

2. Technical solution: the present invention is achieved through the following technical solutions:

The invention provides a method for improving the image quality of helical tomographic radiotherapy based on the retina-cerebral cortex theory, which is a method for improving the image quality of helical tomographic radiotherapy images by using the bilateral filtering method, the retina-cerebral cortex theory and the Gauss-Seidel iterative method Methods. The concrete steps of this method are as follows:

Step 1: first output the computer digital image through the helical tomotherapy instrument, then use the imread function in the Matlab language to read the image, and convert its information into a Matlab matrix form, so that the Matlab language can process it;

The spiral tomotherapy image in the present invention is a digital image of 512 pixels*512 pixels*c channel, that is, the matrix data read in is 512*512*c dimension; wherein the symbol c represents the number of faults contained in the image, and each fault It is a gray-scale image, and the tomographic images of the c-channel number are stacked to form a complete helical tomotherapy image; this method does not use the correlation information between channels, so the following steps are all completed on a single tomographic image; for the convenience of explanation, The capital letter I is used below to represent a certain channel of the matrix data after the image is read by Matlab (that is, a matrix of 512*512 dimensions); the superscript number indicates the intermediate image matrix data of different steps in this method, such as I ⁰ represents the original Image matrix data; use a subscript to represent a certain element in the matrix (the pixel on a certain coordinate in the corresponding image), that is, I _{x, y} represent the element at the coordinates (x, y) in the matrix I;

Step 2: In the Matlab software, perform the following operations on the matrix (i.e. I ⁰ ) read in in the previous step:

Among them, Ω represents the neighborhood range of (x, y) coordinates, and coordinates (i, j) satisfy (i, j) ∈ Ω; the value of the weight coefficient w is the range weight and spatial domain weights The product; this step is the bilateral filtering operation of the image, which can have a good denoising effect while maintaining the edge of the image;

Step 3: In Matlab software, perform the following operations on the matrix in the previous step:

Among them, a, b, c are hyperparameters, set manually, and have:

Here, the symbol * represents convolution, and the symbol G represents the Gaussian kernel, that is, the denominator is the image blurred by the numerator Gaussian kernel; this step is the improved Retinex algorithm, which does not use the logarithmic form used in the previous common Retinex method, but is to use a new curve function form;

Step 4: At first, use the Sobel operator function in Matlab, matrix I ² is operated, and the edge image matrix that obtains is marked as Gra;

Then, the edge image matrix Gra is traversed pixel by pixel to find the maximum value of the edge in the neighborhood, and the matrix Gra _M is obtained to mark whether the current position is the maximum value of the neighborhood in the matrix Gra, and then pass:

Update the matrix Gra, where f _{x, y} is a self-defined function that takes the distance from the current point (x, y) to the maximum point of the nearest edge as an independent variable, and the value of this function is greater than 1; when the distance is larger, it is less than 1, so that the distribution of the edge in the edge image corresponding to the new Gra ¹ can be narrowed and the peak value can be increased;

Finally, using the Gauss-Seidel iterative method, the matrix Gra ¹ is used to deduce the edge-enhanced image, and the steps are as follows:

1. Carry out the differential operation again on the matrix Gra ¹ to obtain the second-order derivative image matrix Lap of the target image;

2. Initialize matrix dst ⁰ = I ²

3. Repeat the following iterations until the number of iterations reaches 5, where n is the number of iterations, n=0,1,2...:

1)

2)

After the above steps are iterated, the final image is obtained, and the quality improvement of the helical tomotherapy image is completed.

3. Advantages and efficacy: This method simultaneously enhances the helical tomotherapy images by bilateral filtering, improved retina-cerebral cortex theory and Gauss-Seidel iterative method. It can be seen from Figure 2 that the pixel gray value change frequency of the processed image decreases and the change range increases, which shows that this method can not only effectively remove the noise in the image, but also improve the contrast, and at the same time make the edge of the image clearer.

(4) Description of drawings:

Fig. 1 is a flow chart of the method of the present invention.

Fig. 2 is a comparative diagram of the method effects of the present invention.

(5) Specific implementation methods:

A method for improving the image quality of spiral tomotherapy based on the retina-cerebral cortex theory of the present invention is shown in Figure 1, and its steps are as follows:

Step 1: first output computer digital images through the helical tomotherapy apparatus, then use the imread function in Matlab language to read the image, and convert its information into Matlab matrix form, so that Matlab language can process it.

The helical tomotherapy image in the present invention is a digital image of 512 pixels*512 pixels*c channel, that is, the read-in matrix data has dimensions of 512*512*c. The symbol c represents the number of slices contained in the image, each slice is a grayscale image, and the slice images with the number of c channels are stacked to form a complete helical tomotherapy image. This method does not use the correlation information between channels, so the following steps are all completed on a single tomographic image. For convenience of description, the capital letter I is used below to represent a certain channel (ie, a matrix of 512*512 dimensions) in the matrix data after the image is read in by Matlab; the superscript numbers indicate the intermediate image matrix data of different steps in the method, such as I ⁰ represents the original image matrix data; use a subscript to represent a certain element in the matrix (corresponding to a pixel on a certain coordinate in the image), that is, I _{x, y} represent the element at the coordinate (x, y) in the matrix I.

Step 2: In the Matlab software, perform the following operations on the matrix (i.e. I ⁰ ) read in in the previous step:

Among them, Ω represents the neighborhood range of (x, y) coordinates, and coordinates (i, j) satisfy (i, j)∈Ω. The value of the weight coefficient w is the range weight and spatial domain weights product of . This step is a bilateral filtering operation of the image, which can have a good denoising effect while maintaining the edge of the image.

Step 3: In Matlab software, perform the following operations on the matrix in the previous step:

Among them, a, b, c are hyperparameters, set manually, and have:

Here, the symbol * represents convolution, and the symbol G represents the Gaussian kernel, that is, the image after the denominator is blurred by the numerator Gaussian kernel. This step is the improved Retinex algorithm, which does not use the logarithmic form used in the previous common Retinex method, but uses a new curve function form.

Step 4: First, use the Sobel operator function in Matlab to operate on the matrix ^I2 , and the obtained edge image matrix is marked as Gra.

Then, the edge image matrix Gra is traversed pixel by pixel to find the maximum value of the edge in the neighborhood, and the matrix Gra _M is obtained to mark whether the current position is the maximum value of the neighborhood in the matrix Gra. Then pass:

Update the matrix Gra, where f _{x, y} is a self-defined function that takes the distance from the current point (x, y) to the maximum point of the nearest edge as an independent variable, and the value of this function is greater than 1. When the distance is larger, it is less than 1, so that the distribution of edges in the edge image corresponding to the new Gra ¹ can be narrowed and the peak value can be increased.

Finally, using the Gauss-Seidel iterative method, the matrix Gra ¹ is used to deduce the edge-enhanced image, and the steps are as follows:

4. Differentiate the matrix Gra ¹ again to obtain the second-order derivative image matrix Lap of the target image;

5. Initialize matrix dst ⁰ = I ²

6. Repeat the following iterations until the number of iterations reaches 5, where n is the number of iterations, n=0,1,2...:

3)

4)

After the above steps are iterated, the final image is obtained, and the quality improvement of the helical tomotherapy image is completed.

Beneficial effect:

Experimental results: In order to verify the effectiveness of the present invention, we used this method to conduct experiments and achieved a better enhancement effect. The data used in the experiment of the present invention is the image output from the helical tomography apparatus, and it can be seen from the analysis of Figure 2 that by using the invented method, a relatively ideal image quality enhancement result is obtained, which not only effectively removes the noise in the image, but also improves the image contrast. Enhanced, the edges are also more defined.

From the experimental results, the method we invented has solved the problem of poor image quality output by the helical tomotherapy instrument. Combining this method with the helical tomotherapy instrument can effectively improve the efficiency of medical staff in using the treatment instrument. It has broad application prospects and value.

Claims (1)

1. a method for improving the image quality of helical tomotherapy, it relates to a method for improving the quality of helical tomotherapy images based on the retina-cerebral cortex theory, it is characterized in that: the specific steps of the method are as follows: Step 1: first output the computer digital image through the helical tomotherapy instrument, then use the imread function in the Matlab language to read the image, convert its information into a Matlab matrix form, and use the Matlab language to process it; The helical tomotherapy image is a digital image of 512 pixels*512 pixels*R channel, that is, the matrix data read in is 512*512*R dimensions; the symbol R indicates the number of faults contained in the image, and each fault is a gray The tomographic image with the number of R channels is stacked into a complete helical tomotherapy image; the capital letter I is used below to indicate a certain channel of the matrix data after the image is read into Matlab, that is, a matrix of 512*512 dimensions; superscript numbers Indicates the intermediate image matrix data of different steps, such as I ⁰ represents the original image matrix data; use a subscript to represent an element in the matrix, that is, a pixel on a certain coordinate in the corresponding image, that is, I _{x, y} represents the position in the matrix I elements of coordinates (x, y); Step 2: In the Matlab software, the matrix read in in the previous step, i.e. I ⁰ is carried out as follows: <mrow><msubsup><mi>I</mi><mrow><mi>x</mi><mo>,</mo><mi>y</mi></mrow><mn>1</mn></msubsup><mo>=</mo><mfrac><mrow><msub><mi>&amp;Sigma;</mi><mi>&amp;Omega;</mi></msub><msubsup><mi>I</mi><mrow><mi>i</mi><mo>,</mo><mi>j</mi></mrow><mn>0</mn></msubsup><mi>w</mi><mrow><mo>(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo>,</mo><mi>i</mi><mo>,</mo><mi>j</mi><mo>)</mo></mrow></mrow><mrow><msub><mi>&amp;Sigma;</mi><mi>&amp;Omega;</mi></msub><mi>w</mi><mrow><mo>(</mo><mi>x</mi><mo>,</mo><mi>y</mi><mo>,</mo><mi>i</mi><mo>,</mo><mi>j</mi>mi><mo>)</mo></mrow></mrow></mfrac></mrow> Among them, Ω represents the neighborhood range of (x, y) coordinates, and coordinates (i, j) satisfy (i, j)∈Ω; the value of the weight coefficient w is the range weight and spatial domain weights The product; this step is the bilateral filtering operation of the image, which has a good denoising effect while maintaining the edge of the image; Step 3: In Matlab software, perform the following operations on the matrix in the previous step: <mrow><msubsup><mi>I</mi><mrow><mi>x</mi><mo>,</mo><mi>y</mi></mrow><mn>2</mn></msubsup><mo>=</mo><mfrac><mrow><mo>(</mo><mfrac><mn>1</mn><mrow><mn>1</mn><mo>+</mo><mi>exp</mi><mrow><mo>(</mo><mo>-</mo><mi>a</mi><mo>*</mo><msub><mi>R</mi><mrow><mi>x</mi><mo>,</mo><mi>y</mi></mrow></msub><mo>+</mo><mi>b</mi><mo>)</mo></mrow></mrow></mfrac><mo>+</mo><mi>c</mi><mo>)</mo></mrow><mrow><mi>c</mi><mo>+</mo><mn>1</mn></mrow></mfrac></mrow> Among them, a, b, c are hyperparameters, set manually, and have: <mrow><msub><mi>R</mi><mrow><mi>x</mi><mo>,</mo><mi>y</mi></mrow></msub><mo>=</mo><mfrac><msubsup><mi>I</mi><mrow><mi>x</mi><mo>,</mo><mi>y</mi></mrow><mn>1</mn></msubsup><mrow><msubsup><mi>I</mi><mrow><mi>x</mi><mo>,</mo><mi>y</mi></mrow><mn>1</mn></msubsup><mo>*</mo><mi>G</mi></mrow></mfrac></mrow> Here, the symbol * represents convolution, and the symbol G represents the Gaussian kernel, that is, the denominator is the image blurred by the numerator Gaussian kernel; this step is the improved Retinex algorithm, which does not use the logarithmic form used in the previous common Retinex method, but is to use a new curve function form; Step 4: At first, use the Sobel operator function in Matlab, matrix I ² is operated, and the edge image matrix that obtains is marked as Gra; Then, the edge image matrix Gra is traversed pixel by pixel to find the maximum value of the edge in the neighborhood, and the matrix Gra _M is obtained to mark whether each element is the maximum value of the neighborhood in the matrix Gra, and then pass: <mrow><msubsup><mi>Gra</mi><mrow><mi>x</mi><mo>,</mo><mi>y</mi></mrow><mn>1</mn></msubsup><mo>=</mo><msub><mi>Gra</mi><mrow><mi>x</mi><mo>,</mo><mi>y</mi></mrow></msub><mo>&amp;CenterDot;</mo><msub><mi>f</mi><mrow><mi>x</mi><mo>,</mo><mi>y</mi></mrow></msub></mrow> Update the matrix Gra, where f _{x, y} is a custom function that takes the distance from the current point (x, y) to the nearest edge maximum point as an argument, if the distance between the current point (x, y) and the nearest edge maximum point When it is small, the value of this function is greater than 1; if the distance between the current point (x, y) and the maximum point of the nearest edge is large, the value of this function is less than 1; in this way, the distribution of the edge in the edge image corresponding to the new Gra ¹ narrower and higher peaks; Finally, using the Gauss-Seidel iterative method, the matrix Gra ¹ is used to deduce the edge-enhanced image, and the steps are as follows: Step 4.1: Perform a differential operation on the matrix Gra ¹ again to obtain the second-order derivative image matrix Lap of the target image; Step 4.2: Initialize matrix dst ⁰ =I ² Step 4.3: Repeat the following iterations until the number of iterations reaches 5, where n is the number of iterations, n=0,1,2...: 1) 2) After the above steps are iterated, the final image is obtained, and the quality improvement of the helical tomotherapy image is completed.