JP4408863B2 - Medical image processing apparatus and method - Google Patents

Description

  The present invention relates to a medical image processing apparatus and method, and more particularly to a medical image processing apparatus and method for creating a shadow enhanced image and a difference image from a medical image.

  In recent years, a large amount of medical images are obtained by a medical image photographing apparatus, and there is a demand for efficient interpretation of a large amount of medical images. Therefore, in order to make it easier to find an affected area such as a cancer shadow from a medical image, the shadow is enhanced by a difference process between images. Specifically, when there is no cancer shadow in an image captured in the past (hereinafter referred to as a past image) and there is a cancer shadow in an image captured thereafter (hereinafter referred to as an image on the day), for example, In the abnormal shadow automatic detection device disclosed in Japanese Patent No. 6870, a true abnormal shadow is extracted from a difference image obtained by subtracting pixels of a current image and a past image corresponding to the same position of the subject.

  However, in the above-described abnormal shadow automatic detection device, since subtraction is performed between corresponding pixels of the current day image and the past image, if the alignment is not sufficient, a false positive shadow (not a real shadow) is included in the subtraction result. (Part; pseudo-shade) will occur.

  However, the above-described abnormal shadow automatic detection device makes no mention of such a case.

The medical image processing apparatus of the present invention includes a first storage means for storing a predetermined part of a subject imaged by the medical image diagnostic apparatus as a first medical image, and a first storage means stored in the first storage means. Second storage means for storing a second medical image of the same subject and different from the date and time when one image was taken, and a first medical image stored by the first storage means A first setting means for setting a plurality of first local regions having at least one pixel as an element, and a plurality of first local regions set by the first setting means, Second setting means for setting a plurality of second local areas having a width equal to or larger than the first local area in the second image stored in the second storage means, and the second setting means Concentration value for each of a plurality of second local areas set by Based on a reference value calculating means for calculating a reference value, a reference value calculated by the reference value calculating means, and a density value of each pixel in a plurality of first local areas set by the first setting means A reference value calculated by the reference value calculation means, and an image creation means for creating an emphasized image for each of the first local regions; Is at least one of a local maximum value, a local average value, and a local median for each second local region set by the second setting means, or a constant multiple thereof, and the reference value calculating means Direction calculating means for calculating the maximum concentration gradient direction of the second local region set by the second setting means, wherein the reference value calculating means calculates the maximum concentration gradient calculated by the direction calculating means. The maximum value of the density value from the pixel value along the line, a value obtained by multiplying the maximum value by a constant, a value obtained by multiplying the maximum value by a function of the density value, a value obtained by multiplying the maximum value by an approximation function of the density distribution, and the maximum A value obtained by multiplying the value by a constant and a function of an angle formed by the density gradient maximum direction of the first image and the density gradient maximum direction of the second image, and a density value in the vicinity of the pixel where the density value is the maximum value The local maximum value is calculated using any one of the average values.

  The reference value here refers to a local maximum value, a local average value, a local median, a constant multiple of the local maximum value or the local average value, a value related to the concentration gradient, and a value related to the concentration gradient according to the characteristics of the medical image. It can be adopted. For example, the adoption of the maximum value is the simplest and most effective value, and the adoption of the average value or median suppresses the noise component of the image from the maximum value, and if the density gradient is used, the values of adjacent pixels and neighboring pixels are rapidly increased. It can cope with the case of changing to.

  Further, the enhanced image may be created by the image creating means based on the difference value between the reference value and the density value of each pixel in the first local region. Although most desirable, it can be performed by any calculation such as addition, multiplication, division, etc. as long as it can function to emphasize only the cancer shadow.

  According to the medical image processing apparatus of the present invention, even if a positional deviation occurs between the current day image and the past image, the shadow of the image is emphasized, and an abnormal shadow such as cancer can be easily found.

  Further, according to a preferred embodiment of the present invention, the image creating means superimposes the contour image generating means for generating a contour image of the first or second medical image, and the contour image and the emphasized image. Image superimposing means, and the display means displays the superimposed image superimposed by the image superimposing means.

  As a result, by superimposing the shadow-enhanced image and the contour, even if it is difficult to determine where in the body the shadow is located in the shadow-enhanced image, while highlighting the abnormal shadow such as cancer, The position can be easily grasped.

  According to a preferred embodiment of the present invention, the image forming apparatus further includes alignment means for aligning the first medical image and the second medical image.

  Thereby, if the so-called image of the day and the past image are combined, the calculation accuracy of abnormal shadows such as cancer is improved, and the position of the abnormal shadows can be found more easily.

  Furthermore, according to a preferred embodiment of the present invention, the pseudo shadow generated by the imperfection of the alignment by the alignment unit is reduced in density compared with the enhanced image generated by the image generation unit, thereby reducing the pseudo shadow. And a density reducing means for creating a reduced image, and the display means displays the pseudo-shadow-reduced image created by the density reducing means.

  As a result, even if the alignment is incomplete, the pseudo-shadow generated thereby is reduced in density, so that an abnormal shadow such as cancer can be found more easily.

  Hereinafter, preferred embodiments of a medical image processing apparatus and method according to the present invention will be described in detail with reference to the accompanying drawings.

  FIG. 1 shows a configuration of a medical image processing apparatus 10 according to an embodiment of the present invention. The medical image processing apparatus 10 includes a central processing unit (hereinafter referred to as a CPU) 11, a main memory 12, a magnetic disk 13, a display memory 14, a CRT 15, a controller 16, a mouse 17, and a keyboard 18. Connected through. Further, the medical image photographing apparatus 30 is connected to the medical image processing apparatus 10 via, for example, the LAN 32.

  The medical imaging apparatus 30 is an X-ray CT (Computed Tomography) apparatus, a PET (Positron Emission Tomography) apparatus, an MRI (Magnetic Resonance Imaging) apparatus, a DR (Digital Radiography) apparatus, an ultrasonic imaging apparatus, a fundus camera, or the like. Although it is good, it is not limited to these, What kind of thing may be sufficient if it is an apparatus which image | photographs a medical image. The medical image to be processed by the medical image processing apparatus 10 includes medical images obtained by the various medical image photographing apparatuses described above.

  The main memory 12 is used as an area for temporary storage and processing of images and data, and processing results are displayed on the CRT 15 via the display memory 14 and stored in the magnetic disk 13 for redisplay and result reference. . The magnetic disk 13 stores medical images captured by the medical image capturing device 30.

  Next, image processing performed by the medical image processing apparatus 10 will be described with reference to FIGS.

  The first medical image (here, also referred to as “image of the day” because the image is taken at a predetermined examination date and time) is obtained by imaging a predetermined part of the subject and patient by the medical image photographing apparatus 30.

  Information on the patient and the imaging region is transmitted to the magnetic disk 13 via the common bus 19 via the network (LAN) 32.

  The second medical image (here, it is taken at a date and time different from the date and time when the image was taken on the day, but is often taken before that, also referred to as a “past image”) is stored on the magnetic disk 13. From the stored state, it is read out to the main memory 12 based on the transmitted patient and imaging part information.

  The current day image is transmitted to the display memory 14 via the network (LAN) 32, the common bus 19, and displayed on the CRT 15.

  An operation screen displayed on the CRT 15 of the medical image processing apparatus 10 is shown in FIG. A simple difference image 47 obtained by the simple difference method for the current day image and the past image, a shadow-enhanced image 48 by a process called “anisotropic difference process” described in the present embodiment, and a circle or the like to clearly indicate the shadow. The marker display image 49 can be selected.

  The first local area a is set to a part of the display area of the current day image by the mouse 17 on the current day image displayed on the CRT 15. Here, an example in which one first local region is set will be described, but a plurality of first local regions may be set.

  In the past image read out to the main memory 12, the CPU 11 sets a second local area A having an area equal to or larger than the local area where the first local area a is set.

  The local reference value of the second local area A is calculated by the local reference value calculation unit 111 of the CPU 11. The local reference value here is a weight such as a case where representative values such as the maximum value, average value, and median in the second local region A are used as they are, a combination of these representative values, or a multiplication of these representative values by a constant. May be multiplied.

  The shadow-enhanced image is created by the image creation unit 112 of the CPU 11 performing a difference between the density value of the first local region a and the local reference value of the second local region A. In addition, the shadow-enhanced image is created by the image creation unit 112 based on the calculated difference value by calculating a difference value between the reference value and the density value of each pixel in the first local region. Although it is most desirable, it may be performed by any calculation such as addition, multiplication, and division as long as the function of enhancing only the cancer shadow can be exhibited.

  The shadow enhanced image is transmitted from the image creating unit 112 to the display memory 14 and displayed on the CRT 15.

  Here, the image density of the medical image is represented by the density value of each pixel constituting the medical image. The density value includes, for example, a CT value, a pixel value, and the like, but is not limited thereto, and includes all values representing image density, transparency, opacity, lightness, darkness, signal intensity, and the like. The density value is inverted between positive and negative and magnitude depending on conditions such as the type of medical image capturing apparatus that captured the medical image, the calculation method and representation method of the density value, and whether the image is a negative image or a positive image. In this specification, when comparing a portion A in which a living tissue is captured and a portion B in which air or water is captured in a certain medical image, the density value of the portion A is not limited regardless of the magnitude of the density value. It is defined as being larger than the density value of part B. According to this definition, for example, in the image shown in FIG. 4, the density value of the portion displayed in white is larger than the density value of the portion displayed in black.

  FIG. 4 explains the principle of the present invention in comparison with the prior art, and image processing was performed on the CT tomogram of the subject's lung field using the simple difference method and the anisotropic difference method, respectively. An example is shown. When there is no cancer shadow in the past image 2 and there is a cancer shadow on the current day image 1 (region b reversed in white compared to the previous image on the lower right of the screen), the past image 2 is subtracted from the current day image 1. Thus, a difference image in which the density of the cancer shadow part is large and the density of the other part is small is obtained. That is, the cancer shadow is emphasized and it is easy to find. In the conventional simple difference processing, since the corresponding pixels of the current day image and the past image are subtracted, if the alignment is not sufficient, a false positive shadow (a part that is not a real shadow; a pseudo shadow) occurs in the difference image 3 It's hard to find real shadows. According to the anisotropic local maximum value difference process according to the method of the present embodiment, which will be described later, a false positive shadow is unlikely to occur in the difference image 4, and a real shadow can be easily found.

  Another example of the simple difference process will be described with reference to FIGS. The image (FIG. 5) in which a false positive shadow is embedded in the current day image corresponding to a light minute lung cancer and the past image (not shown) are aligned so that the correlation of each part of the image is the highest. FIG. 6 shows the result of subtracting the past image from the current day image after alignment. In FIG. 6, cancer shadows are difficult to find.

  Next, a subtraction method in the medical image processing apparatus 10 according to the present embodiment will be described with reference to FIG. This flowchart is performed on behalf of only one pixel in the image. For example, in the main memory 12 that stores the difference image, a sufficiently large value is recorded so that the difference result does not become negative.

(Step 90)
A direction having the maximum density gradient is obtained at the pixel point (X, Y) in the past image. A partial region 52 of the image 51 in FIG. Among these, a small area 54 of, for example, a 3 × 3 image matrix wider than the pixels of the image of the day is taken, and a direction having the maximum density gradient is obtained among the vertical, horizontal, and diagonal directions passing through the center point of the small area 54. Here, as the small region 54, a region of three vertical pixels and three horizontal pixels centering on the point A22 is considered, and abs (A11-A33), abs (A21-A23), abs (A31-A13), abs (A12- From the maximum value among the absolute values of A32), the maximum concentration gradient direction at point A22 is obtained. The maximum value among the absolute values of abs (A11-A22), abs (A21-A22), abs (A12-A22), and abs (A31-A22) may be used.

  FIG. 9 shows a case where the concentration gradient in the A11 and A33 directions is the largest.

  Of the small regions 54, a region along the maximum concentration gradient direction is referred to as a “specific direction region (i, j) (or anisotropic region (i, j))”. Here, A11, A22, and A33 are those, but B11, B44, and the like in the same direction may be included. Bij indicates that the area for obtaining the concentration gradient and the subtraction target area do not necessarily have to be the same, and is included in Aij when the area for obtaining the concentration gradient and the subtraction target area are the same. . Therefore, B55 (not shown) can be provided outside B44.

  In short, FIG. 9 shows the principle of determining whether or not the maximum density gradients of the current day image 901, the past image 902, and the difference image 903 are the same, and the respective density gradient maximum directions match. The procedure is as follows.

(Step 91)
Next, as shown in the example of FIG. 10B, the maximum value max (local maximum value) of the density values in the specific direction area (i, j) of the past image 902 shown in FIG. 9 is obtained.

(Step 92)
C × max is subtracted for each pixel (i, j) in the specific direction area (i, j) of the day image 901 shown in FIG. The result is sa1. C is a constant including 1.0 and may be set according to the image. That is, after the maximum values max of A11, A22, A33, B11, and B44 of the past image 902 are obtained, for example, max × constant is subtracted from the pixel A11 of the image 901 on the day to obtain sa1. In the simple difference process, as shown in the example of FIG. 10A, subtraction is performed between the pixels of the current-day image 901 and the past image 902 corresponding to the same position of the subject, The difference image 903 is obtained as (image data) = ba.

(Step 93)
It is determined whether or not “pixel (i, j) value of difference image 903> sa1”. That is, the subtraction result sa1 is compared with each current memory SUB (same coordinates as A11).

(Step 94)
As a result of the subtraction in step 93, when “pixel (i, j) value of difference image SUB> sa1”, pixel (i, j) of the difference image = sa1. That is, the smaller value sa1 is stored in SUB (same coordinates as A11). If the subtraction result is negative, it may be replaced with a specific value, for example, zero.

  The same processing is performed for the other pixels A22, A33, B11, and B44 in the specific direction area (i, j).

  The processes in steps 90 to 94 described above are executed for all pixels in the image. At this time, it is preferable that the small areas 54 move while overlapping each other little by little. When moving in this way, subtraction for one pixel is performed a plurality of times, and the result of the smallest difference is stored.

  The results are shown in FIG. The shadow is emphasized, and it is easier to find the abnormal shadow than the result of the simple difference shown in FIG. When the high density portion of FIG. 11 is detected and a marker is attached to the original CT image (FIG. 5), the result is shown in FIG.

  FIG. 13 is an image on the day containing a real cancer shadow. From this, the past images are subtracted according to the procedure of the preferred embodiment of the present invention as shown in FIG.

  When the position of the cancer shadow enhanced by the subtraction process is difficult to understand, as shown in the example of FIG. 15, if an outline image is separately extracted and combined with the difference image, an abnormal shadow is obtained as shown in FIG. The positional relationship between the locations becomes easier to understand. FIG. 17 shows a synthesis procedure for each pixel of the difference image and the contour image.

(Step 20)
A threshold value is determined as to whether the contour image and the emphasized image can be superimposed. If the result of the determination is that the value is within the threshold value range, the process proceeds to step 21;

(Step 21)
Anisotropic difference processing is executed.

(Step 22)
The difference result is stored in the memory, and the process is terminated.

(Step 23)
The contour density is stored in the memory, and the process is terminated.

  FIG. 18 shows the case where the concentration gradient in the A21 and A23 directions is the largest in the small region. Subtract the maximum value x constant of A21, A22, A23, 21, B24 of the past image 902 from the pixel A21 of the image 901 on the day, compare the subtraction result with the current SUB (same coordinates as A21), and calculate the smaller value Store in SUB (same coordinates as A21). The same applies to A22, A23, B21, and B24.

  In the above description, the past image 902 is subtracted from the current day image 901. However, the result is the same even if the current day image 901 is subtracted from the past image 902 and the sign is inverted at the end.

In addition, the subtraction method is not limited to the above example. In addition, (1) subtract the local maximum value (including a constant multiple) (2) (function of CT value corresponding to each pixel value) × local Subtract maximum value (3) (statistical function of concentration distribution in specific direction area (i, j)) x subtract local maximum value (4) subtract average value (including constant multiple) near local maximum value (5) (A function of the angle formed by the gradient maximum direction) × local maximum value can be subtracted.

Here, the “function of the angle formed by the maximum gradient direction” is a function of the angle θ formed by the maximum direction of the past image density gradient and the maximum direction of the image density gradient of the day, and changes in the same manner as cos θ. However, it is not necessary to be exact cos θ,
When θ = 0 degree: 1.0
When θ = 45 degrees: 0.8
When θ = 90 degrees: 0.5
It is good also as values, such as.

As described above, the subtraction method is not limited to the procedure shown in FIG. That is,
(Step 100)
The maximum value max of the density value in the specific direction area (i, j) of the past image of the central pixel of the small area 54 is obtained.

(Step 101)
C × max is subtracted for each pixel (i, j) in the specific direction area (i, j) of the image 901 on the day. The result is sa1.

(Step 102)
It is determined whether or not “pixel (i, j) value of difference image 903> sa1”.

(Step 103)
When “pixel (i, j) value of difference image 903> sa1”, “pixel (i, j) of difference image 903 = sa1”.

  Note that the method of obtaining the maximum density gradient direction is not limited to FIG. 8, and as shown in FIG. 20, pixel density differences between two points passing through a circle C are sequentially compared, and the maximum density gradient is determined from the comparison result. It is also possible to determine the direction.

  Further, in the above description, the small area 54 is 3 vertical pixels × 3 horizontal pixels, but the small area 54 may have different sizes such as 4 vertical pixels × 4 horizontal pixels and 5 vertical pixels × 5 horizontal pixels. . Further, the size of the small region 54 may be changed depending on the type of image such as a CT image or a DR image.

  In the subtraction, the second and third largest values may be used instead of the local maximum value of the density value, or a constant multiple of the average value may be used according to the size of the small region 54. . Even in such a case, the effect of shadow enhancement can be obtained. Moreover, if data regarding the average value is used, a shadow-enhanced image with less noise than the local maximum value can be obtained.

  In the above description, the specific direction area is an area (3 pixels) in the maximum density gradient direction. However, as shown in FIG. 21, the specific direction area may be a small rectangular area or a small circular area. It is good. Further, the shape and size of the specific direction area may be different between the past image 902 and the current day image 901. Furthermore, in one of the past image 902 and the current day image 901, the specific direction area may have a size of 1 vertical pixel × 1 horizontal pixel, that is, 1 pixel. In this case, the density value of that one pixel becomes the density local maximum value as it is.

  In the above description, tomographic images such as CT images and MRI images are used. However, the present invention can also be applied to DR images as shown in FIG. In the DR image shown in FIG. 22, the density value of the part displayed in black is larger than the density value of the part displayed in white. In this case, subtraction according to a preferred embodiment of the present invention between the current day image 180 and the past image 181 results in 183 in FIG. 22, and 184 when combined with the contour image 182 that has been subjected to contour extraction processing.

In the above, the case where it is not necessary to align the image of the day and the past image has been described.
However, if the registration of the current day image and the past image is combined, a clearer shadow image can be obtained. This combination of alignment processing may be performed at any time as long as the effect of alignment is finally obtained, but is most preferably performed before the inter-image difference processing. Next, an idea when the inter-image registration process before the inter-image difference process is combined will be described with reference to FIGS. In alignment, image division, enlargement, reduction, movement, rotation, and the like are performed (see FIGS. 23 and 24). You may align as a whole, without dividing | segmenting an image.

  Further, every time an image is divided, a displacement vector is calculated, the divided image is deformed, and this is repeated until a predetermined number of divisions is reached to obtain a final displacement vector. In the alignment, an image obtained by deforming an original image (an image before division or deformation) with a final displacement vector is used.

  The current day image “a” and the past image “b” in FIG. 25 are aligned. Each of a and b is reduced (reduction ratio r) to a1 and b1, and correlation 1 between a1 and b1 (subtraction is performed to determine whether the difference is minimized) is taken to translate b1 b2 is obtained. Next, correlation between images after rotation of a1 and b2 is taken to obtain an image b3.

  Based on this information, as shown in FIG. 26, the image b is translated by dx / reduction ratio r and dy / reduction ratio r, and further an image b4 rotated by an angle ang is obtained. A displacement vector for alignment is obtained from the images a and b4. First, a is divided into four to be A4, and b4 is divided into B4. The correlation between the divided images aij and bij is obtained, and the displacement vector 4 is obtained. The image B4 is deformed with the displacement vector 4.

  Thereafter, further 16 divisions are performed, and correlation between aij and bij after division is obtained to obtain a displacement vector 16. Such division is repeated, and a displacement vector N (FIG. 27) for each pixel is obtained by spline interpolation of a displacement vector of 32 × 32 divisions having improved accuracy.

  Finally, as shown in FIG. 27, the image b4 (original image before division and deformation) is deformed using the displacement vector N, and an image after alignment is obtained.

  In this way, accurate image alignment can be performed by gradually finely dividing the image, and false positive shadows (pseudo shadows) due to imperfect alignment can be made difficult to occur.

  Note that the number of image divisions is not limited to 32 × 32 divisions, and different values may be set according to the images.

  Next, FIG. 28 and FIG. 29 show the flow of screen operations.

(Step 51)
The current day image 41 and the past image 42 are displayed.

(Step 52)
Judgment is “Next slice display button pressed?”. If yes, jump to step 57. If NO, go to step 53.

(Step 53)
Judge “Push the previous slice display button?”. If yes, jump to step 58. If NO, go to step 54.

(Step 54)
Judgment of “Positioning & difference button pressed?” If yes, jump to step 59. If NO, go to step 55.

(Step 55)
Judge whether "Difference FP delete button is pressed?" If yes, jump to step 60. If NO, go to step 56.

(Step 56)
Judge whether "Is the end button pressed?" If yes, exit. If NO, return to step 52.

(Step 57)
Display the next slice and jump to step 52.

(Step 58)
Display the previous slice and jump to step 52.

(Step 59)
Alignment & difference processing and jump to step 52.

(Step 60)
The false positive shadow after the difference (pseudo shadow) is reduced in density and jumped to step 52. As a method of reducing the concentration,
There are a density reduction method using a shadow feature value, a density reduction method using an edge region, and a density reduction method using a two-point replacement process.

  First, the density reduction method using the shadow feature amount will be described with reference to FIG.

(Step 61)
Perform difference processing.

(Step 62)
Perform differential image enhancement & binarization.

(Step 63)
Since there are cases where a plurality of regions are connected after the binarization processing, the narrowed portion of the binarized region is cut to make an isolated region.

(Step 64)
It is determined whether or not “the ratio of the length and width of each isolated region is calculated value> predetermined value 1”. If Yes, jump to Step 69 and reduce the concentration as a false positive.

(Step 65)
It is determined whether or not “the circularity of each isolated region is calculated value> predetermined value 2”. If Yes, jump to Step 69 and reduce the concentration as a false positive.

(Step 66)
Other determination processing is performed. If Yes, jump to Step 69 and reduce the concentration as a false positive.

(Step 67)
A circle is marked as a cancer shadow candidate at a high density portion of the difference image.

(Step 68)
It is determined whether all shadows have been determined. If yes, end. If NO, jump to step 64.

(Step 69)
Reduce the concentration of false positive shadows.

  Next, a method for reducing the density using the edge region will be described with reference to FIGS. In the example of FIG. 31, when the difference process between the current day image 1 and the past image 2 is performed, an image 3 is obtained. If there is no cancer shadow in the past image and there is a cancer shadow in the image on the day, the difference image mainly displays the cancer shadow and the portion due to the positional deviation of the entire image. Here, it is easier to find cancer shadows if only the cancer shadows are displayed.

  FIG. 32 is a binarized image after emphasizing the edges of the CT images of the current day image and the past image, and when these images are correlated, an image 20 is obtained. Using the image 20, the density of the region where the value of the edge image 20 is 1 in 3 of FIG. Thereby, the pseudo-shadow-reduced image 21 in which the false positive after the difference is reduced in density is obtained.

  Note that the image 20 can be obtained by correlating regions having a density greater than the threshold value without binarization in advance.

  FIG. 33 shows a case where this process is applied to a DR image.

  Images with reduced false positives may contain shadows that are suspected of being cancerous. Without subtraction, it is also possible to ask for the judgment of the doctor by combining and displaying the result of detecting an abnormal shadow candidate only with each image.

  Next, the density reduction process by the two-point replacement process will be described with reference to FIG.

  When the values of the pixels p1 and p5 are smaller than the predetermined value, p1 to p5 are replaced with the minimum value of p1 to p5. In this case, the effect of reducing the concentration can be obtained even if p2 to p4 are replaced with the minimum values of p2 to p4.

  Also, when the values of the pixels p1 and p5 are larger than a predetermined value, p1 to p5 may be replaced with the maximum value of p1 to p5. In this case, the effect of reducing the concentration can be obtained even if p2 to p4 are replaced with the maximum values of p2 to p4.

  In this way, the medical image processing apparatus 10 makes it difficult to generate false positive shadows (pseudo shadows) by accurately aligning the current day image and the past image, and lowers the density of the generated false positive shadows. Abnormal shadows such as cancer can be easily found.

  Next, another example of the operation screen of the medical image processing apparatus 10 is shown in FIGS. A CT image 21 by the simple difference method and a shadow enhanced image 22 by the anisotropic difference method are displayed on the CRT 15. On the CRT 15, the cursor of the mouse 17 is displayed on the CT image 21 and the shadow-enhanced image 22, and these cursors are also interlocked when the mouse 17 is moved, and the corresponding positions are displayed. It is easy to grasp.

  In addition, as shown in FIG. 36, three images of the current day image 31, the past image 32, and the difference image 33 after alignment may be displayed and the cursor of the mouse 17 may be linked.

  In general, in the method of finding the maximum value in the search area, the same address of the Mth image of the past image having the highest correlation with the Nth image of the current day is scanned with dots (sequential scanning points). In the scanning, a two-dimensional image parallel to the maximum density gradient direction as shown in FIG. 8, 9 or 18 is assumed.

  However, if only the two-dimensional image has a slightly different blood vessel positional relationship between the current day image and the past image, it is difficult to cope with it. For example, the body shape of the subject may change due to fatness or thinness between when the past image is captured and when the current day image is captured. Therefore, in addition to the two-dimensional image, a concept of a three-dimensional image formed by stacking a plurality of continuous two-dimensional images is introduced.

  For example, the (N + 1) th and (N-1) th images before and after (or above and below) the Nth image of the current day image shown in FIG. 37, and the (before and after) (or above and below) images of the Mth past image shown in FIG. The three-dimensional processing is performed using the M + 1 and M−1 images.

  Specifically, as shown in FIG. 39, when the coordinates X, Y of the Nth image are represented as (N, X, Y), for example, the subtraction value of (N, X1, Y1) points is The density value V1 of the past image (M, x1, y1) is subtracted from the density value V1 of the image (N, X1, Y1) point, and the subtraction value of the (N, X2, Y2) point is the current day image (N , X2, Y2) is a value obtained by subtracting the density value v2 of the past image (M-1, x2, y2) from the density value V2 of the point. Here, the search area (plane) is parallel to the maximum density gradient direction of the sequential scanning points of the Mth image. If the subtraction value is negative, the subtraction value may be replaced with zero or the absolute value of the subtraction value. In the subtraction, the current day image may be subtracted from the past image.

  As shown in FIG. 40, the search area for finding the maximum value may be a cube including scanning points. As shown in FIG. 41, the boundary of the search area may be on an interpolated image. As shown in FIG. 42, when the slice thickness of the tomographic image is thin, the boundary of the search area may extend over a plurality of original images. At this time, as shown in FIG. 43, it is preferable to obtain the local maximum value in consideration of the three-dimensional concentration gradient maximum direction.

  FIG. 44 shows a case where a MIP (Maximum Intensity Projection) image is created from at least two multiple images including past images at slice positions corresponding to the current image.

  Due to the influence of the movement of the subject, there is a blood vessel or the like that appears in only one image, and the shadow of this blood vessel or the like may remain after the inter-image difference processing. This means that blood vessels are shown in an image adjacent to an image in which no blood vessels are shown. Therefore, for example, 3 images (the number of sheets depends on the thickness of the slice depending on the slice thickness). If it is thin, the number of images increases.) If the MIP image is made, the missing portion can be compensated.

  After imaging, the position of the blood vessel or the like is not aligned, so the alignment calculation is performed.To solve the problem that it is difficult to align the position completely by the alignment calculation, in the method according to the present embodiment, The maximum value of the pixel values in the vicinity of (x, y) of the past image b is subtracted from the pixel coordinates (x, y) of the small area of the image a on the current day. The principle is the same as that described with reference to FIG.

  Considering both of the above, in the method according to the present embodiment, the maximum value of the pixel values in the vicinity of (x, y) of the MIP image including the past image b is subtracted from the (x, y) pixel value of the image a on the day.

  The MIP image B1 is created from at least two images including the image b3 corresponding to the image a1. The MIP image B1 may be created by reconstructing a thick slice image from the measurement data. The MIP image Bn is created similarly.

  The difference image 1 is a difference between the image a1 and the MIP image B1, and the difference image 2 is a difference between the image a2 and the MIP image B2. Similarly, the difference image n is a difference between the image an and the MIP image Bn.

  FIG. 45 shows a case where the MIP image Bn is created from four images including an image b6 corresponding to the image a4 in a vertically asymmetric manner as a modified example of the interpolation image creation.

  Next, a subtraction method which is a main process in the medical image processing apparatus 10 will be described with reference to FIG.

(Step 161)
A pair of images having the highest correlation between image groups with different shooting dates (a is the current day image and b is the past image) is obtained.

(Step 162)
Next, the target part is extracted from a plurality of images including the image b whose date is earlier. This step may be omitted when the region is the head.

(Step 163)
An MIP image c is created using the extracted image. If the extraction step is omitted, the MIP image c is created as it is. Furthermore, when creating a MIP image, the range in which the maximum value is found, that is, the range in which the maximum value is found from adjacent slice images may be set as the CT values CT1 to CT2. Further, a tomographic image having a thick slice thickness may be reconstructed and used instead of the MIP image c.

(Step 164)
Image a and image c are aligned. At the time of this alignment, if necessary, the alignment process is facilitated by matching the average density and the density histogram.

(Step 165)
Substitute initial values for x and y. The initial value in this case means the start address of the image read address.

(Step 166)
The maximum pixel value (including a value obtained by multiplying the subtraction value by a constant value) is subtracted from the pixel value at the point (x, y) of the image a in the small area including (x, y) of the image c.

(Step 167)
Using the subtracted value, it is converted to a value suitable for display and stored in the display memory.

(Step 168)
Add 1 to x. That is, the image readout address x is updated.

(Step 169)
It is determined whether or not the image read address x is the maximum value. If it is not the maximum value, the process proceeds to step 166, and if it is the maximum value, the process proceeds to step 16A.

(Step 16A)
Add 1 to y. That is, the image read address y is updated.

(Step 16B)
It is determined whether the image read address y is the maximum value. If it is not the maximum value, the process proceeds to step 16C, and if it is the maximum value, the process is terminated.

(Step 16C)
Substitute an initial value for x. That is, x is set to an initial address by updating the image read address y.

  Next, a method of combining the result of the detection process using the original image and the result of the detection process using the difference image will be described with reference to FIG.

(Step 171)
A detection process (A) using the original image is performed.

(Step 172)
A detection process (B) using the difference image is performed.

(Step 173)
It is determined whether the operator has selected the detection rate with priority. If detection priority is selected, the process proceeds to step 174. If not selected, the process proceeds to step 175.

(Step 174)
An OR process is performed on the detected portions of both A and B, and then the process ends.

(Step 175)
It is determined whether or not the operator preferentially selects false positive reduction. If false positive reduction priority is selected, the process proceeds to step 176. If not selected, the process is terminated.

(Step 176)
An AND process is performed on the detected portions of both A and B, and then the process ends.

  Furthermore, the difference processing is not performed between all images of the current day image and the past image, but only the past image corresponding to the image where the abnormal part is detected by the abnormal shadow detection processing using the original day image is automatically databased. May be read out from the image and the difference process with the image of the day may be performed.

  In the processing described above, the past image and the current day image may be exchanged, and the same result can be obtained.

  As described above, according to the present invention, the shadow of an image is enhanced, and an abnormal shadow such as cancer can be easily found.

FIG. 1 is a diagram showing a configuration of a medical image processing apparatus according to an embodiment of the present invention; FIG. 2 is a functional block diagram of FIG. FIG. 3 is a diagram showing an operation screen of the medical image processing apparatus; FIG. 4 is a diagram showing an outline of the difference processing; FIG. 5 is a diagram showing an example of the day image; FIG. 6 is a diagram showing an example of a simple difference image; FIG. 7 is a diagram showing an example of the procedure of difference processing; FIG. 8 is a diagram showing an example of obtaining the maximum concentration gradient direction; FIG. 9 is a diagram showing another example of obtaining the maximum concentration gradient direction; 10 (a) and 10 (b) are diagrams showing a comparison between simple difference processing and anisotropic difference processing; FIG. 11 is a diagram showing an example of an image by anisotropic difference processing; FIG. 12 is a diagram showing an image with an abnormal shadow circled; FIG. 13 is a diagram showing another example of the day image; FIG. 14 is a diagram showing another example of an image by difference processing according to the present embodiment; FIG. 15 is a diagram illustrating the synthesis of separately processed difference images and contour images; FIG. 16 is a diagram illustrating an example of a composite image of a difference processed image and a contour image; FIG. 17 is a diagram showing a synthesis procedure for each pixel of the difference image and the contour image; 18 is a diagram showing an example in which the maximum concentration gradient direction is different from FIG. 9; FIG. 19 is a diagram showing another example of the difference processing procedure; FIG. 20 is a diagram showing another example of obtaining the maximum concentration gradient direction; FIG. 21 is a diagram showing another example of the specific direction area; FIG. 22 is a diagram showing an example of difference processing in a DR image; FIG. 23 is a diagram showing a schematic alignment between the current day image and the past image by the simple difference method; FIG. 24 is a diagram showing the detailed alignment of the current day image and the past image by the simple difference method; FIG. 25 is a diagram showing schematic alignment of images according to the present embodiment; FIG. 26 is a diagram showing detailed registration of an image according to the present embodiment; FIG. 27 is a diagram showing another example of detailed image alignment according to the present embodiment; FIG. 28 is a diagram showing a screen example of the medical image processing apparatus; FIG. 29 is a diagram showing a flow of screen operations of the medical image processing apparatus; FIG. 30 is a diagram showing a density reduction process using feature amounts; FIG. 31 is a diagram showing a false positive concentration reduction process using a correlation image between edge images; FIG. 32 is a diagram showing another example of false positive density reduction processing using correlation images between edge images; FIG. 33 is a diagram showing a density reduction process using a correlation between edge images in which edges of a DR image are emphasized; FIG. 34 is a diagram showing a concentration reduction process by a two-point replacement process; FIG. 35 is a diagram showing an example of cursor-linked display on the display screen; FIG. 36 is a diagram showing another example of cursor-linked display on the display screen; FIG. 37 is a diagram showing an example of how blood vessels are captured in a plurality of images on the day; FIG. 38 is a diagram showing an example of how blood vessels are captured in a plurality of past images; FIG. 39 is a diagram for explaining the difference processing between the current day image and the past image in FIGS. 37 and 38; FIG. 40 is a diagram showing an example of how to take different local regions; FIG. 41 is a diagram showing another example of how to take different local regions; FIG. 42 is a diagram showing another example of how to take different local regions; FIG. 43 is a diagram showing another example of how to take different local regions; FIG. 44 is a diagram showing another example of how to take different local regions; FIG. 45 is a diagram showing another example of how to take different local regions; FIG. 46 is a diagram showing another processing example of shadow enhancement; FIG. 47 is a diagram showing a subroutine used in the processing of step 162 in FIG.

Claims (9)

First storage means for storing a predetermined portion of the subject imaged by the medical image diagnostic apparatus as a first medical image;
Second storage means for storing a second medical image of the same subject that is different from the date and time when the first image stored in the first storage means was taken;
First setting means for setting a plurality of first local regions having at least one pixel as an element in the first medical image stored by the first storage means;
A plurality of second local areas corresponding to each of the plurality of first local areas set by the first setting means and having a width larger than those of the first local areas are stored in the second storage means. Second setting means for setting the second image stored in
Reference value calculating means for calculating a reference value of density value for each of the plurality of second local regions set by the second setting means;
An enhanced image for each first local region based on the reference value calculated by the reference value calculating unit and the density value of each pixel of the plurality of first local regions set by the first setting unit Image creation means for creating
Display means for displaying the emphasized image created by the image creating means,
The reference value calculated by the reference value calculating means is at least one of a local maximum value, a local average value, a local median for each second local region set by the second setting means, or a constant multiple thereof. Yes,
The reference value calculating means further includes a direction calculating means for calculating the maximum concentration gradient direction of the second local region set by the second setting means, and the reference value calculating means is calculated by the direction calculating means. The maximum value of the density value from the pixel value along the density gradient maximum direction, a value obtained by multiplying the maximum value by a constant, a value obtained by multiplying the maximum value by a function of the density value, and an approximation function of the density distribution to the maximum value. A value obtained by multiplying the multiplied value, the maximum value by a constant, and a function of an angle formed by the density gradient maximum direction of the first image and the density gradient maximum direction of the second image becomes the maximum value. A medical image processing apparatus, wherein the local maximum value is calculated using any one of average values of density values in the vicinity of a pixel. The image creating means calculates a difference value between the reference value calculated by the reference value calculating means and the density value of each pixel of the plurality of first local regions set by the first setting means, The medical image processing apparatus according to claim 1, wherein an emphasized image is created for each of the first local regions based on the calculated difference value. The image creating means includes a contour image generating means for generating a contour image of the first or second medical image, and an image superimposing means for superimposing the contour image and the emphasized image,
The medical image processing apparatus according to claim 1, wherein the display unit displays the superimposed image superimposed by the image superimposing unit. The medical image processing apparatus according to claim 1, further comprising an alignment unit configured to align the first medical image and the second medical image. Further comprising a density reducing means for creating a pseudo-shadow-reduced image by reducing the density of the pseudo-shadow generated by the incomplete alignment of the positioning means in the enhanced image created by the image creating means,
The medical image processing apparatus according to claim 4, wherein the display unit displays a pseudo-shadow-reduced image created by the density reduction unit. A pointing device for operating a cursor displayed on the display means;
The display means displays in parallel at least two images among the first medical image, the second medical image, and the emphasized image created by the image creation means, and the pointing is displayed on each of the images displayed in parallel. Displays the cursor operated by the device,
The medical image processing apparatus according to claim 1, wherein the cursor is interlocked by an operation of the pointing device on each of the images displayed in parallel. The reference value calculation means includes three-dimensional information including MIP images formed by the plurality of second medical images adjacent to each other in the slice direction for the plurality of second local regions set by the second setting means. The medical image processing apparatus according to claim 1, wherein the reference value is calculated by: 2. The display condition setting means for setting whether to give priority to an abnormal shadow detection rate or false positive reduction as a display condition of an emphasized image displayed on the display means. Medical image processing apparatus. A first storage step of storing a predetermined portion of the subject imaged by the medical image diagnostic apparatus as a first medical image;
A second storage step of storing a second medical image of the same subject that is different from the date and time when the first image stored in the first storage step was taken;
A first setting step of setting a plurality of first local regions having at least one pixel as an element in the first medical image stored in the first storage step;
A plurality of second local regions corresponding to each of the plurality of first local regions set by the first setting step and having a width equal to or larger than those first local regions are stored in the second storage step. A second setting step for setting the second image stored in
A reference value calculating step for calculating a reference value of the density value for each of the plurality of second local regions set by the second setting step;
The enhanced image for each first local region based on the reference value calculated by the reference value calculating step and the density value of each pixel of the plurality of first local regions set by the first setting step Image creation process to create
A display step for displaying the emphasized image created by the image creation step;
With
The reference value calculated by the reference value calculating step is at least one of a local maximum value, a local average value, a local median for each second local region set by the second setting step, or a constant multiple thereof. Yes,
The reference value calculating step further includes a direction calculating step of calculating a concentration gradient maximum direction of the second local region set by the second setting step, and the reference value calculating step is calculated by the direction calculating step. The maximum value of the density value from the pixel value along the density gradient maximum direction, a value obtained by multiplying the maximum value by a constant, a value obtained by multiplying the maximum value by a function of the density value, and an approximation function of the density distribution to the maximum value. A value obtained by multiplying the multiplied value, the maximum value by a constant, and a function of an angle formed by the density gradient maximum direction of the first image and the density gradient maximum direction of the second image becomes the maximum value. A medical image processing method, wherein the local maximum value is calculated using any one of average values of density values in the vicinity of a pixel.

Images