TWI593392B - Metal Detection and Artifact Removal Methods in Computerized Tomography Images - Google Patents

Description

Metal detection and artifact elimination method for computerized tomography image

The invention relates to the detection technology of metal substances in computer tomographic images, in particular to a metal detection and artifact elimination method for computer tomography images.

Computerized tomography images in medical imaging, which have the advantages of high resolution, fine detection range and flexibility, are commonly used to screen cancer and check for small lesions. In addition, computed tomography images can be assisted surgically. Doctors perform diagnosis of organ damage or swelling and internal bleeding, so they are widely used in clinical applications. The use of computed tomography to reconstruct 3D images is now widely used in medical diagnostics, such as virtual endoscopy, plastic reconstruction surgery, and dental implant correction planning systems. However, the problems caused by artifacts reconstructed by computerized tomography have yet to be resolved.

The more common clinical artifacts are the so-called metal artifacts, which occur because the tissue part of the scan contains metal implants, such as dental implants, bone nails or bone plates. This causes several artifacts such as Beam Hardening, Scatter, and the like. These will now obscure other body tissues on the image, which in turn affects the authenticity and reliability of the information that the image could have provided to us. As a result, it has lost the clinical use of medical images to enhance and assist physicians. Diagnostic accuracy or success rate of surgery and other related medical aid software development intentions.

The main purpose of the present invention is to provide a metal detection and artifact removal method for computer tomographic scanning images, which can effectively remove metal artifacts in a computer tomographic image.

According to the present invention, a metal detection and artifact removal method for a computerized tomography image according to the present invention is mainly performed by a computer, and the method comprises the following steps: A) obtaining a computed tomography image: Scanning a physical body with a computerized tomography device to obtain a plurality of computerized tomographic images, or downloading from a predetermined file source to obtain a plurality of computerized tomographic images corresponding to the physical body; B) Metal detection: The computerized tomographic images are converted into a histogram containing information on the number of tissues of the body, and the number of tissues in the body when the body has a metal substance is smaller than the body when the body does not have a metal substance The number is increased by 1; then the histogram is filtered by a predetermined filtering method to obtain a filtered histogram, and then the one-dimensional derivative of the filtered histogram is calculated, and the positive and negative changes of the one-dimensional derivative are used. Determining peaks and valleys in the filtered histogram, if the number of troughs of the filtered histogram is less than that of the body tissue The number of weaves is increased by one, that is, it is determined that the computerized tomographic image corresponding to the filtered histogram contains metal information; after that, the rightmost trough position in the filtered histogram is taken, and the trough position is set. a threshold value, and filtering each of the computerized tomographic images according to the threshold value to generate each binarized image, each of the binarized images simply displaying the metal substance; C) the Radden conversion to obtain projection data : performing the Ryden conversion on each of the computerized tomographic images to obtain respective original projection data corresponding to each of the computerized tomographic images; and performing a Ryden conversion on each of the binarized images to obtain each of the two Each of the two-bit projection data corresponding to the imaged image; D) linear interpolation and weighted interpolation: combining the original projection data and the two-bit projection data corresponding to each of the computerized tomographic images, Thereby, the metal portion in each of the original projection data is removed, and linear interpolation and weight interpolation are performed on each of the combined projection data metal portions; after linear interpolation The obtained result is defined as linear interpolation projection data, and the result obtained after weighted interpolation is defined as weighted interpolation projection data; when linear interpolation is performed, the following equation (1) is used; Wherein x and y are values to be interpolated, and x ₀ , x ₁ and y ₀ , y ₁ are known values; when performing weighted interpolation, the following equation (2) is used; Equation (2) where C _n is the projection data label, g ^WI (C _n ) is the weighted interpolation projection data, g(C _n ) is the original projection data, N is the proportional coefficient, and g _R is the projection data outside the metal interval The right value, g _L is the outer left value of the projection data metal interval, N _metal is the total number of metal interval values; E) the back projection imaging: the linear interpolation projection data in the aforementioned step D) is back-projected to form a complex number Linearly interpolating the image, and performing back projection imaging on each of the weighted interpolation projection data in the foregoing step D) to form a complex weight interpolated image; F) establishing a preprocessed image: for each of the linear interpolated images An edge-preserving blur filter performs filtering processing to obtain each linear interpolation filter image; in addition, each computerized tomographic image is combined with its corresponding binarized image to remove pixel values of the metal region, and then fill in Entering a pixel value of a metal region in each of the linear interpolation filter images to obtain each metal removed image; and then subtracting each of the metal removed images from each of the linear interpolation filter images to obtain each After the first difference image is normalized, the first difference image is normalized, and then the fusion operation is performed to obtain each pre-processed image, wherein when normalizing, the following formula (3) is used: (3) where D _max and D _min are the maximum and minimum values of the pixel values in the first difference image; and when performing the fusion operation, the following equations (4) and (5) are performed: Formula (4) Equation (5) where 0 < t ≤ 1 and n > 0, I _prec is the preprocessed image, I _{LI -filtered} is the linear interpolation filter image, I _{M etal -removed} is the metal removal image; G) the fusion image is created And converting each of the preprocessed images into respective preprocessed projection data by performing Ryden conversion; and performing slope weight calculation on each of the preprocessed projection data and each of the original projection data to perform metal region and metal region bonding, and after bonding The data is added to the corresponding original projection data and averaged, and then linearly interpolated and reconstructed and imaged to obtain each slope image; in addition, each of the original projection data is subtracted from each of the pre-processed projection data to obtain Each difference projection data is linearly interpolated for each of the difference projection data metal sections to obtain linear interpolation projection data of each difference, and then the preprocessed projection data is superimposed on each of the difference linear interpolation projection data. Then, performing back projection imaging, that is, obtaining each preprocessed superimposed image; then, subtracting each preprocessed superimposed image from each of the slope images to obtain each second difference a value image; then, using the foregoing equations (3), (4), and (5) to normalize each of the second difference images, each of the slope images and each of the slope images are blurred by edges. The image obtained after filtering is subjected to a fusion operation to obtain each of the fused images; and H) the metal portion of the fused image is backfilled: the values belonging to the metal portion of the fused images are interpolated from the same position in the image by the weights Replace it.

Thereby, not only the metal detection effect but also the metal artifact in the computer tomographic image can be effectively removed.

In order to explain the technical features of the present invention in detail, the following preferred embodiments will be described with reference to the drawings, wherein:

As shown in FIG. 1 , a metal detection and artifact removal method for a computerized tomography image according to a preferred embodiment of the present invention is mainly implemented by a computer, and the method includes the following steps:

A) Obtaining a computerized tomographic image: as shown in Figures 2 and 3, a computerized tomography device scans a physical body to obtain a plurality of computed tomography images, or is downloaded from a predetermined file source. A plurality of computed tomography scan images corresponding to the body are obtained.

B) Metal detection: converting the computed tomography scan image into a histogram, as shown in Fig. 4, the histogram contains information on the number of tissues of the body, and the quantity information of the tissue when the body has a metal substance The number of tissues is one more than the number of tissues when the body does not have a metal substance. For example, in the human mouth, the body tissues are bones, teeth, and flesh, so the number of tissues is 3, and when there is a metal substance, the tissue is made. The quantity becomes 4; then the histogram is filtered by a predetermined filtering method, and then a filtered histogram is obtained, and then the one-dimensional derivative of the filtered histogram is calculated, and the positive and negative changes of the one-dimensional derivative are used to judge The peaks and valleys in the filtered histogram, if the number of troughs of the filtered histogram is one more than the number of tissues when the body tissue does not have a metal substance, that is, the computer corresponding to the filtered histogram is determined. The tomographic image contains metal information; after that, the rightmost trough position in the filtered histogram is taken, and the trough position is set to a threshold value, and Threshold to filter each of the computer tomography image generating respective binary images, as shown in FIG. 5, an example system represented in an image, each of the binarized image that is displayed pure metal species.

C) Ryden conversion to obtain projection data: each of the computerized tomographic images is subjected to Ryden conversion, and each original projection data corresponding to each of the computerized tomographic images is obtained, as shown in Fig. 6(A). Taking a data image as an example; and performing a Ryden conversion on each of the binarized images, and obtaining each two-bit projection data corresponding to each binarized image, as shown in FIG. 6(B) Take a data image as an example.

D) linear interpolation and weighted interpolation: combining the original projection data and the two-bit projection data corresponding to each of the computerized tomographic images to remove the metal portion of each of the original projection data, Linear interpolation and weight interpolation are performed on each of the combined projection data metal parts; the result obtained after linear interpolation is defined as linear interpolation projection data, as shown in Fig. 7, The data image is taken as an example; the result obtained after the weighted interpolation is defined as the weighted interpolation projection data. As shown in Fig. 8, a data image is taken as an example; linear interpolation is performed. When using the following formula (1):

Formula 1)

Where x and y are the values to be interpolated, and x ₀ , x ₁ and y ₀ , y ₁ are known values. When performing weighted interpolation, it is performed by the following formula (2);

Formula (2)

Where C _n is the projection data label, g ^WI (C _n ) is the weighted interpolation projection data, g(C _n ) is the original projection data, N is the proportional coefficient, and g _R is the outer value of the projection data metal interval, g _L is the outer left value of the projection data metal interval, and N _metal is the total number of metal interval values.

E) back projection imaging: back-projecting each of the linear interpolation projection data in step D) to form a complex linear interpolation image, as shown in FIG. 9 , taking an image as an example; Each of the weighted interpolation projection data in the foregoing step D) is subjected to back projection imaging to form a complex weight interpolated image. As shown in FIG. 10, an image is taken as an example.

F) establishing a pre-processed image: filtering each of the linear interpolation images with an edge-preserving blur filter to obtain each linear interpolation filter image, as shown in FIG. 11 , taking an image as an example; In addition, combining the computed tomography scan image with the corresponding binarized image, thereby removing the pixel value of the metal region, and then filling in the pixel values of the metal regions in each of the linear interpolation filter images, thereby obtaining each The metal removes the image, as shown in FIG. 12, taking an image as an example; and then subtracting each of the metal-removed images by each of the linear interpolation filter images to obtain each first difference image, such as As shown in Fig. 13, an image is taken as an example; after the first difference image is normalized, a fusion operation is performed, that is, each preprocessed image is as shown in Fig. 14, which is a The image is shown as an example. Among them, when normalization is performed, the following formula (3) is used:

Formula (3)

Wherein D _max and D _min are maximum and minimum values of pixel values in the first difference image.

When performing the fusion operation, the following equations (4) and (5) are used:

Formula (4)

Formula (5)

Where 0 < t ≤ 1 and n > 0, I _prec is the preprocessed image, I _{LI -filtered} is the linear interpolation filter image, and I _{M etal -removed} is the metal removal image.

G) establishing a fused image: converting each of the preprocessed images into a respective pre-processed projection data by performing a Ryden conversion; performing a ramp weight operation on each of the pre-processed projection data and each of the original projection data to perform a metal region and a metal region After the joint is joined, the combined data is added to the corresponding original projection data and averaged, and then linearly interpolated and reconstructed and imaged to obtain each slope image. As shown in Fig. 15, an image is taken as an image. In addition, each of the original projection data is subtracted from each of the original projection data to obtain each difference projection data, and then the metal segments of the difference projection data are linearly interpolated to obtain linear interpolation projections of the differences. Data, and then superimposing each of the pre-processed projection data on the linear interpolation projection data, and performing back-projection imaging, that is, obtaining each pre-processed superimposed image, as shown in Fig. 16, taking an image as an example. Representing; then, subtracting each of the preprocessed superimposed images from each of the slope images to obtain respective second difference images, as shown in FIG. For example, the second difference image is first normalized by using the above formulas (3), (4), and (5), and each of the slope images and each of the slope images are blurred by edges. The image obtained after the filtering is subjected to a blending operation to obtain each of the fused images. As shown in FIG. 18, an image is taken as an example.

H) backfilling the metal part of the fused image: the values belonging to the metal part of the fused image are replaced by the values of the same position in the weighted interpolated image, that is, each completed image is obtained, as shown in Fig. 19 Take an image as an example.

In the present embodiment, in the foregoing step B), the predetermined filtering method is exemplified by a moving average filtering method, which is performed by using the following formula (6):

Formula (6)

Where H _I ( i ) is the frequency of the histogram, i is the gray scale value of the histogram, and H _MA (i) is the histogram frequency processed by the moving average method.

In addition, in the foregoing step G), the edge-preserving fuzzy filter is a nonlinear filter, and the filtering effect is mainly obtained by performing the following formulas (7) and (8):

Formula (7)

Formula (8)

The image pixel reconstructed after linear interpolation of the projection data is b _BF (i,j ) , and the filtered pixel is b _AF (i,j ) , and the filtering calculation range is -v to v , where N is within the filter calculation range. The number of rms pixels, T is the threshold set by the user.

According to the above steps and description, the metal detection and artifact removal method of the computerized tomographic image provided by the present invention can separate the metal parts in the image in the step of weight interpolation. The artifact removal is performed on the portion of the fused image, and finally the value belonging to the metal portion of the weighted interpolation image is backfilled into the fused image. Thereby, the value of the metal part backfilled in the image is extremely clear and clear, not only achieves the effect of metal detection, but also effectively removes metal artifacts in the computerized tomographic image.

It should be noted that, in addition to the foregoing steps, the present invention may further include technical features for performing background removal operations on respective images to be processed, and technical features for performing reduction interval interval operations. The background operation is mainly used to eliminate background noise that may be caused when performing back projection imaging; when performing the background removal operation, the background position of each image to be processed is recorded, and the background position is determined to be processed. The position of the non-human tissue value in the image, and the maximum value of the pixel value of the background position is used as the background threshold, and filtering is performed according to the background threshold, thereby filtering the background noise of each image to be processed. Fig. 20 (A) and Fig. 20 (B) show images before and after background removal, respectively. The reduction of the numerical interval is mainly to eliminate the problem of pixel value amplification which may be caused when performing back projection imaging. When performing the operation of restoring the numerical interval, it is for each pixel of the image to be processed that has been subjected to background removal. The pixel value is adjusted by the numerical interval, and the numerical interval is calculated by using a scaling factor, and then the average pixel value of each of the computed tomographic images and the corresponding average pixel value of each of the to-be-processed images are performed. Dividing is performed to generate a plurality of adjustment coefficients, and each of the to-be-processed images is operated by each of the adjustment coefficients.

Based on the above techniques for removing the background and reducing the numerical interval, we can first perform the processing of the restored numerical interval on each of the fused images in step H), and remove the imaging background and the restored numerical interval for each of the weighted interpolated images. After the processing, the action of backfilling the metal portion is performed. This allows the background noise of the image to be removed, and the pixel values of the image can be adjusted so that they are not amplified during the conversion process. In addition, the background removal and the reduction value interval can also be used between step E) and step F) to remove background noise and adjust image pixel values in the correct interval.

BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a flow chart of a preferred embodiment of the present invention. 2 is a schematic diagram of an image of a preferred embodiment of the present invention showing a plurality of computed tomography scan images. Figure 3 is a schematic view of a preferred embodiment of the present invention showing a computed tomography scan image. Figure 4 is a histogram of a preferred embodiment of the present invention. Figure 5 is a schematic diagram of an image of a preferred embodiment of the present invention showing a binarized image. Fig. 6(A) is a schematic view showing an image of a preferred embodiment of the present invention, showing original projection data. Figure 6(B) is a schematic diagram of an image of a preferred embodiment of the present invention showing binary projection data. Figure 7 is a schematic diagram of an image of a preferred embodiment of the present invention showing linear interpolation projection data. Figure 8 is a schematic diagram of an image of a preferred embodiment of the present invention showing weighted interpolation projection data. Figure 9 is a schematic illustration of a preferred embodiment of the present invention showing a linear interpolated image. Figure 10 is a schematic diagram of an image of a preferred embodiment of the present invention showing weighted interpolated images. Figure 11 is a schematic diagram of an image of a preferred embodiment of the present invention showing a linear interpolation filtered image. Figure 12 is a schematic illustration of a preferred embodiment of the present invention showing a metal removed image. Figure 13 is a schematic diagram of an image of a preferred embodiment of the present invention showing a first difference image. Figure 14 is a schematic illustration of a preferred embodiment of the present invention showing a pre-processed image. Figure 15 is a schematic illustration of a preferred embodiment of the present invention showing a ramp image. Figure 16 is a schematic diagram of an image of a preferred embodiment of the present invention showing a pre-processed overlay image. Figure 17 is a schematic diagram of an image of a preferred embodiment of the present invention showing a second difference image. Figure 18 is a schematic illustration of a preferred embodiment of the present invention showing a fused image. Figure 19 is a schematic diagram of an image of a preferred embodiment of the present invention showing the completed image. Figure 20 (A) is a schematic view of a preferred embodiment of the present invention showing an image before background removal. Figure 20 (B) is a schematic view of a preferred embodiment of the present invention showing an image after background removal.

Claims (5)

A metal detection and artifact removal method for a computerized tomographic image is mainly performed by a computer, and the method comprises the following steps: A) obtaining a computerized tomographic image: a computerized tomography device The body scans to obtain a plurality of computerized tomographic images, or is downloaded from a predetermined file source to obtain a plurality of computerized tomographic images corresponding to the physical body; B) Metal detection: converting the computerized tomographic images into a histogram, the histogram contains information on the number of tissues of the body, and the number information of the organization when the body has a metal substance is 1 more than the number of tissues when the body does not have a metal substance; The filtering method filters the histogram, and then obtains a filtered histogram, and then calculates the one-dimensional derivative of the filtered histogram, and uses the positive and negative changes of the one-dimensional derivative to determine the filtered histogram. Peak and trough, if the number of troughs of the filtered histogram is one more than the number of tissues when the flesh tissue does not have a metal substance, it is judged that The computerized tomographic image corresponding to the histogram of the wave contains metal information; after that, the rightmost trough position in the filtered histogram is taken, and the trough position is set to a threshold value, and according to the valve The value is used to filter each of the computerized tomographic images to generate each binarized image, and each of the binarized images simply displays the metal substance; C) the Ryden conversion obtains the projection data: each of the computerized tomographic images is scanned Performing a Ryden conversion to obtain each original projection data corresponding to each of the computerized tomographic images; and performing a Ryden conversion on each of the binarized images to obtain each of the two bits corresponding to the binarized image Meta-projection data; D) linear interpolation and weighted interpolation: combining each of the original projection data and the binary projection data corresponding to each of the computerized tomographic images, thereby making each of the original projection data The metal portion is removed, and the metal portions of the combined projection data are linearly interpolated and weighted, respectively; the results obtained after linear interpolation are defined as Interpolation of projection data is performed after the weight of the weighting interpolation results obtained that is defined as the weight of the interpolated projection data; carrying out linear interpolation, based using the following formula (1); Wherein x and y are values to be interpolated, and x ₀ , x ₁ and y ₀ , y ₁ are known values; when performing weighted interpolation, the following formula (2) is used; Equation (2) where C _n is the projection data label, g ^WI (C _n ) is the weighted interpolation projection data, g(C _n ) is the original projection data, N is the proportional coefficient, and g _R is the projection data outside the metal interval The right value, g _L is the outer left value of the projection data metal interval, N _metal is the total number of metal interval values; E) the back projection imaging: the linear interpolation projection data in the aforementioned step D) is back-projected to form a complex number Linearly interpolating the image, and performing back projection imaging on each of the weighted interpolation projection data in the foregoing step D) to form a complex weight interpolated image; F) establishing a preprocessed image: for each of the linear interpolated images An edge-preserving blur filter performs filtering processing to obtain each linear interpolation filter image; in addition, each computerized tomographic image is combined with its corresponding binarized image to remove pixel values of the metal region, and then fill in Entering a pixel value of a metal region in each of the linear interpolation filter images to obtain each metal removed image; and then subtracting each of the metal removed images from each of the linear interpolation filter images Each of the first difference images is normalized, and then the first difference image is normalized, and then the fusion operation is performed to obtain each pre-processed image, wherein when normalizing, the following formula (3) is used: (3) where D _max and D _min are the maximum and minimum values of the pixel values in the first difference image; and when performing the fusion operation, the following equations (4) and (5) are performed: Formula (4) Equation (5) where 0 < t ≤ 1 and n > 0, I _prec is the preprocessed image, I _{LI -filtered} is the linear interpolation filter image, I _{M etal -removed} is the metal removal image; G) the fusion image is created And converting each of the preprocessed images into respective preprocessed projection data by performing Ryden conversion; and performing slope weight calculation on each of the preprocessed projection data and each of the original projection data to perform metal region and metal region bonding, and after bonding The data is added to the corresponding original projection data and averaged, and then linearly interpolated and reconstructed and imaged to obtain each slope image; in addition, each of the original projection data is subtracted from each of the pre-processed projection data to obtain Each difference projection data is linearly interpolated for each of the difference projection data metal sections to obtain linear interpolation projection data of each difference, and then the preprocessed projection data is superimposed on each of the difference linear interpolation projection data. Then, performing back projection imaging, that is, obtaining each preprocessed superimposed image; then, subtracting each preprocessed superimposed image from each of the slope images to obtain each second a value image; then, using the foregoing equations (3), (4), and (5) to normalize each of the second difference images, each of the slope images and each of the slope images are subjected to edge-preserving blur filtering The acquired images are then subjected to a fusion operation to obtain the respective fused images; and H) the fused image is backfilled with metal portions: the values belonging to the metal portions of the fused images are numerically calculated from the same positions in the weighted interpolated images. Replace. According to the metal detection and artifact elimination method of the computerized tomography image according to the first application of the patent scope, the predetermined filtering method is a moving average filtering method, which is performed by using the following formula (6): Equation (6) where H _I ( i ) is the frequency of the histogram, i is the gray scale value of the histogram, and H _MA (i) is the histogram frequency processed by the moving average method. According to the metal detection and artifact elimination method of the computerized tomography image according to the first application of the patent scope, in the step G), the edge preserves the fuzzy filter, which is a nonlinear filter, mainly for execution The following equations (7) and (8) are calculated to obtain a filtering effect: Formula (7) Equation (8), wherein the image pixel reconstructed after linear interpolation of the projection data is b _BF (i,j ) , and the filtered pixel point is b _AF (i,j ) , and the filtering calculation range is -v to v , N is rms calculating the number of pixels within the scope of the filter, T is the user's own set of thresholds. According to the metal detection and artifact elimination method of the computerized tomography image according to the first application of the patent scope, the method further includes the technical feature of removing the background operation for each image to be processed, and then performing the reduction value interval operation. Technical feature; when performing the operation of removing the background, recording the background position of each image to be processed, the background position is the position of the non-human tissue value in each image to be processed, and finding the maximum pixel value of the background position The value is used as the background threshold, and is filtered according to the background threshold to filter the background noise of each image to be processed; when performing the operation of restoring the numerical interval, each of the background operations has been removed. The pixel values of the pixels of the image to be processed are adjusted in the numerical interval, and the numerical interval is calculated by using a scaling factor, and then the average pixel value of each of the computer tomographic images is scanned and corresponding to each The average pixel value of the image to be processed is divided to generate a plurality of adjustment coefficients, and each of the to-be-processed images is performed by using the adjustment coefficient. Operation. According to the metal detection and artifact elimination method of the computerized tomography image according to the fourth application of the patent application, in the step H), the processing of the reduction value interval is performed on each of the fusion images, and the weights are respectively The sexual interpolated image is processed to remove the imaging background and restore the numerical interval.