Image processing apparatus and image processing method

Abstract

To accurately highlight an attention area in an image.SOLUTION: An image processing apparatus has: an inner region setting unit 102 and a peripheral region setting unit 103 that set a peripheral region in a medical image with respect to an inner region in the medical image; an intensity value group probability distribution calculation unit 105 that calculates an inner region histogram about the inner region, and also calculates a peripheral region histogram about the peripheral region; a probability difference calculation unit 106 that calculates a differential value between the inner region histogram and the peripheral region histogram about prescribed respective intensity values, and thereby calculates a probability difference distribution; a tissue image processing unit 104 that generates tissue images on the basis of the medical image; a detection range setting unit 107 that sets a detection range with respect to each tissue image; a mask setting unit 108 that selects pixels for highlighting each tissue image on the basis of the detection range; and a display processing unit 110 that outputs the highlighted tissue images.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to techniques for an image processing device and an image processing method. [Background technology]

[0002] In order to identify lesions such as cerebral infarction in black-and-white contrast images of organs such as the brain, currently, the judgment of an experienced specialist is required. In addition, the task of marking (coloring) the image based on this judgment is also cumbersome because it is done manually.

[0003] Patent Document 1 discloses a control method, information terminal, and program in which "when the lesion included in the designated target image is a texture-type lesion, the probability image calculation unit 102 calculates a probability value that each pixel of the designated target image is included in a lesion area. The output unit 103 calculates, in the probability image acquired from the probability image calculation unit 102, an area consisting of pixels having a probability value equal to or greater than a first threshold as a candidate area, and calculates an area consisting of pixels having a probability value within a predetermined probability range including the first threshold as a correction area. The input unit 104 detects a user's input to pixels of the correction area displayed on the display by the output unit 103 by operating the operation device. The lesion area designation unit 105 designates a lesion area based on the probability image acquired from the probability image calculation unit 102, the candidate area and correction area acquired from the output unit 103, and the user's operation information acquired from the input unit 104" (see abstract). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2018-102916 A Summary of the Invention [Problem to be solved by the invention]

[0005] The technology described in Patent Document 1 does not disclose an algorithm for identifying a lesion site (for example, it states that machine learning is sufficient). In other words, the technology described in Patent Document 1 does not disclose an algorithm for identifying a lesion site, making it difficult to improve the accuracy of determining the lesion site.

[0006] In addition, (semi)automatic techniques for determining the location of lesions using a predetermined threshold value for intensity data on an image, and (semi)automatic techniques using machine learning with a large number of images have been proposed, but the accuracy (performance) of identifying lesion locations cannot be said to be high (60-80%).

[0007] In addition, the way diseased areas are depicted differs for each tissue that makes up the brain; that is, some tissues are depicted bright (white) and some tissues are depicted dark (black); however, conventional technology does not take such a phenomenon into account.

[0008] The present invention has been made in view of the above background, and an object of the present invention is to enhance a region of interest in an image with high accuracy. [Means for solving the problem]

[0009] In order to solve the above-mentioned problems, the present invention provides a region setting unit that sets, in a first image, a second region that is a region adjacent to a first region that is set in the first image as a region including a region of interest in the first image and does not include the first region; an intensity value frequency distribution calculation unit that calculates, for the first region, a first intensity value frequency distribution that is information regarding a distribution of the number of pixels having a predetermined intensity value, and calculates, for the second region, a second intensity value frequency distribution that is information regarding the distribution of the number of pixels having a predetermined intensity value; the difference value being a detection range setting unit that applies a detection range setting algorithm to each of the second images and sets a detection range of the intensity value of the first image for each of the second images based on the difference value; a pixel selection unit that selects pixels to be highlighted for each of the second images based on the detection range set for each of the second images; and an output processing unit that outputs to an output unit a third image in which the pixels selected by the pixel selection unit are highlighted. Other solutions will be described in the embodiments as appropriate. Effect of the Invention

[0010] According to the present invention, it is possible to highly accurately highlight a region of interest in an image. [Brief description of the drawings]

[0011] [Figure 1] 1 is a diagram illustrating an example of the configuration of an image processing system according to a first embodiment. [Diagram 2] FIG. 1 illustrates a configuration of a computer. [Diagram 3] FIG. 2 is a diagram illustrating a hardware configuration of a computer. [Figure 4] 13 is a flowchart showing the overall procedure of a lesion highlighting process performed by a computer. [Diagram 5] 13 is a flowchart showing a detailed procedure of an inner region setting process. [Figure 6] 10 is a flowchart showing a detailed procedure of a surrounding area setting process. [Figure 7] FIG. 13 is a diagram showing an example of an inner region and a peripheral region. [Figure 8] 10 is a flowchart showing the detailed steps of a tissue decomposition process. [Figure 9A] FIG. 1 is a diagram illustrating an example of a medical image. [Figure 9B] FIG. 1 is a diagram showing an example of a tissue image (part 1). [Figure 9C] FIG. 2 is a diagram showing an example of a tissue image (part 2). [Figure 9D] FIG. 3 shows an example of a tissue image (part 3). [Figure 9E] FIG. 4 shows an example of a tissue image. [Figure 10] 13 is a flowchart showing a detailed procedure for calculating the intensity value group probability distribution. [Figure 11A] FIG. 1 is a diagram showing an example of an intensity value group probability distribution (part 1). [Figure 11B] FIG. 2 is a diagram showing an example of an intensity value group probability distribution (part 2). [Figure 11C] FIG. 3 is a diagram showing an example of an intensity value group probability distribution (part 3). [Figure 11D] FIG. 4 is a diagram showing an example of an intensity value group probability distribution. [Figure 12A] This is a diagram showing the probability difference distribution (part 1). [Figure 12B] This is a diagram showing the probability difference distribution (part 2). [Figure 12C] This is a diagram showing the probability difference distribution (part 3). [Figure 13] 10 is a flowchart showing a detailed procedure for setting a detection range. [Figure 14] 10 is a flowchart showing a detailed procedure for determining the position of an intensity section. [Figure 15] 13 is a flowchart showing a detailed procedure for expanding an intensity section. [Figure 16]13 is a flowchart showing a detailed procedure of strength section adjustment. [Figure 17] 1 is a flowchart showing the detailed procedure of the forward method. [Figure 18] FIG. 11 is a diagram (part 1) showing an example of the cumulative intensity value group probability and cumulative probability difference of the surrounding area obtained using the forward method. [Figure 19] 1 is a flowchart showing a detailed procedure of the backward method. [Figure 20] FIG. 13 is a diagram (part 1) showing an example of the cumulative intensity value group probability and cumulative probability difference of the surrounding area obtained using the backward method. [Figure 21] FIG. 2 is a diagram (part 2) showing an example of the cumulative intensity value group probability and cumulative probability difference of the surrounding area obtained using the forward method. [Figure 22] FIG. 2 is a diagram (part 2) showing an example of the cumulative intensity value group probability and cumulative probability difference of the surrounding area obtained using the backward method. [Diagram 23] FIG. 13 is a diagram showing a mask for the lesion highlighting process that is generated in accordance with the set detection range. [Figure 24A] FIG. 1 shows a masked tissue image (part 1). [Figure 24B] FIG. 2 shows a masked tissue image (part 2). [Figure 24C] FIG. 3 shows a masked tissue image (part 3). [Figure 24D] FIG. 4 shows a masked tissue image. [Figure 24E] FIG. 13 is a diagram showing an example of a mask image. [Diagram 25] FIG. 4 is a diagram showing an example of a display screen displayed on a display device. [Figure 26] FIG. 11 is a diagram illustrating a configuration of a computer according to a second embodiment. [Figure 27A] FIG. 1 is a diagram (part 1) showing an outline of a method for generating a three-dimensional image. [Figure 27B] FIG. 2 is a diagram (part 2) showing an outline of a method for generating a three-dimensional image. [Figure 28] 13 is a flowchart showing a process for highlighting a lesion area in a three-dimensional image. [Figure 29] 11 is a flowchart illustrating a detailed procedure of downward detection. [Diagram 30] 13 is a flowchart showing a detailed procedure of upward detection. [Figure 31A] FIG. 1 shows a mask image. [Figure 31B] FIG. 2 shows a mask image (part 2). [Figure 31C] FIG. 3 shows a mask image. [Figure 31D] FIG. 4 shows a mask image. [Figure 31E] FIG. 5 shows a mask image. [Diagram 32] 13 is a flowchart showing the overall procedure of lesion site highlighting processing performed in the third embodiment. [Diagram 33] 1 is a flowchart showing a detailed procedure for generating a mixed tissue image. [Figure 34A] FIG. 2 shows a medical image before being decomposed. [Figure 34B] FIG. 13 shows an example of a mixed tissue image. [Diagram 35] FIG. 13 is a diagram showing an example of a probability distribution of intensity values ​​from a mixed tissue image. [Diagram 36] FIG. 13 is a diagram showing an example of a probability difference distribution of a mixed tissue image. [Figure 37] 10 is a flowchart showing a detailed procedure for setting a detection range for a mixed tissue image. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Next, a mode for carrying out the present invention (referred to as an "embodiment") will be described in detail with reference to the drawings as appropriate.

[0013] [First embodiment] (Image Processing System Z) FIG. 1 is a diagram showing an example of the configuration of an image processing system Z according to the first embodiment. The image processing system Z includes a medical image capturing unit M. The image processing system Z also includes an image processing apparatus 1 that processes an image captured by the medical image capturing unit M. The image processing apparatus 1 sets a mask that highlights a lesion site in a medical image 201 (see FIG. 7). The image processing apparatus 1 includes a computer 100, a storage device 120, an input device 131 that is an input unit, and a display device 132 that is an output unit. The computer 100 will be described in detail later. The storage device 120 stores the medical image 201 acquired from the medical image capturing unit M, and information such as a threshold value and a mask that will be described later. In the example shown in FIG. 1, the storage device 120 is integrated with the image processing apparatus 1, but the storage device 120 may be a device separate from the image processing apparatus 1, such as a database. The input device 131 includes a mouse, a keyboard, and the like. The display device 132 is a display.

[0014] The medical imaging device M is an MRI (Magnetic Resonance Imaging) imaging device, a CT (Computed Tomography) imaging device, etc. The medical image 201 is an MRI image, a CT image, etc.

[0015] (Computer 100) FIG. 2 is a diagram showing the configuration of the computer 100. The computer 100 includes an image acquisition unit 101 and an inner region setting unit 102 which is a region setting unit. The computer 100 also includes a peripheral region setting unit 103 which is a region setting unit. The computer 100 also includes a component image processing unit 104 which is an image processing unit. The computer 100 also includes an intensity value group probability distribution calculation unit 105 which is an intensity value frequency distribution calculation unit.

[0016] The computer 100 also includes a probability difference calculation unit 106. And a detection range setting unit 107. The computer 100 also includes a mask setting unit 108 which is a pixel selection unit, a mask compilation unit 109 which is also a pixel selection unit, and a display processing unit 110 which is an output processing unit.

[0017] The image acquisition unit 101 acquires a medical image 201 (see FIG. 7), which is a first image, from the storage device 120. The inner region setting unit 102 sets an inner region 212 (see FIG. 7), which is a first region, in the medical image 201. In addition, the surrounding region setting unit 103 sets a surrounding region 222 (see FIG. 7), which is a second region, in the medical image 201. The inner region 212 and the surrounding region 222 will be described later.

[0018] The tissue image processing unit 104 registers an anatomical model or the like in the medical image 201, and decomposes the medical image 201 into a tissue image 401 (see Figs. 9B to 9E) which is an image derived from anatomical tissue and is a second image. The tissue is anatomical tissue of the brain such as gray matter, white matter, cerebrospinal fluid, soft tissue, etc. However, an image other than the brain may be used as the medical image 201, and in such a case, tissue other than the brain is applied to the tissue.

[0019] 11A to 11D, calculates an inner region histogram 501 which is a first intensity value frequency distribution, and a surrounding region histogram 502 which is a second intensity value frequency distribution. The inner region histogram 501 and the surrounding region histogram 502 are distributions of the number of pixels having brightness values ​​within predetermined bins for the set inner region 212 and the set surrounding region 222, respectively.

[0020] Probability difference calculation unit 106 calculates probability difference 601 by subtracting peripheral region histogram 502 from inner region histogram 501 for each intensity value group, and generates probability difference distribution 600 (see FIGS. 12A to 12C), which is a probability difference frequency distribution. Detection range setting unit 107 sets an intensity threshold that represents the region of the lesion site (lesion region 202) (see FIG. 7) in medical image 201. Mask setting unit 108 highlights pixels in medical image 201 based on the set detection range.

[0021] The intensity value group probability distribution calculation unit 105, the probability difference calculation unit 106, the detection range setting unit 107, and the mask setting unit 108 execute steps for each tissue. The mask editing unit 109 synthesizes masks generated from all tissues. The display processing unit 110 displays on the display device 132 a mask image 1003 (see FIG. 24E) in which pixels corresponding to the mask synthesized by the mask editing unit 109 are highlighted.

[0022] (Hardware) FIG. 3 is a diagram showing the hardware configuration of the computer 100. As shown in FIG. 3, the computer 100 includes a memory 111 including a RAM (Random Access Memory) and the like. The computer 100 also includes a processing unit 112 including a CPU (Central Processing Unit) and a GPU (Graphic Processing Unit). A communication device 113 transmits and receives information to and from the medical imaging apparatus M shown in FIG.

[0023] Then, the program stored in the storage device 120 is loaded into the memory 111, and the loaded program is executed by the arithmetic unit 112. In this way, the image acquisition unit 101 to the display processing unit 110 shown in FIG.

[0024] (Overall processing) Fig. 4 is a flow chart showing the overall procedure of the lesion highlighting process, which is an image processing method performed by computer 100. Fig. 2 will be referred to as appropriate. Details of each step will be described later. First, a medical image 201 (see FIG. 7) captured by a medical imaging device M using medical equipment is acquired by the computer 100. The acquired medical image 201 is stored in the storage device 120. Next, the image acquisition unit 101 reads out the medical image 201 stored in the storage device 120 (Read medical image: S101). Next, the display processing unit 110 displays the read medical image 201 on the display device 132 (Display medical image: S102).

[0025] After that, the inner region setting unit 102 sets the inner region 212 (see FIG. 7) (Set inner region: S110: region setting step). The process of step S110 will be described later. Next, the peripheral region setting unit 103 sets the peripheral region 222 (see FIG. 7) based on the set inner region 212 (Set peripheral region: S120: region setting step). The process of step S120 will be described later. Then, the tissue image processing unit 104 decomposes the medical image 201 into images of each tissue (Decompose components from medical image: S130: tissue decomposition step). At this time, the tissue image processing unit 104 determines pixels for each tissue of the medical image 201. The image decomposed for each tissue is called a tissue image 401 (see Figs. 9B to 9E). The processing of step S130 will be described later. Note that at the stage of step S130, a mixed tissue image 260 (see Fig. 34B) may be generated by combining only tissue images 401 related to tissues in which the user is interested. The mixed tissue image 260 will be described later in the third embodiment. Next, the intensity value group probability distribution calculation unit 105 calculates the intensity value group probability distribution 500 (see FIGS. 11A to 11D) for the inner region 212, and further calculates the intensity value group probability distribution 500 for the peripheral region 222 (Calculate intensity value group probability distribution: S140: intensity value frequency distribution calculation step). The processing of step S140 will be described later.

[0026] Thereafter, the probability difference calculation unit 106 calculates a probability difference 601 (see FIGS. 12A to 12C) by subtracting the surrounding region histogram 502 (see FIG. 11) of the surrounding region 222 from the inner region histogram 501 (see FIG. 11) of the inner region 212 (Calculate Probability Difference: S151: probability difference calculation step). The processing of step S151 will be described later. Then, the detection range setting unit 107 sets a detection range based on the probability difference 601 (Set detection range: S160: detection range setting step), and the mask setting unit 108 sets a mask (Set mask: S171: pixel selection step). The processing of step S160 will be described later. As a result, the lesion site is highlighted (masked) in each tissue image 401. The processes of steps S140, S151, S160, and S171 are performed for each tissue image 401. That is, the processes of steps S140, S151, S160, and S171 are looped until the processes for all tissue images 401 are completed. Next, the mask editing unit 109 compiles (composes) the mask (Compile mask: S181). In step S181, masks that highlight the lesion site in each of the tissue images 401 are composed. As a result, masks that highlight the lesion sites detected in all of the tissue images 401 are composed. Note that when the mixed tissue image 260 (see FIG. 35B) shown in the third embodiment is used, the process of step S181 is omitted. Then, the display processing unit 110 displays (outputs) on the display device 132 a mask image 1003 (see FIG. 24E) in which the mask synthesized by the mask editing unit 109 is highlighted on the medical image 201 (Display: S191: output step).

[0027] (Setting the inner region 212) Next, the process of setting the inner region 212 by the inner region setting section 102 will be described with reference to FIGS. Fig. 5 is a flowchart showing a detailed procedure of the inner region setting process for setting the inner region 212 shown in Fig. 7. That is, Fig. 5 shows detailed processing of step S110 in Fig. 4. Fig. 7 is a diagram showing an example of the inner region 212 and the peripheral region 222. Fig. 7 shows a medical image 201 relating to the brain. Figs. 2 and 3 will be referred to as appropriate. First, the user designates an inner line 211 in a medical image 201 shown in FIG. 7 (Designate inner line: S111 in FIG. 5). The user designates the inner line 211 as a line that is considered to be a lesion site in the medical image 201 and surrounds a lesion area 202, which is an area of ​​interest. The lesion area is visually determined in advance by a user (such as a doctor). The inner line 211 is designated by the user dragging or clicking a plurality of points with a mouse (input device 131) or the like. At this time, the user designates the inner line 211 so as to include as much of the lesion area 202 as possible.

[0028] The region inside the inner line 211 is referred to as the inner region 212. In this manner, the inner region setting unit 102 sets the inner region 212 in the medical image 201 (Set inner region: S112 in FIG. 5). In this manner, the inner line 211 is set via the input device 131, and the inner region setting unit 102 sets the inner region 212 in the medical image 201. After completion of step S112 in FIG. 5, the computer 100 returns the process to step S120 in FIG. 4.

[0029] In this manner, the inner region setting unit 102 sets the inner region 212 in the medical image 201 as a region that includes the lesion region 202 in the medical image 201 .

[0030] Incidentally, the inner line 211 can be set not only as a polygon as shown in FIG. 7, but also as an ellipse.

[0031] (Setting the surrounding area 222) Next, the process of setting the surrounding area 222 will be described with reference to FIGS. Fig. 6 is a flowchart showing a detailed procedure of the surrounding region setting process for setting the surrounding region 222. That is, Fig. 6 shows detailed processing of step S120 in Fig. 4. Fig. 7 is a diagram showing an example of the inner region 212 and the surrounding region 222. Fig. 7 shows a medical image 201 relating to the brain. Figs. 2 and 3 will be referred to as appropriate. In the process shown in FIG. 6, the surrounding area setting unit 103 sets the surrounding area 222 based on the inside line 211 . First, the surrounding region setting unit 103 acquires an enlargement parameter (x; x is a real number satisfying x>1) input in advance by the user via the input device 131 (Recognize input of enlargement parameter: S121 in FIG. 6).

[0032] Next, the surrounding region setting unit 103 sets an enlarged region 231 (region adjacent to the first region) that is an area of ​​the inner region 212 shown in Fig. 7 multiplied by x (Expand inner region by enlargement parameter: S122 in Fig. 6). As shown in Fig. 7, the enlarged region 231 is surrounded by an outer line 221. Also, the example shown in Fig. 7 shows a case where x=1.2. Furthermore, the peripheral region setting unit 103 sets the region between the outer line 221 and the inner line 211 in the medical image 201 as the peripheral region 222 (Set peripheral region: S123 in Fig. 6). As shown in Fig. 7, the peripheral region 222 is a region in the vicinity of the inner region 212, and is set by the peripheral region setting unit 103 as a region that does not include the inner region 212. The peripheral region 222 is then set as a region that does not include a lesion site.

[0033] In this manner, the surrounding region setting unit 103 sets, in the medical image 201 , a surrounding region 222 that is a region adjacent to the inner region 212 and does not include the inner region 212 .

[0034] After completing step S123 in FIG. 6, the computer 100 returns the process to step S130 in FIG.

[0035] (Tissue decomposition processing) Next, the tissue decomposition process in step S130 of FIG. 4 will be described with reference to FIG. 8 and FIGS. 9A to 9D. Fig. 8 is a flowchart showing a detailed procedure of the tissue decomposition process for generating a tissue image 401 (see Figs. 9B to 9E) from the medical image 201 in step S130 in Fig. 4. Fig. 9A shows an example of the medical image 201, and Figs. 9B to 9E show an example of the tissue image 401. First, the tissue image processing unit 104 loads an anatomical tissue model (not shown) stored in the storage device 120 (Load anatomical component model: S131 in FIG. 8). Next, the tissue image processor 104 calculates a probability tissue map related to tissue information of each pixel according to the read anatomical tissue model (Compute probability component model: S132 in FIG. 8). As a result, the feature amount of the pixel for each tissue is calculated. Then, the tissue image processor 104 refers to the probability tissue map for each pixel of the medical image 201, and determines which tissue the pixel currently being processed most likely belongs to (predetermined criterion), thereby generating a tissue image 401. The tissue image processor 104 registers the generated tissue image 401 in the storage device 120 shown in Fig. 1 (Register component image: S133 in Fig. 8). As a result, the medical image 201 is decomposed into tissue images 401 for each tissue (gray matter, white matter, etc.).

[0036] It should be noted that the generation of the tissue image 401 (decomposition of the medical image 201) is not limited to the procedure of steps S131 to S133.

[0037] (Organization Image 401) FIG. 9A shows a medical image 201, FIG. 9B shows a tissue image 401a (401), FIG. 9C shows a tissue image 401b (401), FIG. 9D shows a tissue image 401c (401), and FIG. 9E shows a tissue image 401d (401). The tissue image 401a shown in FIG. 9B is an image showing, for example, a tissue of gray matter, and the tissue image 401b shown in FIG. 9C is an image showing a tissue of white matter. The tissue image 401c shown in FIG. 9D is an image showing cerebrospinal fluid, and the tissue image 401d shown in FIG. 9E is an image showing soft tissue. Incidentally, when the tissue images 401a to 401d are synthesized, the medical image 201 shown in FIG. 9A is obtained. Thus, in this embodiment, the medical image 201 is a medical image 201 captured by the medical image capturing device M, and is an image related to the brain.

[0038] 4. The inner region 212 and the peripheral region 222 set in steps S110 and S120 in FIG. 4 are also shown in the medical image 201 shown in FIG. 9A. The inner region 212 and the peripheral region 222 are also set in the tissue images 401a to 401d shown in FIG. 9B to 9E. That is, the inner region 212 and the peripheral region 222 set in steps S110 and S120 in FIG. 4 are carried over to the tissue images 401a to 401d. In FIG. 9B to FIG. 9D, as indicated by reference numeral 241, the tissue exists so as to straddle the inner region 212 and the peripheral region 222. However, in FIG. 9E, the tissue indicated by reference numeral 241 exists only in the inner region 212.

[0039] In this way, the tissue image processing unit 104 generates a plurality of tissue images 401 that are decomposed based on the medical image 201 according to a predetermined criterion (which tissue it belongs to).

[0040] After completing step S133 in FIG. 8, the computer 100 returns the process to step S140 in FIG.

[0041] (Intensity value group probability distribution calculation process) Next, the calculation of the intensity value group probability distribution 500 in step S140 in Fig. 4 will be described in detail with reference to Fig. 10 and Fig. 11A to Fig. 11D, and Fig. 9B to Fig. 9E will be referenced as appropriate. Fig. 10 is a flowchart showing a detailed procedure of calculating the intensity value group probability distribution in step S140 in Fig. 4. Figs. 11A to 11D are diagrams showing an example of the intensity value group probability distribution 500 in the inner region 212 and the peripheral region 222 for each tissue image 401.

[0042] First, the intensity value group probability distribution calculation unit 105 counts the number of pixels having each pixel intensity (number of pixels) for the inner region 212 of the tissue image 401 being processed (Count number of pixels related to each intensity value group in inner region: S141). For example, the intensity value group probability distribution calculation unit 105 counts the number "n1" of pixels having pixel intensities "488" to "499" (referred to as intensity value group "500") in the inner region 212 of the tissue image 401a in Fig. 9B. The counting result corresponds to the number of pixels having a predetermined intensity value.

[0043] Then, the intensity value group probability distribution calculation unit 105 divides the number of pixels of the intensity value group by the total number of pixels (the number of pixels constituting the inner region 212) in the inner region 212. As a result, for example, the intensity value group probability distribution calculation unit 105 calculates the probability (referred to as the intensity value group probability) related to pixels of the intensity value group "500" in the inner region 212 of the tissue image 401a (Calculate intensity value group probability in inner region: S142 in FIG. 10).

[0044] The definition of the intensity value group will be described below. A group of a plurality of intensity values ​​is called an intensity value group. For example, the intensity values ​​grouped as "310-321", "322-333", ... "704-715", "716-727" are called an intensity value group. Note that although the bin width is "12" in Figs. 11A to 11D, the bin width is not limited to "12". Furthermore, the intensity value groups do not necessarily need to be aligned as shown by symbols 503a, 503b, 503c, and 503d in Figs. 11A to 11D, but may be aligned between tissues. Moreover, an intensity value group corresponding to "n to m" (n and m are natural numbers including 0) is referred to as an intensity value group called "m+1." For example, an intensity value group corresponding to intensity values ​​"320 to 321" is written as an intensity value group "322."

[0045] For example, for inner region 212 of tissue image 401a shown in Fig. 9B, assume that the total number of pixels included in this inner region 212 is "N1." Also assume that the number of pixels belonging to the intensity value group "500" in inner region 212 of tissue image 401a is "n1." In this case, the intensity value group probability of the intensity value group "500" in inner region 212 of tissue image 401a is "n1 / N1."

[0046] As a result, the intensity value group probabilities are shown as bins of the histogram 501a in the intensity value group "500".

[0047] The intensity value group probability distribution calculation unit 105 performs the processes of steps S141 and S142 for all intensity value groups. 11A, a histogram 501a for the inner region 212 of the tissue image 401a is calculated. In this manner, the intensity value group probability distribution calculation section 105 calculates the histogram 501a, which is information on the distribution of the number of pixels having a predetermined intensity value, for the inner region 212.

[0048] Next, the intensity value group probability distribution calculation section 105 also calculates the intensity value group probability of the peripheral region 222 for the tissue image 401 being processed. The intensity value group probability distribution calculation unit 105 counts the number of pixels related to each intensity value group in the peripheral region 222 of the tissue image 401 by the same procedure as in step S141 (Count number of pixels related to each intensity value group in peripheral region: S143 in FIG. 10).

[0049] Then, the intensity value group probability distribution calculation unit 105 calculates the probability of pixels having each intensity value group in the peripheral region 222 of the tissue image 401 (Calculate intensity value group probability in peripheral region: S144 in FIG. 10). At this time, the intensity value group probability distribution calculation unit 105 divides the number of pixels of each intensity value group by the total number of pixels in the peripheral region 222 (the number of pixels constituting the peripheral region 222).

[0050] For example, the intensity value group probability distribution calculation section 105 calculates the intensity value group probability of the peripheral region 222 for the tissue image 401a shown in FIG. 9B. Then, the intensity value group probability distribution calculation unit 105 counts the number of pixels for each intensity value group in the peripheral region 222 of the tissue image 401a by the same procedure as in step S141 (Calculate intensity value group probability in peripheral region: S143 in FIG. 10). The result of step S143 corresponds to the number of pixels having a predetermined intensity value.

[0051] Then, the intensity value group probability distribution calculation unit 105 calculates the probability of pixels having each intensity value group in the peripheral region 222 of the tissue image 401a (S144 in FIG. 10). At this time, the intensity value group probability distribution calculation unit 105 divides the number of pixels of each intensity value group by the total number of pixels in the peripheral region 222 (the number of pixels constituting the peripheral region 222). This causes a histogram 502a for the peripheral region 222 of the tissue image 401a to be calculated as shown in FIG 11. The histogram 502a corresponds to a second intensity value frequency distribution, which is information on the distribution of the number of pixels having a predetermined intensity value for the peripheral region 222.

[0052] The intensity value group probability distribution calculation unit 105 calculates the intensity value group probability distribution 500 for the other tissue images 401b to 401c in a similar procedure. As a result, histograms 501b, 501c, and 501d for the inner region 212 are generated for the tissue images 401b to 401d, respectively (see FIGS. 11B to 11D). In addition, histograms 502b and 502c for the peripheral region 222 are generated for the tissue images 401b to 401c (see FIGS. 11B and 11C).

[0053] The histograms 501a to 501d and 502a to 502c shown in FIGS. 11A to 11D are referred to as an intensity value group probability distribution 500.

[0054] Note that Fig. 11D shows the intensity value group probability distribution 500 for the tissue image 401d. As indicated by reference numeral 241 in Fig. 9E, in the tissue image 401d, no tissue (soft tissue) exists in the peripheral region 222. Therefore, in Fig. 11D, there is no histogram corresponding to the intensity value group probability of the peripheral region 22.

[0055] After completing step S144 in FIG. 10, the computer 100 returns the process to step S151 in FIG.

[0056] (Intensity value group probability distribution 500) 11A to 11D are diagrams showing an example of an intensity value group probability distribution 500. FIG. 9B to FIG. 9E are also referred to as appropriate. Figure 11A shows the probability distribution 500 of the intensity values ​​in tissue image 401a, and Figure 11B shows the probability distribution 500 of the intensity values ​​in tissue image 401b, Figure 11C shows the probability distribution 500 of the intensity values ​​in tissue image 401c, and Figure 11D shows the probability distribution 500 of the intensity values ​​in tissue image 401d. 11A to 11D, the horizontal axis represents the intensity value group, and the vertical axis represents the intensity value group probability. In addition, in FIG. 11A to 11D, histograms 501a to 501d represent the intensity value group probability in the inner region 212 of each of the tissue images 401a to 401d. Furthermore, histograms 502a to 502c represent the intensity value group probability in the peripheral region 222 of each of the tissue images 401a to 401c.

[0057] Histograms 501a-501d relating to the inner region 212 are collectively referred to as inner region histograms 501, and histograms 502a-502c relating to the peripheral region 222 are collectively referred to as peripheral region histograms 502. In other words, the inner region histogram 501 corresponds to a first intensity value frequency distribution which is information relating to the distribution of the number of pixels having a predetermined intensity value for the inner region 212. And the peripheral region histogram 502 corresponds to a second intensity value frequency distribution which is information relating to the distribution of the number of pixels having a predetermined intensity value for the second region. In this way, the intensity value group probability distribution calculation unit 105 calculates the inner region histogram 501 and the peripheral region histogram 502.

[0058] As described above, in the tissue image 401d shown in Fig. 9E, the tissue (soft tissue) indicated by reference symbol 241 does not exist in the peripheral region 222. Therefore, in the intensity value group probability distribution 500 shown in Fig. 11D, there is no histogram of the intensity value group probability in the peripheral region 222. Therefore, since the processing of steps S151 to S171 in Fig. 4 cannot be performed on the tissue image 401d, steps S151 to S171 are performed only on the tissue images 401a to 401c.

[0059] (Probability Difference Distribution 600) Figures 12A to 12C are diagrams showing a probability difference distribution 600. Figures 12A to 12C show a probability difference distribution 600 which is a result of the processing in step S151 in Figure 4. Figures 9B to 9E will be referred to as appropriate. 12A to 12C, the horizontal axis indicates the intensity group, and the vertical axis indicates the probability difference 601. Fig. 12A shows the probability difference distribution 600 for tissue image 401a, Fig. 12B shows the probability difference distribution 600 for tissue image 401b, and Fig. 12C shows the probability difference distribution 600 for tissue image 401c. As described above, tissue image 401d does not have tissue in the surrounding region 222, and as shown in Fig. 11D, there is no intensity value group probability for the surrounding region 222, so the process of step S151 is not performed.

[0060] In step S151 of FIG. 4, in the intensity value group probability distribution 500 shown in FIGS. 11A to 11C, the surrounding region histogram 502 is subtracted from the inner region histogram 501 of each intensity value group. For example, the probability difference distribution 600 shown in Fig. 12A is calculated by subtracting the histogram 502a from the histogram 501a shown in Fig. 11A. Note that the value obtained by subtracting the surrounding region histogram 502 from each of the inner region histograms 501 is referred to as the probability difference 601 (difference value), and the distribution of the probability differences 601 for all intensity value groups is referred to as the probability difference distribution 600.

[0061] Similarly, probability difference distribution 600 shown in Fig. 12B is calculated by subtracting histogram 502b from histogram 501b shown in Fig. 11B. Moreover, probability difference distribution 600 shown in Fig. 12C is calculated by calculating the difference values ​​between histograms 501c and 502c shown in Fig. 11C for each intensity value group (predetermined intensity value).

[0062] In the probability difference distribution 600 shown in FIGS. 12A to 12C, the bins indicate the probability differences 601 for each intensity value group.

[0063] In the probability difference distribution 600 shown in Figure 12C, the bins with positive values ​​are biased toward the lower intensity values. This is because Figure 12C shows that there are many dark pixels in the inner region 212 of the tissue image 401c. Therefore, in the tissue image 401c, the dark pixels may correspond to the lesion.

[0064] The reference numerals 603 to 607, 607a, 607b, 611, 611a, and 611b shown in FIGS. 12A and 12B will be described later.

[0065] (Detection range setting) Next, step S160 (setting of the detection range) in FIG. 4 will be described with reference to FIGS. 12A to 12C and 13. FIG. Fig. 13 is a flowchart showing a detailed procedure for setting the detection range in step S160 in Fig. 4. The process shown in Fig. 13 corresponds to a detection range setting algorithm. First, the detection range setting unit 107 detects the intensity interval 607 (Detect intensity interval: S161). The detection range setting unit 107 performs the process of step S162 by setting the range in the probability difference distribution 600 where the probability difference 601 (the bins in FIGS. 12A to 12C) is positive as the intensity interval 607. That is, the detection range setting unit 107 sets the interval in the probability difference distribution 600 where the probability difference 601 has a positive value as the intensity interval 607 (detection range). The range in the probability difference distribution 600 where the probability difference 601 is positive indicates that the intensity value group corresponding to this range is more prevalent in the inner region 212 than in the peripheral region 222. Next, the detection range setting unit 107 detects an intensity group having a local maximum value (Detect Intensity Group with Local Maxima of Probability Difference (Imax): S162). The local maximum value is the maximum value of the probability difference 601 in a certain intensity section 607.

[0066] Furthermore, as shown in FIG. 12B, the number of intensity intervals 607 for one probability difference distribution 600 is not limited to one.

[0067] Hereinafter, the intensity value group in which the probability difference 601 is maximum will be referred to as a maximum intensity value group 604 (Imax) (see FIGS. 12A to 12C).

[0068] Then, the detection range setting unit 107 detects a group of intensity values ​​related to the minimum value 605 (see FIG. 12A: Ilow) and the maximum value 606 (see FIG. 12A: Ihigh) of the intensity section 607 (Detect intensities related to lower end (Ilow) and upper end (Ihigh) of Imax: S163). The detection of the minimum value 605 and the maximum value 606 may be performed by linear extrapolation or linear interpolation of the group of intensity values, but is not limited to "linear".

[0069] Next, the detection range setting unit 107 calculates the length (ΔI) of the intensity interval 607 from the minimum value 605 (Ilow) to the maximum value 606 (Ihigh) (Calculate length of intensity interval (ΔI) from lower end (Ilow) to upper end (Ihigh): S164 in FIG. 13).

[0070] Multiple intensity intervals 607 may be detected depending on the shape of the probability difference distribution 600. In the example of Fig. 12B, two intensity intervals 607a and 607b are detected.

[0071] After step S164, the detection range setting unit 107 determines whether the number of detected intensity intervals 607 is more than one (Is more than 1 intensity interval detected?: S165 in FIG. 13). If only one strength section 607 has been detected (S165→no), the detection range setting unit 107 advances the process to step S200.

[0072] If more than one intensity interval 607 is detected (S165→yes), the detection range setting unit 107 selects the intensity interval 607 with the largest area of ​​the distribution (distribution area 611) (see FIG. 12B) formed by the bins corresponding to the intensity interval 607 (Select intensity interval with the greatest area under plot: S166 in FIG. 13). With reference to the example shown in FIG. 12B, the distribution area 611b corresponding to the intensity interval 607b is larger than the distribution area 611a corresponding to the intensity interval 607a. Therefore, in the example shown in FIG. 12B, the detection range setting unit 107 selects the intensity interval 607b in step S166 in FIG. 12. In step S166, if a plurality of intensity intervals 607 are detected, the detection range setting unit 107 selects the intensity interval 607 with the largest area of ​​the probability difference distribution 600 in the intensity interval 607.

[0073] Next, detection range setting section 107 determines the location of intensity interval 607 (Determine intensity interval location: S200 in FIG. 13). The location is a position in probability difference distribution 600. The processing of step S200 will be described later.

[0074] Then, the detection range setting unit 107 judges whether the intensity interval 607 is close to any of the edges (Is intensity interval near to intensity edge?: S167). The edges are the lower end 602 (see FIGS. 12A to 12C: I0) and the upper end 603 (see FIGS. 12A to 12C: Iend) of the probability difference distribution 600. Specifically, it is judged whether the minimum value 605 of the intensity interval 607 is within a predetermined range from the lower end 602 of the probability difference distribution 600, or whether the maximum value 606 of the intensity interval 607 is within a predetermined range from the upper end 603 of the probability difference distribution 600.

[0075] If the intensity section 607 is not close to an edge (S167→no), that is, if it is determined that the position of the intensity section 607 is not close to either edge, the detection range setting unit 107 returns the process to step S171 in Fig. 4. As a result, no detection range is set for the tissue image 401 in which it is determined that the intensity section 607 is not close to an edge. If the intensity interval 607 is close to any edge (S167→yes), the detection range setting unit 107 extends the intensity interval 607 to the nearest edge (Extend intensity interval to the nearest intensity edge: S300 in FIG. 13). The process of step S300 will be described later. Next, the detection range setting unit 107 adjusts the intensity interval 607 that constitutes the determined detection range (Adjust intensity interval: S400 in FIG. 13). In step S400, the intensity interval 607 is adjusted so that pixels in the surrounding area 222 are not detected as a lesion site as much as possible. The processing in step S400 will be described later. After completing step S400 in FIG. 13, the detection range setting unit 107 sets the adjusted intensity interval 607 as the detection range, and the computer 100 returns the process to step S171 in FIG.

[0076] 13, the detection range setting unit 107 applies a detection range setting algorithm to each of the multiple tissue images 401. Then, the detection range setting unit 107 sets a detection range for the intensity value of the medical image 201 for each tissue image 401 based on the probability difference.

[0077] (Determining the position of intensity section 607) Next, step S200 in FIG. 13 (determining the position of the intensity section 607) will be described with reference to FIGS. 12A to 12C and 14. FIG. 14 is a flowchart showing the detailed procedure of step S200 (determining the position of the intensity section 607) in FIG. First, detection range setting unit 107 calculates minimum value 605-lower end 602 (Ilow-I0) in probability difference distribution 600 (Calculate Ilow minus I0 (Lob): S201 in FIG. 14). The value calculated as a result of step S201 is defined as Lob. Then, detection range setting unit 107 calculates upper end 603-maximum value 606 (Iend-Ihigh) in probability difference distribution 600 (Calculate Iend minus Ihigh (Hib): S202 in FIG. 14). The value calculated as a result of step S202 is set as Hib. Next, the detection range setting unit 107 calculates the difference between the upper end 603 and the lower end 602 (Iend-I0) minus the length (ΔI) of the intensity section 607 (Calculate (Iend-I0) minus ΔI (d): S203). As a result of step S203, the calculated value is set as d.

[0078] Next, detection range setting section 107 determines four conditions in steps S204 to S207. First, the detection range setting unit 107 judges whether the first condition ("Is ratio between Lob and Hib is < 0.2?": S204 in FIG. 14). In this judgment, it is judged whether (minimum value 605 - lower end 602 (Lob)) / (upper end 603 - maximum value 606 (Hib)) is less than 0.2. Note that the threshold value "0.2" in step S204 may be any value less than 0.5.

[0079] If the first condition is satisfied (S204→yes), the detection range setting unit 107 determines that the intensity section 607 is close to the lower end 602 (I0) (Location is near I0 edge: S211), and returns the process to step S167 in FIG. If the first condition is not satisfied (S204→no), the detection range setting unit 107 judges the second condition ("Is ratio between Lob and d < 0.05?": S205). In this judgment, it is judged whether (minimum value 605-lower end 602 (Lob))) / ((upper end 603-lower end 602)-length of intensity section 607 ((Iend-I0)-ΔI=d)) is less than 0.05. Note that the threshold value in step S205 is not limited to 0.05.

[0080] If the second condition is satisfied (S205→yes), the detection range setting unit 107 determines that the intensity section 607 is close to the lower end 602 (I0) (Location is near I0 edge: S211), and returns the process to step S167 in FIG.

[0081] If the second condition is not satisfied (S205→no), the detection range setting unit 107 judges the third condition ("Is ratio between Hib and Lob < 0.2?": S206). In step S206, it is judged whether (upper end 603-maximum value 606 (Hib)) / (minimum value 605-lower end 602 (Lob)) is less than 0.2. Note that the threshold value "0.2" in step S206 may be any value less than 0.5.

[0082] If the third condition is satisfied (S206→yes), the detection range setting unit 107 determines that the intensity section 607 is near the upper end 603 (Iend) (Location is near Iend edge: S212 in FIG. 14), and returns the process to step S167 of step S13. If the third condition is not satisfied (S206→no), the detection range setting unit 107 judges the fourth condition (whether or not the ratio of Hib to d is <0.05) (Is ratio between Hib and d < 0.05?: S207). In step S206, it is judged whether or not (upper end 603-maximum value 606(Hib)) / ((upper end 603-lower end 602)-length of intensity section 607((Iend-I0)-ΔI=d)). Note that the threshold in step S105 is not limited to 0.05.

[0083] If the fourth condition is satisfied (S207→yes), the detection range setting unit 107 determines that the intensity section 607 is close to the upper end 603 (Iend) (Location is near Iend edge: S212), and returns the process to step S167 in FIG. If the fourth condition is not satisfied (S207→no), the detection range setting unit 107 determines that the strength section 607 is not close to either the upper end 603 or the lower end 602 as shown in Fig. 12A (Location is far to any edges: S221). After that, the detection range setting unit 107 returns the process to step S167 in Fig. 13.

[0084] In this way, the first and second conditions are conditions for determining whether the intensity interval 607 is close to the lower end 602 (I0), and the third and fourth conditions are conditions for determining whether the intensity interval 607 is close to the upper end 603 (Iend).

[0085] For example, the intensity section 607 shown in FIG. 12A is determined to be far from both the lower end 602 (I0) and the upper end 603 (Iend) (S221 in FIG. 14). On the other hand, the intensity section 607b (intensity value group "975" to "1300") shown in FIG. 12B is determined to be close to the upper end 603 (Iend) (S212 in FIG. 14). And the intensity section 607 (intensity value group "4" to "260") shown in FIG. 12C is determined to be near the lower end 602 (I0) (S211 in FIG. 4).

[0086] As shown in Fig. 12A, the presence of intensity interval 607 in the middle of probability difference distribution 600, that is, intensity interval 607 is far from both upper end 603 and lower end 602, indicates the following. In this case, as shown in intensity value group probability distribution 500 in Fig. 11A, intensity interval 607 is neither a bright region nor a dark region, and therefore cannot be characterized as a lesion site. In other words, a lesion site is generally detected as a location with a large intensity value (bright) or a location with a small intensity value (dark). Therefore, tissue image 401 in which intensity interval 607 is in the middle of probability difference distribution 600 as shown in Fig. 12A is excluded from the detection range setting.

[0087] (Strength range expansion) Fig. 15 is a flowchart showing a detailed procedure for expanding the strength section 607 in step S300 in Fig. 13. Figs. 12A to 12C will be referred to as appropriate. First, in the process shown in FIG. 13, the detection range setting unit 107 determines whether the intensity section 607 being processed is determined to be close to the lower end 602 (I0) or the upper end 603 (Iend) (S167 in FIG. 13). If the intensity section 607 being processed is far from both the lower end 602 (I0) and the upper end 603 (Iend) (S167→no in FIG. 13), the detection range setting unit 107 returns the process to step S171 in FIG. 4 and performs processing on the next tissue image 401. In this way, the intensity section 607 that is far from both the lower end 602 (I0) and the maximum value (Iend) is not subjected to subsequent processing.

[0088] The detection range setting unit 107 determines whether the position of the intensity interval 607 is close to the lower end 602 (I0) (Is intensity interval near to I0 edge?: S301 in FIG. 15). When the intensity interval 607 is close to the lower end 602 (I0) (S301→yes), the detection range setting unit 107 modifies the minimum value 605 (Ilow) to the lower end 602 (I0) (Modify Ilow to I0: S302 in FIG. 15). In other words, the detection range setting unit 107 extends the intensity interval 607 to the lower end 602 (I0). In this way, when the intensity interval 607 is close to the lower end 602 in the probability difference distribution 600, the detection range setting unit 107 extends the minimum value 605 of the intensity interval 607 to the lower end 602.

[0089] If the intensity section 607 is close to the upper end 603 (Iend) (S301→no), the detection range setting unit 107 modifies the maximum value 606 (Ihigh) to the upper end 603 (Iend) (Modify Ihigh to Iend: S303 in FIG. 15). In other words, the detection range setting unit 107 extends the intensity section 607 to the upper end 603 (Iend). In this way, After steps S302 and S303 in Fig. 15 are completed, detection range setting unit 107 returns the process to step S400 in Fig. 13. If intensity interval 607 is close to upper end 603 in probability difference distribution 600, detection range setting unit 107 extends maximum value 606 of intensity interval 607 to upper end 603.

[0090] In general, the intensity interval 607 detected in steps S162 to S164 often does not include the lower end 602 or the upper end 603. As described above, a lesion is indicated by a bright pixel or a dark pixel. Therefore, if bright pixels are to be indicated as lesions, all pixels having an intensity equal to or greater than the maximum value 606 (Ihigh) are indicated as lesions in step S303. Similarly, if dark pixels are to be indicated as lesions, all pixels having an intensity equal to or less than the minimum value 605 (Ilow) are indicated as lesions in step S302.

[0091] (Strength section adjustment process) Fig. 16 is a flowchart showing a detailed procedure of the strength section adjustment in step S400 in Fig. 13. Figs. 18, 20, 21, and 22 will be referred to as appropriate. In order to explain the procedure shown in FIG. 16, the explanation will be given with reference to the intensity section 607b (intensity value group "975" to "1300") in the tissue image 401b shown in FIG. 12B. First, the detection range setting unit 107 executes the forward method, which is the first method (Compute forward method: S410 in FIG. 16). The process of step S410 will be described later. In step S410, an association strength 704, which is a selected strength value, and a selection cumulative probability 705, which is a first selection cumulative strength value probability, are calculated. The association strength 704 and the selection cumulative probability 705 will be described later.

[0092] Next, the detection range setting unit 107 obtains an associated strength 704 and a selected cumulative probability 705 corresponding to the associated strength 704 from the result of the forward method in step S410 (Obtain associated intensity and selected cumulative probability difference from forward method: S421 in FIG. 16). The definitions of the associated strength 704 and the selected cumulative probability 705 will be described later with reference to FIG. 18. In the example of FIG. 18, the associated strength 704 and the selected cumulative probability 705 are "1010" and "0.007", respectively.

[0093] Next, the detection range setting unit 107 executes the backward method, which is the second method (Compute backward method: S430 in FIG. 16). The process of step S430 will be described later. In step S430, an association strength 804, which is the selected strength value, and a selection cumulative probability 805, which is the second selection cumulative strength value probability, are calculated. The association strength 804 and the selection cumulative probability 805 will be described later. The detection range setting unit 107 obtains an associated strength 804 and a selected cumulative probability 805 by the backward method (Obtain associated intensity and selected cumulative probability difference from backward method: S441 in FIG. 16). In the example shown in FIG. 20, the associated strength 804 and the selected cumulative probability 805 are "990" and "0.141", respectively.

[0094] Thereafter, the detection range setting unit 107 selects the forward method or the backward method that has the larger selected cumulative probability 705, 805 (Select method with the greatest selected cumulative probability: S451 in FIG. 16). That is, the detection range setting unit 107 compares the value of the selected cumulative probability 705 in FIG. 18 with the selected cumulative probability 805 in FIG. 20 to determine which is larger.

[0095] Then, the detection range setting unit 107 determines whether or not the forward method has been selected (Is forward method selected?: S461). When the forward method is selected (S461→yes in FIG. 16), the detection range setting unit 107 modifies the maximum value 606 (Ihigh) to the associated intensity 704 (see FIG. 18) acquired by the forward method (Modify Ihigh to associated intensity of forward method: S471 in FIG. 16). When the backward method is selected (S461→no in FIG. 16), the detection range setting unit 107 modifies the minimum value 605 (Ilow) to the associated intensity 804 obtained by the backward method (FIG. 18) (Modify Ilow to associated intensity of backward method: S481 in FIG. 16).

[0096] Detection range setting unit 107 outputs strength interval 607 corrected in step S471 or step S481 as the detection range.

[0097] 18 and 20, the selection cumulative probability 705 by the forward method is "0.007" (see FIG. 18), and the selection cumulative probability 805 by the backward method is "0.141" (see FIG. 20). Therefore, the backward method is selected in step S461 in FIG. 16. Then, the strength value group "975" which is the minimum value 605 (Ilow) in FIG. 12B is corrected to "990" (association strength 804 by the backward method (see FIG. 18)).

[0098] 4. Then, in step S481, detection range setting unit 107 sets strength section 607 (that is, detection range) with the group of strength values ​​"990" to "1300." After completing steps S471 and S481 in FIG. 16, detection range setting unit 107 returns the process to step S171 in FIG.

[0099] (Forward method) Fig. 17 is a flowchart showing the detailed procedure of the forward method in step S410 in Fig. 16. Fig. 11B and Fig. 12B are referenced as appropriate. First, the detection range setting unit 107 calculates the cumulative intensity value group probability 701 (first cumulative intensity value probability: see FIG. 18) of the peripheral region 222 in the intensity interval 607 from the set minimum value 605 (Ilow) to the maximum value 606 (Ihigh) (Compute cumulative intensity value group probability of peripheral region in intensity interval (from Ilow to Ihigh): S411 in FIG. 17). In step S411, the intensity value group probability of the peripheral region histogram 502 (FIGS. 11A to 11C) is accumulated from the minimum value 605 (Ilow) to the maximum value 606 (Ihigh). The set minimum value 605 (Ilow) to maximum value 606 (Ihigh) are the minimum value 605 and maximum value 606 set in the process shown in FIG. 15.

[0100] In the example shown in FIG. 11B, the surrounding area histogram 502 (histogram 502b) is accumulated for each bin from the group of intensity values ​​“975” of the minimum value 605 (Ilow) in the intensity range 607 to the group of intensity values ​​“1300” of the maximum value 606 (Ihigh).

[0101] In this way, in step S411, the detection range setting unit 107 calculates the cumulative intensity value group probability 701 for the surrounding area 222 by accumulating the surrounding area histogram 502 from the minimum value 605 side of the intensity section 607 toward the maximum value 606 side of the intensity section 607.

[0102] Returning to the explanation of Figure 17. After step S411, the detection range setting unit 107 calculates the cumulative probability difference 702 (second cumulative intensity value probability: see FIG. 18) from the minimum value 605 (Ilow) side to the maximum value 606 (Ihigh) (Compute cumulative probability difference in intensity interval (from Ilow to Ihigh): S412 in FIG. 17). For example, in the example of the probability difference distribution 600 shown in FIG. 12B, the probability difference 601 for the tissue image 401b is accumulated for each bin from the minimum value 605 (Ilow: intensity value group "975") to the maximum value 606 (Ihigh: intensity value group "1300") of the intensity interval 607b.

[0103] In step S412, the detection range setting unit 107 calculates an accumulated probability difference 702 by accumulating the probability difference 601 from the minimum value 605 side of the strength section 607 toward the maximum value 606 side of the strength section 607.

[0104] Returning to the explanation of Figure 17. The detection range setting unit 107 sets a predetermined threshold value "0.05" (reference numeral 703 in FIG. 18) for making the number of pixels detected in the surrounding area 222 as small as possible. Then, the detection range setting unit 107 selects an associated intensity 704 as an intensity value based on a predetermined threshold (reference number 703 in FIG. 18) and the cumulative intensity value group probability 701. Then, the detection range setting unit 107 determines the cumulative intensity value group probability 701 corresponding to the selected associated intensity 704 as a selected cumulative probability 705, which is a first selected cumulative intensity value group probability. After that, the detection range setting unit 107 outputs the associated intensity 704 and the selected cumulative probability 705 (Output associated intensity and selected cumulative probability: S413 in FIG. 17). After step S413 in FIG. 17 is completed, the detection range setting unit 107 returns the process to step S421 in FIG.

[0105] (Results of the forward method) Fig. 18 is a diagram showing an example of a cumulative intensity value group probability 701 and a cumulative probability difference 702 for the surrounding area 222 (see Fig. 9C) obtained using the forward method. The results shown in Fig. 18 are based on the results shown in Figs. 11B and 12B. In the graph shown in FIG. 18, the horizontal axis indicates the intensity value group. The vertical axis indicates the cumulative intensity value group probability 701. A threshold value (reference number 703) of "0.05" (reference number 703) is set for the cumulative intensity value group probability 701 of the surrounding area 222. An associated intensity 704 (for example, an intensity value group "1010"), which is an intensity value group associated with the forward method, is defined by an intensity value group (horizontal axis) at which the threshold value (reference number 703) of "0.05" and the cumulative intensity value group probability 701 of the surrounding area 222 intersect. That is, the detection range setting unit 107 selects the associated intensity 704 based on a predetermined threshold value (reference number 703 in FIG. 18) and the cumulative intensity value group probability 701.

[0106] A calculation of the intensity value group at which the threshold value of "0.05" (symbol 703) and the cumulative intensity value group probability 701 of the surrounding area 222 intersect may be performed by linear extrapolation or linear interpolation, but is not limited to "linear". Furthermore, a selected cumulative probability 705 corresponding to an association strength 704 in the forward method is defined by the value of a cumulative probability difference 702 corresponding to the association strength 704 ("0.007" in the example of FIG. 18). In this way, the detection range setting unit 107 determines the cumulative probability difference 702 corresponding to the selected association strength 704 as the selected cumulative probability 705.

[0107] The detection range setting unit 107 outputs a selection cumulative probability 705 and an association strength 704 .

[0108] (Backward Method) Fig. 19 is a flowchart showing the detailed procedure of the backward method in step S430 in Fig. 16. Figs. 11B, 12B and 20 are referenced as appropriate. First, the detection range setting unit 107 calculates the cumulative intensity value group probability 801 (third cumulative intensity value probability: see FIG. 20) of the peripheral region 222 in the intensity interval 607 from the maximum value 606 (Ihigh) to the minimum value 605 (Ilow) (Compute cumulative intensity value group probability of peripheral region in intensity interval (from Ihigh to Ilow): S431 in FIG. 19). In step S431, the intensity value group probability of the peripheral region histogram 502 (FIGS. 11A to 11C) is accumulated from the maximum value 606 (Ihigh) to the minimum value 605 (Ilow). The set maximum value 606 (Ihigh) to minimum value 605 (Ilow) are the maximum value 606 and minimum value 605 set in the process shown in FIG. 15.

[0109] In the example shown in FIG. 11B, the surrounding area histogram 502 (histogram 502b) is accumulated for each bin from the maximum value 606 (Ihigh: intensity value group “1300”) to the minimum value 605 (Ilow: intensity value group “975”) in the intensity range 607.

[0110] In this way, in step S431, the detection range setting unit 107 calculates the cumulative intensity value group probability 801 for the surrounding region 222 by accumulating the surrounding region histogram 502 from the maximum value 606 side of the intensity section 607 toward the minimum value 605 side of the intensity section 607.

[0111] Next, the detection range setting unit 107 calculates a cumulative probability difference 802 (fourth cumulative intensity value probability: see FIG. 20) from the maximum value 606 (Ihigh) side toward the minimum value 605 (Ilow) (Compute cumulative probability difference in intensity interval (from Ihigh to Ilow): S432 in FIG. 19). For example, in the example of the probability difference distribution 600 shown in FIG. 12B, the probability difference 601 for the tissue image 401b is accumulated for each bin from the maximum value 606 (Ihigh: intensity value group "1300") of the intensity interval 607b toward the minimum value 605 (Ilow: intensity value group "975").

[0112] In step S432, the detection range setting unit 107 calculates an accumulated probability difference 802 by accumulating the probability difference 601 from the maximum value 606 side of the strength section 607 toward the minimum value 605 side of the strength section 607.

[0113] Then, the detection range setting unit 107 sets a predetermined threshold value of "0.05" (reference symbol 803) for the cumulative intensity value group probability 801 (see FIG. 20) of the surrounding region 222 to make the portion of the surrounding region 222 within the intensity interval 607 as small as possible. Next, the detection range setting unit 107 selects an associated strength 804 (strength value) based on a predetermined threshold (reference number 803 in FIG. 20) and the cumulative strength value group probability 801. After that, the detection range setting unit 107 determines the cumulative probability difference 802 corresponding to the selected associated strength 804 as a selected cumulative probability 805, which is a second selected cumulative strength value probability. Then, the detection range setting unit 107 outputs the associated strength 804 and the selected cumulative probability 805 (Output associated intensity and selected cumulative probability: S433 in FIG. 19). After step S433 in FIG. 19 is completed, the detection range setting unit 107 returns the process to step S441 in FIG.

[0114] (Results of backward method) FIG. 20 is a diagram showing an example of a cumulative intensity value group probability 801 and a cumulative probability difference 802 of the surrounding area 222 obtained using the backward method. In the graph shown in FIG. 20, the horizontal axis indicates the intensity value group. The vertical axis indicates the cumulative intensity value group probability 801. A threshold of "0.05" (reference number 803) is set for the cumulative intensity value group probability 801 of the surrounding area 222. An association intensity 804 ("990" in the example of FIG. 20) using the backward method is defined by the intensity value group (horizontal axis) where the limit of "0.05" (reference number 803) and the cumulative intensity value group probability 801 of the surrounding area 222 intersect. That is, the detection range setting unit 107 selects the association intensity 804 based on a predetermined threshold (reference number 803 in FIG. 20) and the cumulative intensity value group probability 801.

[0115] A calculation of the intensity value group (horizontal axis) at which the threshold value of "0.05" (symbol 803) and the cumulative intensity value group probability 801 of the surrounding area 222 intersect can be performed by linear extrapolation or linear interpolation, but is not limited to "linear". Furthermore, a selected cumulative probability 805 (e.g., "0.141") corresponding to the association strength 804 using the backward method is defined by the value of the cumulative probability difference 802 corresponding to the association strength 804. In this way, the detection range setting unit 107 determines the cumulative probability difference 802 corresponding to the selected association strength 804 as the selected cumulative probability 805.

[0116] The detection range setting unit 107 outputs a selection cumulative probability 805 and an association strength 804 .

[0117] 16, when the forward method is selected in step S461, the detection range setting unit 107 corrects the maximum value 606 (Ihigh) to the association strength 704 (see FIG. 18) by the forward method (S471 in FIG. 16). In other words, when the selected cumulative probability 705 is greater than the selected cumulative probability 805, the detection range setting unit 107 sets the maximum value 606 of the strength interval 607 as the association strength 704 selected by the forward method.

[0118] 18, pixels having the corresponding intensity value are detected in the surrounding region 222 with a probability of "0.05". By performing the process shown in step S471, the probability of false positive detection of a lesion site in the surrounding region 222 can be suppressed to less than "0.05".

[0119] 16, when the backward method is selected in step S461, the detection range setting unit 107 corrects the minimum value 605 (Ilow) to the association strength 804 (see FIG. 18) by the backward method (S481 in FIG. 16). In other words, when the selected cumulative probability 805 is greater than the selected cumulative probability 705, the detection range setting unit 107 sets the minimum value 605 of the strength interval 607 as the association strength 804 selected by the backward method.

[0120] In the group of intensity values ​​larger than the selection cumulative probability 805 in Fig. 20, pixels having the corresponding intensity value are detected in the surrounding region 222 with a probability of "0.05". By performing the process shown in step S481, it is possible to suppress the probability of false detection of a lesion site in the surrounding region 222 to less than "0.05". However, as described above, in the examples shown in Figs. 18 and 20, the backward method is selected.

[0121] In this manner, detection range setting section 107 corrects the detection range in step S471 or step S481 in FIG.

[0122] When the intensity value group probability or the probability difference based on the tissue image 401b shown in Fig. 11B or Fig. 12B is used, the backward method is selected in step S461. Therefore, the detection range setting unit 107 corrects the minimum value 605 (Ilow) to the associated intensity 804 (see Fig. 18) by the backward method (S481 in Fig. 16).

[0123] (Results of the forward method and the backward method (part 2)) 21 and 22 are diagrams showing the results of the forward method and the backward method performed using the intensity section 607 of FIG. 12C for the intensity value group probability distribution 500 (see FIG. 11C) and the probability difference distribution 600 (FIG. 12C) by the tissue image 401c shown in FIG. 9D. FIG. 21 shows the results of the forward method, and FIG. 22 shows the results of the backward method. In other words, the results shown in FIG. 21 and FIG. 22 show an example in which the forward method and the backward method are applied to the intensity section 607 of FIG. 12C (intensity value group "4" to "260"). Note that in FIG. 21 and FIG. 22, the horizontal and vertical axes are the same as those in FIG. 18 and FIG. 20.

[0124] 21 shows a cumulative intensity value group probability 701 and a cumulative probability difference 702 of the surrounding region 222 (see FIG. 9D). A threshold value of "0.05" (reference number 703 in FIG. 21) is set for the cumulative intensity value group probability 701 of the surrounding region 222 in order to minimize the number of pixels that are erroneously detected in the surrounding region 222. An association strength 704 using the forward method is defined by the intensity value (horizontal axis) at which the threshold value of "0.05" (reference number 703) and the cumulative intensity value group probability 701 of the surrounding region 222 intersect.

[0125] A method for calculating the intensity value at which the threshold value of "0.05" (symbol 703) intersects with the cumulative intensity value group probability 701 of the surrounding area 222 may be linear extrapolation or linear interpolation, but is not limited to the "linear" method. Furthermore, a selection cumulative probability 705 corresponding to an association strength 704 in the forward method is defined by the value of a cumulative probability difference 702 corresponding to the association strength 704.

[0126] 22 also shows a cumulative intensity value group probability 801 and a cumulative probability difference 802 of the surrounding region 222. A threshold value of "0.05" (reference symbol 803) is set for the cumulative intensity value group probability 801 of the surrounding region 222 in order to minimize the number of pixels that are erroneously detected in the surrounding region 222. An association strength 804 using the backward method is defined by an intensity value group (horizontal axis) at which the threshold value of "0.05" (reference symbol 803) and the cumulative intensity value group probability 801 of the surrounding region 222 intersect.

[0127] A method for calculating the intensity value at which the threshold value (symbol 803) of "0.05" intersects with the cumulative intensity value group probability 801 of the surrounding area 222 may be linear extrapolation or linear interpolation, but is not limited to the "linear" method. Furthermore, a selection cumulative probability 805 corresponding to an association intensity 804 by the backward method is defined by the value of a cumulative probability difference 802 corresponding to the association intensity 804.

[0128] 21 and 22, the selection cumulative probability 705 by the forward method is 0.059, and the selection cumulative probability 805 by the backward method is 0.006. Therefore, in step S461 in FIG. 16, the forward method is selected.

[0129] Furthermore, in step S471 in Fig. 16, the value of the intensity value group "260" in the maximum value 606 (Ihigh) in the intensity section 607 in Fig. 12C is corrected to "55", which is the value of the associated intensity 704 in the forward method in Fig. 21. Therefore, the intensity section 607 (detection range) of the tissue image 401c is set to "4" to "55".

[0130] In this way, by using the selection cumulative probabilities 705 and 805, erroneous detection of a lesion site outside the inner region 212 (the peripheral region 222) can be suppressed.

[0131] (mask) 23 is a diagram showing a mask for the lesion highlighting process generated in accordance with the set detection range, as a result of the process performed in step S171 in FIG. 23 shows a tissue image 401B in which a mask has been applied to the tissue image 401b. In the tissue image 401B, pixels having intensity values ​​corresponding to the intensity interval 607 (detection range) set in step S160 of FIG. 4 are highlighted as a mask region 1002 (selected pixels), which is a predetermined region. Specifically, the mask setting unit 108 selects pixels to be highlighted for each of the multiple tissue images 401 based on the intensity interval 607 (i.e., the detection range) set for each tissue image 401. Then, the mask setting unit 108 highlights the selected pixels as the mask region 1002.

[0132] (Mask Compositing) Next, the mask synthesis process performed in step S181 of FIG. 4 will be described with reference to FIGS. 24A to 24E. FIG. 24A is a third image for tissue image 401a shown in FIG. 9B, and is a diagram showing masked tissue image 401A. FIG. 24B is a third image for tissue image 401b shown in FIG. 9C, and is a diagram showing masked tissue image 401B. FIG. 24C is a third image for tissue image 401c shown in FIG. 9D, and is a diagram showing masked tissue image 401C. FIG. 24D is a third image for tissue image 401d shown in FIG. 9E, and is a diagram showing masked tissue image 401D. And the mask image 1003 shown in FIG. 24E is a composite of tissue images 401A to 401D shown in FIG. 24A to FIG. 24D.

[0133] Mask setting unit 108 highlights pixels according to the detection range set by detection range setting unit 107, thereby generating tissue images 401A to 401D shown in FIGS. 24A to 24D. Furthermore, tissue images 401B and 402C display mask regions 1002 in which pixels having intensity values ​​corresponding to the detection range are highlighted. However, in tissue image 401d shown in FIG. 9E, as described in FIG. 11D, there is no tissue (soft tissue) in peripheral region 222 in original tissue image 401d. Therefore, there is no peripheral region histogram 502 based on tissue image 401d, and therefore the detection range setting process is not performed. Therefore, in tissue image 401D shown in FIG. 24D, mask regions 1002 as shown in FIGS. 24B and 24C are not displayed. That is, tissue image 401D shown in FIG. 24D is the same as tissue image 401d shown in FIG. 9E.

[0134] In addition, in the tissue image 401a shown in FIG. 9B, the intensity interval 607 is located in the center of the probability difference distribution 600 as shown in FIG. 12A. That is, the intensity interval 607 is far from both the lower end 602 (I0) and the upper end 603 (Iend). Therefore, the process of setting the detection range is not performed (S160 in FIG. 4 is interrupted, and S171 is not performed). As a result, a mask such as that shown in FIG. 24B and FIG. 24C is not set in the tissue image 401A shown in FIG. 24A. Therefore, the tissue image 401A shown in FIG. 24A is the same image as the tissue image 401a shown in FIG. 9B.

[0135] Detection ranges are set for tissue image 401b shown in Fig. 9C and tissue image 401c shown in Fig. 9D. Therefore, as shown in Fig. 24B and Fig. 24C, pixels having intensity values ​​corresponding to the respective detection ranges are highlighted, so that mask setting unit 108 displays mask region 1002.

[0136] As shown in Figures 24A and 24D, a mask is not generated for all tissue images 401. As shown in the example of Figures 24B and 24C, a mask region 1002 indicating a highlighted lesion is generated only for tissue images 401b and 401c.

[0137] Then, in step S181 of Fig. 4, the mask editing unit 109 synthesizes the generated tissue images 401A to 401D, thereby synthesizing all the masks. As a result, a mask image 1003 (fourth image) having a single mask region 1002A is generated as shown in Fig. 24E. The pixels highlighted by the mask region 1002A correspond to the mask region 1002 in Fig. 24B and the mask region 1002 in Fig. 24C. In this manner, the mask editing unit 109 generates the mask image 1003 by synthesizing the pixels selected by the mask setting unit 108.

[0138] (Display screen 1100) 25 is a diagram showing an example of a display screen 1100 displayed on the display device 132. Note that FIG. 25 is displayed as a result of the display process in step S191 in FIG. As shown in FIG. 25, a display screen 1100 includes screens 1101 and 1102. A medical image 201 that is read in step S101 in FIG. 4 and displayed in step S102 in FIG. 4 is displayed on screen 1101. Incidentally, the medical image 201 displayed on screen 1101 is the same as the medical image 201 shown in FIG. 7. On screen 1101, a portion that is suspected to be a lesion site is highlighted as lesion area 202. Note that the portion highlighted in lesion area 202 is a portion that a doctor has determined to be a lesion site, as described above. As described above, a portion that is suspected to be a lesion site is a portion that is determined by a doctor. Also, color bar 1101a corresponds to the intensity value on screen 1101.

[0139] 4. A mask image 1003, which is the result of image processing (lesion site emphasis processing) shown in Fig. 4, is also displayed on the screen 1102. Note that a color bar 1102a corresponds to the luminance value on the screen 1102.

[0140] As shown in Fig. 25, a display screen 1100 displays a screen 1101 displaying the medical image 201 before the lesion area highlighting process, and a screen 1102 displaying a mask image 1003 after the lesion area highlighting process. In the mask image 1003 displayed on the screen 1102, the synthesized mask is highlighted as a mask area 1002A. In this manner, the display processing unit 110 displays (outputs) the tissue image 401, in which the pixels (mask area 1002) selected by the mask setting unit 108 are highlighted, as the mask image 1003 on the display device 132.

[0141] Lesion area 202 on screen 1101 roughly shows the area that the doctor has determined to be the lesion area. In contrast, on screen 1102, the accuracy of the lesion area display (mask area 1002A) is improved by lesion area highlighting processing. Furthermore, according to the lesion area highlighting processing of this embodiment, not only bright pixels but also dark pixels can be detected as the lesion area.

[0142] 25, the user can compare the medical image 201 and the mask image 1003 before and after the lesion area emphasis processing. After the user checks the result displayed on the screen 1102, the mask image 1003 displayed on the screen 1102 is saved as lesion data together with the lesion name, etc.

[0143] According to the first embodiment, by setting a criterion (detection range) for determining a lesion site for each tissue, it becomes possible to render the lesion site more accurately.

[0144] Furthermore, if only one threshold is set, and there are pixels outside the inner region 212 (peripheral region 222) that exceed the threshold, there is a possibility that pixels in the peripheral region 222 will also be highlighted as a lesion site. Furthermore, with one threshold, either bright pixels or dark pixels in the medical image 201 are highlighted as a lesion site. However, in reality, lesion sites in the medical image 201 often contain a mixture of bright pixels and dark pixels depending on the tissue.

[0145] In the first embodiment, a region (inner region 212) corresponding to the lesion site in the medical image 201 is specified by receiving a user's input. Then, a detection range of pixel intensity is determined using the occurrence frequency of the intensity of all pixels inside the inner region 212 and the occurrence frequency of the intensity of pixels in the peripheral region 222 (for example, 20% outside the inner region 212). At this time, in the first embodiment, the tissue is decomposed into tissue images 401 for each tissue, and a detection range is set for each tissue image 401. This makes it possible to highlight the lesion site with high accuracy. Furthermore, by combining tissue images 401A to 401D in which the lesion site is highlighted (masked), it becomes possible to display a mask image 1003 in which the lesion site is displayed with high accuracy. In this way, according to the first embodiment, the region of interest in the image can be highlighted with high accuracy.

[0146] Thus, in the first embodiment, in order to simplify these specialized viewpoints and complicated tasks, the area range (inner area 212) corresponding to the lesion site according to the user's estimation is received as an input. Then, computer 100 compares the intensity distribution of the selected area with that of the surrounding area 222, thereby highlighting the lesion site within the specified area.

[0147] 16 to 22 are performed, and the method that results in the larger selected cumulative probability 705, 805 is selected. Then, the detection range is corrected by the association strength 704, 804 calculated by the selected forward method or backward method. In this way, it is possible to prevent erroneous detection of a lesion site in the surrounding area 222.

[0148] [Second embodiment] Next, a second embodiment of the present invention will be described with reference to FIGS. 26 to 31E. In the first embodiment, the processing of a two-dimensional image (for example, a two-dimensional image extending in an x-axis-y-axis plane) has been described, but it can also be expanded to three dimensions using the same inner line 211 (see FIG. 7). That is, an inner line 211 is set for a plurality of two-dimensional medical images 201 adjacent in the z direction in FIG. 27A, and a mask image 1003 is generated based on the set inner line 211. Then, a three-dimensional image 250 (a fifth image) is constructed by stacking the mask images 1003. (Computer 100A) Fig. 26 is a diagram showing the configuration of a computer 100A in the second embodiment. In Fig. 26, the same components as those in Fig. 2 are given the same reference numerals and the description thereof will be omitted. In the second embodiment, a plurality of mask images 1003 exist. The computer 100A has a 3D image processing unit 110A that generates a 3D image 250 (see FIG. 27B), which is a fifth image, by stacking a plurality of mask images 1003 as shown in FIG. 27B.

[0149] (3D image 250) 27A and 27B are diagrams showing an outline of a method for generating a three-dimensional image 250. In FIG. In the first embodiment, the lesion site is detected for each tissue. However, detection using a procedure similar to the detection range setting (S160) in FIG.

[0150] The photographing is performed continuously along the z-axis direction of the medical image 201. In addition, the medical image 201 shows a lesion area 202 suspected of being a lesion site, an inner line 211 set by the user, and an outer line 221 extending the inner line 211. The inner region setting unit 102 projects the inner line 211 of the medical image 201 onto each of the medical image 201a adjacent to the medical image 201 in the z-axis direction and the medical image 201b adjacent to the medical image 201a in the z-direction. As a result, the inner lines 211a and 211b are set on the medical images 201a and 201b, respectively. Furthermore, the surrounding region setting unit 103 sets the outer line 221 on the medical image 201, and projects the outer line 221 set on the medical image 201 onto the medical images 201a and 201b adjacent to the medical image 201 in the Z-axis direction. As a result, outer lines 221a and 221b are set on the medical images 201a and 201b, respectively.

[0151] Then, as shown in Fig. 27B, the three-dimensional image processing unit 110A generates a three-dimensional image 250, which is a fifth image, by stacking a plurality of mask images 1003, 1003a, and 1003b (fourth images). The mask images 1003, 1003a, and 1003b are generated from the medical images 201, 201a, and 201b shown in Fig. 27A. Note that a mask region 1002A is displayed in the mask image 1003 shown in Fig. 27B. Also, each of the medical images 201, 201a, and 201b shown in Fig. 27A will be referred to as a slice image.

[0152] (Emphasis processing of lesion area on 3D image 250) Fig. 28 is a flowchart showing a lesion site highlighting process for the three-dimensional image 250. The process shown in Fig. 28 is an example of an extension procedure for the first embodiment by the computer 100. Fig. 27B will be referred to as appropriate. First, the computer 100 detects a lesion site at z0 where the set inner line 211 exists according to steps S110 to S191 in FIG. 4 (Detect lesion site at z0 of set inner region: S501 in FIG. 28). z0 is the position on the z axis where the reference medical image 201 exists. z0 is set by the user based on the size of the lesion site (mask) and the like. The three-dimensional image 250 is generated by connecting the mask images 1003 in the vertical direction with the mask image 1003 located at z0 as the reference. In particular, the mask image 1003 located at z0 (a certain fourth image) is appropriately referred to as the reference mask image.

[0153] Next, three-dimensional image processor 110A counts the first lesion site pixel count (V0), which is the first pixel count, from detected mask region 1002A (see FIG. 27B) (Count pixels with detected lesion site (V0): S502 in FIG. 28). The first lesion site pixel count (V0) is the pixel count of highlighted pixels (mask region 1002A), which are pixels that are highlighted in the fourth image.

[0154] Next, three-dimensional image processing unit 110A calculates the first lesion site center (C0) which is the first center (Compute lesion site center (C0): S503 in FIG. 28). The first lesion site center (C0) is calculated by using the average value of the x and y positions of the pixels corresponding to the lesion site in the reference mask image (i.e., in mask area 1002A). Specifically, the first lesion site center (C0) is the center of gravity of the pixels that three-dimensional image processing unit 110A finds to be in mask area 1002A. The first lesion site center (C0) is the center of mask area 1002A, as indicated by reference numeral 1203 in FIG. 31A. The first lesion site center (C0) is the center of the highlighted display pixel (mask area 1002A).

[0155] Next, the three-dimensional image processor 110A calculates the first lesion site spread (S0), which is the first spread (Compute lesion site spread (S0): S504 in FIG. 28). The first lesion site spread (S0) is the average of the distances between the first lesion site center (C0) calculated in step S503 and each of the pixels constituting the lesion site (mask region 1002A). In this way, the first lesion site spread (S0) is the average of the distances between the first lesion site center (C0) and each of the pixels constituting the highlighted display pixel (mask region 1002A).

[0156] In this manner, in steps S502 to S504, three-dimensional image processing unit 110A calculates the first lesion pixel count (V0), the first lesion center (C0), and the first lesion diffuseness (S0). Thereafter, three-dimensional image processor 110A detects a lesion site in the adjacent image for z0 downward (Detect lesion site for z0 downward: S510 in FIG. 28). The downward direction is the negative direction of the z axis with z0 as the reference point. After that, the three-dimensional image processor 110A detects a lesion site in the adjacent image for z0 upward (Detect lesion site for z0 upward: S520 in FIG. 28). The information is the positive direction of the z axis with z0 as the reference. The processes in steps S510 and S520 will be described later.

[0157] (Downward detection) Fig. 29 is a flowchart for explaining the detailed procedure of downward detection performed in step S510 in Fig. 28. Fig. 31 will be referred to as necessary. First, three-dimensional image processing unit 110A sets Δz to 1 (Set Δz to 1: S5101 in FIG. 29). Next, the computer 100 detects a lesion site in the adjacent slice image below z0 (i.e., the medical image 201 at z0-Δz) using the inner region 212 and the peripheral region 222 set in the medical image 201 to be processed according to steps S110 to S181 in FIG. 4 (Detect lesion site at z0-Δz: S5102 in FIG. 29). Specifically, a mask region 1002A is set in the medical image 201 to be processed, thereby generating a mask image 1003. Here, the generated mask image 1003 corresponds to the fourth image different from the fourth image in which the first lesion pixel count (V0), the first lesion center (C0), and the first lesion diffuseness (S0) are calculated.

[0158] Note that the distribution of tissues and the distribution of diseased areas differ for each medical image 201. Therefore, the probability difference 601 differs for each tissue. That is, the detection range may differ across z for each tissue. However, for example, by calculating the average of the minimum values ​​605 (see FIG. 12A) of the detection range or the average value of the maximum values ​​606 (see FIG. 12A), the detection range can be maintained at the same range across each mask image 1003.

[0159] Next, three-dimensional image processing unit 110A counts the second lesion site pixel number (V), which is the second pixel number, in mask area 1002A generated in step S5102 (Count pixels with detected lesion site (V): S5103 in FIG. 29). In step S5103, the same process as in step S502 in FIG. 28 is performed. The second lesion site pixel number (V) is the pixel number of highlighted pixels (mask area 1002A).

[0160] Next, three-dimensional image processing unit 110A determines whether the ratio between the second lesion site pixel number (V) calculated in step S5103 of FIG. 29 and the first lesion site pixel number (V0) calculated in step S502 of FIG. 28 is less than 0.05 (a first value, which is a predetermined value) (Is ratio between V and V0 < 0.05?: S5104 of FIG. 29). If the ratio of the second lesion pixel count (V) to the first lesion pixel count (V0) is less than 0.05 (S5104→yes), the detected lesion is small as shown in mask area 1002A in Fig. 31B and is therefore considered negligible. Therefore, no further detection in the downward direction is performed, and three-dimensional image processing unit 110A returns the process to step S520 in Fig. 28.

[0161] If the ratio of the second lesion pixel count (V) to the first lesion pixel count (V0) is 0.05 or more (S5104→no), the lesion is considered to remain prominent, as shown in mask region 1002A in FIG. 31C. Therefore, three-dimensional image processing unit 110A calculates the second lesion site center (C), which is the second center, for mask area 1002A generated in step S5102 (Compute lesion site center (C): S5105 in FIG. 29). This process is similar to the process performed in step S503 in FIG. 28. The second lesion site center (C) is the same as the first lesion site center (C0), and the average value of the x position and y position of each pixel corresponding to the lesion site, etc. is used. The second lesion site center (C) is the center of the highlighted display pixel (mask area 1002A).

[0162] Next, three-dimensional image processor 110A determines whether the distance between the second lesion center (C) and the first lesion center (C0) is greater than 0.5 × inner region diameter (Is distance between C and C0 > (0.5 × inner region diameter)?: S5106 in FIG. 29). The inner region diameter is the major axis of the ellipse that has the smallest area among the ellipses that include inner line 211 (see FIG. 7) therein. Note that, because inner region 212 is composed of not only circles but also polygons, the inner region diameter is approximated by the following formula (1).

[0163]

number

[0164] In formula (1), “inner area” is the area of ​​the inner region 212 .

[0165] For example, three-dimensional image processor 110A determines whether the distance between the first lesion center (1203:C0) in Fig. 31A and the second lesion center (1203:C) in Fig. 31C is greater than 0.5 × the inner region diameter (a second value that is a predetermined value). Specifically, three-dimensional image processor 110A calculates the distance between the location (1203 in Fig. 31C) where the first lesion center (1203:C0) in Fig. 31A is projected onto mask image 1003B in Fig. 31C, and the second lesion center (C) (1203b) in Fig. 31C.

[0166] When the distance between the first lesion center (C0) and the second lesion center (C) is greater than 0.5×the inner region diameter (S5106→yes), the following factors are considered. That is, in such a case, as shown in FIG. 31A and FIG. 31D, the lesion having the first lesion center (C0) (reference number 1203 in FIG. 31A) and the lesion having the second lesion center (C) (reference number 1203c in mask image 1003C in FIG. 31D) are considered to belong to unrelated lesions. When "yes" is determined in step S5105 in FIG. 29, no further processing is performed on the medical image 201 in the downward direction. Therefore, the three-dimensional image processing unit 110A proceeds to the processing in step S520 in FIG. 28. In this way, it is possible to prevent erroneous detection of an unrelated lesion with respect to the lesion shown in mask image 1003 located at z0.

[0167] When the distance between the first lesion site center (C0) and the second lesion site center (C) is 0.5 × inner region diameter or less (S5106 → no), the three-dimensional image processing unit 110A calculates the second lesion site spread (S), which is the second spread, for the mask image 1003 being processed (Compute lesion site spread (S): S5107 in FIG. 29). The second lesion site spread (S) is the average of the distance between the second lesion site center (C) calculated in step S5105 and each pixel constituting the lesion site (mask region 1002A) in the mask image 1003 being processed. The second lesion site spread (S) is the average of the distance between the second lesion site center (C) and each pixel constituting the highlighted display pixel (mask region 1002A).

[0168] In this manner, in steps S5103, S5105, and S5107, three-dimensional image processing unit 110A calculates the second lesion pixel count (V), the second lesion center (C), and the second lesion diffuseness (S), respectively.

[0169] Then, the three-dimensional image processor 110A judges whether the ratio between the second lesion site diffusivity (S) and the first lesion site diffusivity (S0) is greater than 1.5 (Is ratio between S and S0>1.5?: S5108 in FIG. 29). If the ratio of the second lesion site diffusion degree (S) to the first lesion site diffusion degree (S0) is greater than 1.5 (a third value that is a predetermined value) (S5108→yes), the mask image 1003 being processed is considered to have detected lesion sites scattered as shown in the mask image 1003D in FIG. 31E. In other words, it is considered to have a lot of noise. Therefore, processing downward from the mask image 1003 being currently processed is not performed. In other words, the three-dimensional image processing unit 110A advances the process to step S520 in FIG. 28. In this way, when the lesion sites are scattered, that is, when noise is observed, it is possible to prevent these lesion sites from being linked.

[0170] When the ratio between the second lesion site diffuseness (S) and the first lesion site diffuseness (S0) is 1.5 or less (S5108→no), three-dimensional image processing unit 110A continues the downward detection by adding 1 to Δz to modify Δz (Modify Δz to Δz+1: S5109 in FIG. 29). That is, three-dimensional image processing unit 110A continues to select mask image 1003 from among multiple mask images 1003 as mask image 1003 constituting three-dimensional image 250 until any of the following (A1) to (A3) becomes true.

[0171] (A1) The ratio of the second lesion site pixel number (V) to the first lesion site pixel number (V0) is less than a predetermined first value ("0.05" in this embodiment) (S5104). (A2) The distance between the second lesion center (C) and the first lesion center (C0) becomes greater than a second value, which is a predetermined value (0.5×the inner area in this embodiment) (S5106). (A3) The ratio of the second lesion site diffusivity (S) to the first lesion site diffusivity (S0) becomes greater than a third value, which is a predetermined value ("1.5" in this embodiment).

[0172] After correcting Δz, three-dimensional image processing unit 110A determines whether Z0-Δz≧1 is satisfied (Is z0-Δz≧1?: S5110 in FIG. 29). If Z0-Δz≧1 is satisfied (S5110→yes), three-dimensional image processing unit 110A returns the process to step S5102 and performs the processes from step S5102 on the next lower medical image 201. Note that z0 (reference medical image 201) is fixed throughout the process in Fig. 29. That is, the number of first lesion site pixels (V0), first lesion site center (C0), and first lesion site diffuseness (S0) used are fixed.

[0173] On the other hand, if Z0-Δz≧1 is not satisfied (S5110→no), the three-dimensional image processor 110A determines that the last processed slice image is the bottom slice image in the z direction. Therefore, no processing is performed below the medical image 201 currently being processed. Then, the three-dimensional image processor 110A advances the process to step S520 in FIG.

[0174] In step S5104, the threshold for the ratio between the second lesion pixel count (V) and the first lesion pixel count (V0) is 0.05, but is not limited to this value. In step S5106, the threshold for the distance between the second lesion center (C) and the first lesion center (C0) is 0.5 × inner region diameter, but is not limited to this value. In step S5108, the threshold for the second lesion diffuseness (S) and the first lesion diffuseness (S0) is 1.5, but is not limited to this value.

[0175] The first lesion pixel count (V0), the first lesion center (C0), and the first lesion diffuseness (S0) are calculated from mask image 1003 at the position of z0 initially specified by inner line 211. Although these values ​​are fixed throughout FIG. 29, the first lesion pixel count (V0), the first lesion center (C0), and the first lesion diffuseness (S0) may indicate the state at an adjacent z. In other words, each time processing moves to an adjacent medical image 201, the first lesion pixel count (V0), the first lesion center (C0), and the first lesion diffuseness (S0) may be recalculated based on mask image 1003 that was previously processed.

[0176] (Upward detection) Fig. 30 is a flowchart showing a detailed procedure of the upward detection performed in step S520 in Fig. 28. Fig. 31 will be referred to as appropriate. First, three-dimensional image processing unit 110A sets the Δz step to 1 (Set Δz to 1: S5201 in FIG. 30). Next, computer 100 detects a lesion site in an upper adjacent slice image of z0 (i.e., z0+Δz) using the same inner region 212 and peripheral region 222 according to steps S110 to S191 in FIG. 4 (Detect lesion site at z0 +Δz: S5202 in FIG. 30).

[0177] Note that the distribution of tissues and the distribution of diseased areas differ for each medical image 201. Therefore, the probability difference 601 differs for each tissue. That is, the detection range may differ across z for each tissue. However, for example, by calculating the average of the minimum values ​​605 (see FIG. 12A) of the detection range or the average value of the maximum values ​​606 (see FIG. 12A), the detection range can be maintained at the same range across each mask image 1003.

[0178] Next, three-dimensional image processing unit 110A counts the second number of pixels (V) of lesion site, which is the second number of pixels, from the mask generated in step S5202 (Count pixels with detected lesion site (V): S5203 in FIG. 30).

[0179] Next, three-dimensional image processing unit 110A determines whether the ratio between the second lesion site pixel number (V) calculated in step S5203 of FIG. 30 and the first lesion site pixel number (V0) calculated in step S502 of FIG. 28 is less than 0.05 (Is ratio between V and V0<0.05?: S5204 of FIG. 30). If the ratio of the second lesion pixel count (V) to the first lesion pixel count (V0) is less than 0.05 (S5204→yes), the detected lesion is small as shown in mask area 1002A in Fig. 31B and can be ignored. Therefore, no further upward detection is performed, and three-dimensional image processing unit 110A ends the processing in Fig. 30 and the processing in Fig. 28.

[0180] If the ratio of the second lesion pixel number (V) to the first lesion pixel number (V0) is 0.05 or more (S5204→no), the lesion is considered to remain prominent, as shown in mask region 1002A in FIG. 31C. Therefore, three-dimensional image processor 110A calculates a second lesion site center (C) which is the second center for the mask generated in step S5202 (Compute lesion site center (C): S5205 in FIG. 30). This process is similar to the process performed in step S503 in FIG. 28, and the average value of the x position and y position of each pixel corresponding to the lesion site, etc., is used for the second lesion site center (C). Next, three-dimensional image processor 110A determines whether the distance between the second lesion center (C) and the first lesion center (C0) is greater than 0.5 × inner region diameter (Is distance between C and C0>(0.5 × inner region diameter)?: S5206 in FIG. 30). The inner region diameter is defined as described above. Note that, since inner region 212 is composed of not only circles but also polygons, the inner region diameter is approximated by the following formula (2).

[0181]

number

[0182] In formula (2), "inner area" is the area of ​​the inner region.

[0183] For example, three-dimensional image processor 110A determines whether the distance between the first lesion center (1203:C0) in Fig. 31A and the second lesion center (1203:C) in Fig. 31B is greater than 0.5 × the inner region diameter. Specifically, three-dimensional image processor 110A calculates the distance between the location (1203 in Fig. 31C) where the first lesion center (1203:C0) in Fig. 31A is projected onto mask image 1003B in Fig. 31C, and the second lesion center (1203b:C) in Fig. 31C.

[0184] If the distance between the first lesion center (C0) and the second lesion center (C) is greater than 0.5×the inner region diameter (S5206→yes), the lesion having the first lesion center (C0) and the lesion having the second lesion center (C) are considered to belong to unrelated lesions. If "yes" is determined in step S5205 of FIG. 30, no further processing on the medical image 201 in the upward direction is performed. Therefore, the three-dimensional image processor 110A ends the processing of FIG. 30 and the processing of FIG. 28. In this way, it is possible to prevent erroneous detection of an unrelated lesion with respect to the lesion shown in the medical image 201 located at z0.

[0185] When the distance between the first lesion site center (C0) and the second lesion site center (C) is 0.5 × inner region diameter or less (S5206 → no), the three-dimensional image processor 110A calculates the second lesion site spread (S), which is the second spread, for the mask image 1003 being processed (Compute lesion site spread (S): S5207 in FIG. 30). The second lesion site spread (S) is the average of the distance between the second lesion site center (C) calculated in step S5205 and each pixel constituting the lesion site (mask) in the mask image 1003 being processed.

[0186] Then, the three-dimensional image processor 110A judges whether the ratio between the second lesion site diffusivity (S) and the first lesion site diffusivity (S0) is greater than 1.5 (Is ratio between S and S0>1.5?: S5208 in FIG. 30). When the ratio of the second lesion site diffusion degree (S) to the first lesion site diffusion degree (S0) is greater than 1.5 (S5208→yes), the mask image 1003 being processed is considered to have detected lesion sites scattered as shown in the mask image 1003D of FIG. 31E. In other words, it is considered to have a lot of noise. Therefore, the processing above the medical image 201 being currently processed is not executed. In other words, the three-dimensional image processing unit 110A ends the processing of FIG. 30 and the processing of FIG. 28. In this way, when the lesion sites are scattered, that is, when noise is observed, it is possible to prevent these lesion sites from being linked.

[0187] If the ratio of the second lesion site diffuseness (S) to the first lesion site diffuseness (S0) is 1.5 or less (S5208→no), three-dimensional image processor 110A adds 1 to Δz to modify Δz (Modify Δz to Δz+1: S5209 in FIG. 30).

[0188] After correcting Δz, three-dimensional image processing unit 110A determines whether or not Z0+Δz≦zmax is satisfied (Is z0+Δz≦zmax?: S5210 in FIG. 30). If Z0+Δz≦zmax1 is satisfied (S5210→yes), three-dimensional image processing unit 110A returns the process to step S5202 and performs the processes from step S5202 onward for the next higher medical image 201. Note that z0 (reference medical image 201) is fixed throughout the process in Fig. 30. That is, the number of first lesion site pixels (V0), first lesion site center (C0), and first lesion site diffuseness (S0) used are fixed.

[0189] If Z0+Δz≦zmax is not satisfied (S5210→no), the three-dimensional image processor 110A determines that the last processed slice image is the top slice image in the z direction. Therefore, no processing is performed above the medical image 201 currently being processed. Then, the three-dimensional image processor 110A ends the processing of FIG. 30 and the processing of FIG. 28.

[0190] In step S5204, the threshold for the ratio between the second lesion pixel count (V) and the first lesion pixel count (V0) is 0.05, but is not limited to this value. In step S5206, the threshold for the distance between the second lesion center (C) and the first lesion center (C0) is 0.5 × inner region diameter, but is not limited to this value. In step S5208, the threshold for the second lesion diffuseness (S) and the first lesion diffuseness (S0) is 1.5, but is not limited to this value.

[0191] The first lesion pixel count (V0), the first lesion center (C0), and the first lesion diffuseness (S0) are calculated from the mask image 1003 at the position of z0 initially specified by the inner line 211. Although these values ​​are fixed throughout FIG. 30, the first lesion pixel count (V0), the first lesion center (C0), and the first lesion diffuseness (S0) may indicate the state at an adjacent z. In other words, each time processing moves to an adjacent medical image 201, the first lesion pixel count (V0), the first lesion center (C0), and the first lesion diffuseness (S0) may be recalculated based on the mask image 1003 that was previously processed.

[0192] The lesion site (mask region 1002A) is not displayed in all mask images 1003. In the second embodiment, three parameters are determined across each mask image 1003: lesion site pixel count, lesion site center, and lesion site spread, but are not limited to these parameters.

[0193] Figures 31A to 31E are diagrams showing mask images 1003 with different z positions. In each of the mask images 1003 shown in Figures 31A to 31E, a mask region 1002A is displayed as a lesion site. The mask image 1003 shown in FIG. 31A is an example based on a reference mask image located at z0. Mask image 1003 shown in Fig. 31A displays mask region 1002A, and the first lesion center (C0) calculated in step S503 of Fig. 28 is indicated by reference numeral 1203. Although not shown in Fig. 31A, the first lesion pixel count (V0) and first lesion diffuseness (S0) are calculated based on mask image 1003 (S502, S503, and S504 of Fig. 28).

[0194] Mask images 1003A to 1003D shown in Figures 31B to 31E are mask images 1003 that exist at different z positions from mask image 1003 based on the reference mask image in Figure 31A. Note that inner line 211 and outer line 221 of mask image 1003 shown in Figure 31A are projected onto mask images 1003A to 1003D shown in Figures 31B to 31E. Mask region 1002A is also shown in Figures 31B to 31E. Based on mask images 1003A to 1003D shown in Figures 31B to 31E, second lesion pixel count (V), second lesion center (C), and second lesion diffuseness (S) are calculated (S5103, S5105, and S5107 in Figure 29).

[0195] The mask image 1003B shown in FIG. 31C has the following conditions: (A1) The ratio of the number of pixels in the second lesion site (V) to the number of pixels in the first lesion site (V0) is 0.05 or more. (A2) The distance between the second lesion center (C) (1203b) and the first lesion center (C0) (1203) is less than or equal to 0.5× the inner region diameter. (A3) The ratio of the second lesion site diffusivity (S) to the first lesion site diffusivity (S0) is 1.5 or less.

[0196] Therefore, in mask image 1003B shown in Fig. 31C, "no" is selected in all of steps S5104, S5106, and S5108 in Fig. 29, or in all of steps S5204, S5206, and S5208 in Fig. 30. In other words, the lesion site (mask area 1002A) shown in mask image 1003B is determined to be the same (contiguous) lesion site as the lesion site (mask area 1002A) shown in mask image 1003 in Fig. 31A.

[0197] A mask image 1003A shown in FIG. 31B satisfies the following conditions. (B1) The ratio of the number of pixels in the second lesion site (V) to the number of pixels in the first lesion site (V0) is less than 0.05. Therefore, in the mask image 1003A shown in FIG. 31B, "yes" is selected in step S5104 in FIG. 29 or in step S5204 in FIG.

[0198] The mask image 1003C shown in FIG. 31D has the following conditions: (C1) The distance between the second lesion center (C) (reference number 1203c) and the first lesion center (C0) (reference number 1203) is greater than 0.5× the inner region diameter. Therefore, in the mask image 1003C shown in FIG. 31D, "yes" is selected in step S5106 in FIG. 29 or in step S5206 in FIG.

[0199] The mask image 1003D shown in FIG. 31E has the following conditions: (D1) The ratio of the second lesion site diffusivity (S) to the first lesion site diffusivity (S0) is greater than 1.5. (D2) The distance between the second lesion center (C) (reference number 1203) and the first lesion center (C0) (reference number 1203d) is less than or equal to 0.5× the inner region diameter. Therefore, in the case of mask image 1003D shown in FIG. 31E, the determination in step S5108 in FIG. 29 or in step S5208 in FIG. 30 is "yes."

[0200] In the example shown in Figure 31E, the lesions are scattered even though the distance between the second lesion center (C) (1203d) and the first lesion center (C0) (1203) is less than 0.5 times the inner region diameter. In such a case, the ratio of the second lesion diffusivity (S) to the first lesion diffusivity (S0) is greater than 1.5 (a predetermined value).

[0201] According to the second embodiment, a three-dimensional image 250 (see FIG. 27) can be constructed by using a plurality of mask images 1003. In particular, the number of lesion pixels (V0, V), the lesion center (C0, C), and the lesion diffuseness (S0, S) are compared, and mask images 1003 that do not satisfy the conditions are removed from the construction of the three-dimensional image 250. In this manner, the mask images 1003 that constitute the three-dimensional image 250 can be quantitatively selected.

[0202] [Third embodiment] Next, a third embodiment of the present invention will be described with reference to Figs. 32 to 37. In the third embodiment, a tissue image 401 is selected, and a mixed tissue image 260 (see Fig. 34B) is generated by combining the selected tissue image 401. Then, a lesion site highlighting process is performed on this mixed tissue image 260. Note that the configuration of the computer 100 in the third embodiment is the same as that in Fig. 2, and is therefore not shown in the figure.

[0203] (Overall processing) Fig. 32 is a flowchart showing the overall procedure of the lesion site highlighting processing by computer 100, which is performed in the third embodiment. In Fig. 32, the same processes as those in Fig. 4 are given the same step numbers and the description thereof will be omitted. In the process shown in FIG. 32, the process of step S130B is performed instead of step S130 in FIG. In step S130B, the tissue image processor 104 generates (Register mixed components) the mixed tissue image 260. The process of step S130B will be described later with reference to FIG. Furthermore, in FIG. 4, the processes of steps S140 to S171 are repeated for each component, but in the process shown in FIG. 32, the processes of steps S140 to S171 are not repeated. 4 is replaced by step S160B, which will be described later. Furthermore, the process of step S181 is omitted.

[0204] (Mixed tissue image generation) FIG. 33 is a flowchart showing a detailed procedure for generating the mixed tissue image 260 (see FIG. 34B) in step S130B of FIG. First, the tissue image processing unit 104 loads an anatomical tissue model (not shown) stored in the storage device 120 (Load anatomical component model: S131 in FIG. 8). Next, the tissue image processor 104 calculates a probability component map relating to the tissue information of each pixel according to the read anatomical tissue model (Compute probability component map: S132 in FIG. 8). This allows the feature amount of the pixel for each tissue to be calculated. Then, the tissue image processor 104 refers to the probability tissue map for each pixel of the medical image 201 and determines which tissue the pixel currently being processed most likely belongs to (predetermined criterion) (Register component to medical image: S133 in FIG. 8). As a result, a tissue image 401 (see FIGS. 9B to 9E) is generated.

[0205] Thereafter, the user selects a tissue of interest (eg, gray matter, white matter) via the input device 131 (Select interest component: S134). Then, the tissue image processor 104 combines (combines) the tissue images 401 of the tissues of interest (e.g., gray matter, white matter) selected in step S134 (Combine image relevant to component of interest: S135). Note that the tissue images 401 of the tissues (e.g., skull, background) not selected in step S134 are ignored. As a result, a mixed tissue image 260 as shown in Fig. 34B is generated as a new second image. After completing step S135 in Fig. 33, the computer 100 returns the process to step S140 in Fig. 4.

[0206] (Example of mixed tissue imaging) FIG. 34A is a diagram showing the medical image 201 before being decomposed in steps S131 to S133 in FIG. In the example shown in FIG. 34A, the inner line 211, the outer line 221, the inner region 212, and the peripheral region 222 that are set in steps S110 and S120 in FIG. 32 are shown. In the medical image 201 shown in Fig. 34A, two tissues are identified: a tissue of interest (reference number 1301) to the user, and a tissue of no interest (reference number 1302). These two components exist in both the inner region 312 and the peripheral region 322. The tissue image processing unit 104 synthesizes only the tissue image 401 (see Figs. 9B to 9E) of the tissue of interest (reference number 1301) to the user, and ignores the tissue image 401 of the tissue of no interest (reference number 1302).

[0207] FIG. 34B is a diagram showing an example of a mixed tissue image 260 generated in step S135 of FIG. In the mixed tissue image 260, the tissue of no interest (reference numeral 1302) in Fig. 34A is omitted. In Fig. 34B, the reference numerals other than the mixed tissue image 206 are the same as those in Fig. 34A, and therefore the description thereof will be omitted.

[0208] In this way, the tissue image processing unit 104 generates an image by combining predetermined tissue images 401 from among the multiple tissue images 401, and sets this image as a mixed tissue image 260 (a new second image).

[0209] (Intensity value group probability distribution 500E) 35 is a diagram showing an example of an intensity value group probability distribution 500E according to the mixed tissue image 260, which is used in the third embodiment. FIG. 35 is generated in step S140 in FIG. 35, the horizontal axis in FIG. 34 indicates the intensity value group, and the vertical axis indicates the intensity value group probability. Furthermore, in FIG. 35, an inner region histogram 501E is the intensity value group probability in the inner region 212 of the mixed tissue image 260. Moreover, a peripheral region histogram 502E is the intensity value group probability in the peripheral region 222 of the mixed tissue image 260.

[0210] (Probability Difference Distribution 600E) FIG. 36 is a diagram showing an example of a probability difference distribution 600E of the mixed tissue image 260 (see FIG. 34B) used in the third embodiment. In FIG. 36, the horizontal axis indicates the intensity group, and the vertical axis indicates the probability difference 601. Fig. 36 is generated in step S151 of Fig. 32. Fig. 36 shows an example of a probability difference 601 obtained by subtracting the intensity value group probability distribution 500E (peripheral region histogram 502E in Fig. 35) of the peripheral region 222 (see Fig. 34A) from the intensity value group probability distribution 500E (inner region histogram 501E in Fig. 35) of the inner region 212 (see Fig. 34A) of the mixed tissue image 260. In the diagram shown in Fig. 36, two peaks are generated. That is, two intensity intervals 607Ea, 607Eb (607) exist.

[0211] (Detection range setting) Fig. 37 is a flowchart showing a detailed procedure for setting the detection range for the mixed tissue image 260 (see Fig. 34B). Fig. 37 shows the details of the process of step S160B in Fig. 32. Fig. 12A and Fig. 36 are referred to as appropriate. As shown in Fig. 36, a plurality of intensity intervals 607 associated with the number of detected maximum values ​​may be detected depending on the shape of the probability difference distribution 600E. Unlike the analysis for each tissue shown in the first embodiment, the lesion site in the mixed tissue image 260 may exhibit both low and high signal characteristics, and is therefore not limited to a single intensity interval 607. That is, in the third embodiment, in the probability difference distribution 600E, the intensity intervals 607 far from both the lower end 602 (see Fig. 12A) and the upper end 603 (see Fig. 12A) are discarded, but the detection range setting process is performed for the other intensity intervals 607.

[0212] First, the detection range setting unit 107 detects the intensity interval 607 (Detect intensity interval: S161 in FIG. 37). The process of step S161 is the same as step S161 in FIG. 37, the processes of steps S162 to S400 are performed for each strength interval 607. First, the detection range setting unit 107 detects an intensity group having a local maxima of the probability difference 601 in the mixed tissue image 260 (Detect Intensity Group with Local Maxima of Probability Difference (Imax): S162 in FIG. 37). The local maxima are defined as intensity groups having the maximum value of the probability difference 601 in each of the intensity intervals 607 being processed. Hereinafter, the intensity value group (maximum value) where the probability difference 601 is maximum will be referred to as a maximum intensity value group 604 (Imax: see FIGS. 12A to 12C).

[0213] Next, the detection range setting unit 107 detects a group of intensity values ​​related to the minimum value 605 (Ilow) and the maximum value 606 (Ihigh) of the intensity section 607 (Detect intensities related to lower end (Ilow) and upper end (Ihigh) of Imax: S163 in FIG. 37). The detection of the minimum value 605 and the maximum value 606 may be performed by linear extrapolation or linear interpolation, but is not limited to a "linear" method.

[0214] Next, the detection range setting unit 107 calculates the length (ΔI) of the intensity interval 607 from the minimum value 605 (Ilow) to the maximum value 606 (Ihigh) (Calculate length of intensity interval (ΔI) from lower end (Ilow) to upper end (Ihigh): S164 in FIG. 37).

[0215] Here, in FIG. 13, if a plurality of strength intervals 607 are detected in step S165, one of the strength intervals 607 is selected in step S166, but in FIG. 37, these processes are not performed.

[0216] In the example of FIG. 36, two intensity intervals 607Ea, 607Eb are found in the probability difference 601 of the mixed tissue image 260.

[0217] 37, the detection range setting unit 107 determines the location of the intensity interval 607. This process is similar to step S200 in FIG. 13 and the process performed in FIG.

[0218] Next, the detection range setting unit 107 judges whether the position of any intensity interval 607 is close to either the lower end 602 (see FIG. 12A: I0) or the upper end 603 (see FIG. 12A: Iend) (Is intensity interval near to intensity edge?: S167 in FIG. 37). The process of step S167 in FIG. 37 is the same as the process of step S167 in FIG. 13. If it is determined that none of the strength sections 607 are located near either edge (S167→no in FIG. 37), the detection range setting unit 107 returns the process to step S162 and performs processing on the next strength section 607. If there is no next strength section 607, the detection range setting unit 107 ends the process. If the intensity interval 607 is close to an edge (S167→yes), the detection range setting unit 107 extends the intensity interval 607 to the nearest edge (Extend intensity interval to the nearest intensity edge: S300 in FIG. 37). Then, the detection range setting unit 107 adjusts the intensity interval 607 that is the determined detection range (Adjust intensity interval: S400 in FIG. 37). Steps S300 and S400 are the same processes as steps S300 and S400 in FIG. 13.

[0219] According to the third embodiment, it is possible to perform a lesion highlighting process on a mixed tissue image obtained by combining a tissue image 401 of a tissue in which the user is interested.

[0220] The present invention is not limited to the above-described embodiment, and includes various modified examples. For example, the above-described embodiment has been described in detail to clearly explain the present invention, and is not necessarily limited to all of the configurations described. In addition, it is possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. In addition, it is possible to add, delete, or replace a part of the configuration of each embodiment with another configuration.

[0221] In addition, in this embodiment, the lesion area highlighting process is performed on the medical image 201, but this is not limited to the above. That is, the image processing device 1 of this embodiment can also be applied to a process for highlighting a predetermined area on an image other than the medical image 201.

[0222] Furthermore, in this embodiment, the medical image 201 is an image relating to the brain, but it may be an image relating to an organ other than the brain.

[0223] Further, the above-mentioned configurations, functions, image acquisition unit 101 to display processing unit 110, three-dimensional image processing unit 110A, storage device 120, etc. may be realized in hardware by designing some or all of them as integrated circuits, for example. As shown in Fig. 3, the above-mentioned configurations, functions, etc. may be realized in software by a processor such as a CPU interpreting and executing a program that realizes each function. Information such as a program, table, file, etc. that realizes each function can be stored in a memory, a recording device such as an SSD (Solid State Drive), or a recording medium such as an IC (Integrated Circuit) card, an SD (Secure Digital) card, or a DVD (Digital Versatile Disc), in addition to being stored in an HD (Hard Disk). In addition, in each embodiment, the control lines and information lines are those that are considered necessary for the explanation, and not all control lines and information lines in the product are necessarily shown. In reality, it can be considered that almost all components are connected to each other. [Explanation of symbols]

[0224] 1 Image processing device 100,100A Computer 101 Image acquisition unit 102 Inner region setting unit (region setting unit) 103 Surrounding area setting unit (area setting unit) 104 Tissue image processing section (image processing section) 105 Intensity value group probability distribution calculation unit (intensity value frequency distribution calculation unit) 106 Probability Difference Calculation Unit 107 Detection range setting section 108 Mask setting section (pixel selection section) 109 Mask editing section (pixel selection section) 110 Display processing section (output processing section) 110A 3D image processing unit 131 Input device (input section) 132 Display device (output section) 201 Medical Images (First Image, Brain-related Images) 202 Lesion area (area of ​​interest) 211,211a,211b inner line 212 Inner Region (First Region) 221,221a,221b Outside line 222 Surrounding Area (Second Area) 231 Expanded area (area adjacent to the first area) 250 3D image (5th image) 260 Mixed Tissue Image (New Secondary Image) 401, 401a to 401d Tissue images (second images) 401A~401D Tissue images (third images) 500,500E Intensity value group probability distribution 501, 501E Inner area histogram (first intensity value frequency distribution) 501a to 501d histogram (first intensity value frequency distribution) 502, 502E Surrounding area histogram (second intensity value frequency distribution) 502a to 502c histogram (second intensity value frequency distribution) 600 Probability Difference Distribution (Probability Difference Frequency Distribution) 601 Probability difference (difference value) 602 Bottom end 603 Upper end 605 Minimum 606 Maximum 607, 607a, 607b, 607Ea, 607Eb Intensity range (detection range) 611,611a,611b Distribution area 701 Cumulative intensity value group probability (first cumulative intensity value probability by first method) 702 Cumulative Probability Difference (Second Cumulative Intensity Value Probability by First Method) 703 code (predetermined threshold in the first method) 704 Associated Strength (Strength value selected in the first method) 705 Selection cumulative probability (first selection cumulative intensity value probability in the first method) 801 Cumulative intensity value group probability (third cumulative intensity value probability by second method) 802 Cumulative Probability Difference (fourth cumulative intensity value probability by the second method) 803 code (predetermined threshold in the second method) 804 Associated Strength (Strength value selected by the second method) 805 Selection cumulative probability (second selection cumulative intensity value probability in the second method) 1002 Mask area (selected pixels, highlighted pixels, specified area) 1002A Mask area (selected pixels) 1003, 1003a, 1003b, 1003A to 1003D Mask image (fourth image, a fourth image) 1100 display screen 1101 Screen 1102 screens M Medical imaging equipment C0 First lesion center (first center) C. Center of the second lesion (second center) S0 First lesion site diffusion (first diffusion) S Second lesion site diffusion (second diffusion) V0 Number of pixels in the first lesion area (number of pixels in the first area) V Number of pixels in the second lesion (Number of pixels in the second lesion) Z Image Processing System S110 Setting the inner region (region setting step) S120 Setting the surrounding area (area setting step) S130 Decomposition of medical images into tissue images (tissue decomposition step) S140 Calculation of intensity value group probability (intensity value frequency distribution calculation step) S151 Calculation of probability difference (probability difference calculation step) S160 Setting the detection range (detection range setting step) S171 Mask setting (pixel selection step) S191 Display (output step)

Claims

1. a region setting unit that sets a second region in the first image, the second region being a region adjacent to a first region set in the first image as a region including a region of interest in the first image, and not including the first region; an intensity value frequency distribution calculation unit that calculates a first intensity value frequency distribution, which is information regarding a distribution of the number of pixels having a predetermined intensity value, for the first region, and calculates a second intensity value frequency distribution, which is information regarding a distribution of the number of pixels having a predetermined intensity value, for the second region; a probability difference calculation unit that calculates a probability difference frequency distribution by calculating a difference value between the first intensity value frequency distribution and the second intensity value frequency distribution for each of the predetermined intensity values; an image processing unit that generates a plurality of second images that are decomposed based on the first image according to a predetermined criterion; a detection range setting unit that applies a detection range setting algorithm to each of the plurality of second images and sets a detection range of the intensity value of the first image for each of the second images based on the difference value; a pixel selection unit that selects pixels to be highlighted for each of the second images based on the detection range set for each of the second images; an output processing unit that outputs to an output unit a third image in which the pixels selected by the pixel selection unit are highlighted; An image processing device comprising:

2. The pixel selection unit is A fourth image is generated by combining the selected pixels.

2. The image processing device according to claim 1,

3. The detection range setting unit is a section in which the difference value has a positive value in the probability difference frequency distribution is set as the detection range; if the detection range is close to a lower end of the probability difference frequency distribution, the minimum value of the detection range is extended to the lower end; If the detection range is close to the upper end of the probability difference frequency distribution, the maximum value of the detection range is extended to the upper end.

2. The image processing device according to claim 1,

4. The detection range setting unit is When a plurality of detection ranges are detected, the detection range having the largest area of ​​the probability difference frequency distribution in the detection range is selected.

4. The image processing device according to claim 3.

5. The detection range setting unit is calculating a first cumulative intensity value probability by accumulating the second intensity value frequency distribution from a minimum value side of the detection range toward a maximum value side of the detection range for the second region; calculating a second cumulative intensity value probability by accumulating the difference values ​​from a minimum value side of the detection range toward a maximum value side of the detection range; selecting the intensity value based on a predetermined threshold and the first cumulative intensity value probability; a first method for determining the second cumulative intensity value probability corresponding to the selected intensity value as a first selected cumulative intensity value probability; Run calculating a third cumulative intensity value probability by accumulating the second intensity value frequency distribution from a maximum value side of the detection range toward a minimum value side of the detection range; calculating a fourth cumulative intensity value probability by accumulating the difference values ​​from a maximum value side of the detection range toward a minimum value side of the detection range for the second region; selecting the intensity value based on a predetermined threshold and the third cumulative intensity value probability; a second method for determining the fourth cumulative intensity value probability corresponding to the selected intensity value as a second selected cumulative intensity value probability; Run comparing the first selected cumulative intensity value probability with the second selected cumulative intensity value probability; If the first selected cumulative intensity value probability is greater than the second selected cumulative intensity value probability, the maximum value of the detection range is set as the intensity value selected by the first method; If the second selected cumulative intensity value probability is greater than the first selected cumulative intensity value probability, the minimum value of the detection range is set as the intensity value selected by the second method.

4. The image processing device according to claim 3.

6. There are a plurality of the fourth images; a three-dimensional image processing unit that generates a fifth image, which is a three-dimensional image, by stacking a plurality of the fourth images; 3. The image processing apparatus according to claim 2, further comprising:

7. The three-dimensional image processing unit includes: calculating, in the fourth image, a first pixel number which is the number of pixels of a highlighted display pixel which is the pixel that is highlighted, a first center which is the center of the highlighted display pixel, and a first diffusion degree which is the average of the distance between the first center and each of the pixels that constitute the highlighted display pixel; calculating a second number of pixels which is the number of pixels of the highlighted display pixel, a second center which is the center of the highlighted display pixel, and a second diffusion degree which is the average of the distance between the second center and each of the pixels which constitute the highlighted display pixel, in the fourth image which is different from the fourth image in which the first number of pixels, the first center, and the first diffusion degree are calculated; The fourth image is continuously selected as the fourth image constituting the fifth image from among the plurality of fourth images until a ratio of the second number of pixels to the first number of pixels becomes less than a first value that is a predetermined value, a distance between the second center and the first center becomes greater than a second value that is a predetermined value, or a ratio of the second diffusion degree to the first diffusion degree becomes greater than a third value that is a predetermined value.

7. The image processing device according to claim 6,

8. The image processing unit includes: An image is generated by synthesizing a predetermined one of the plurality of second images, and the synthesized image is set as a new second image.

3. The image processing device according to claim 2.

9. The first image is a medical image captured by a medical image capture device.

2. The image processing device according to claim 1,

10. The first image is an image of a brain.

2. The image processing device according to claim 1,

11. An image processing device that highlights a predetermined area of ​​a first image, a region setting step of setting a second region in the first image, the second region being a region adjacent to a first region set in the first image as a region including a region of interest in the first image, the second region being a region not including the first region; an intensity value frequency distribution calculation step of calculating a first intensity value frequency distribution, which is information regarding a distribution of the number of pixels having a predetermined intensity value, for the first region, and calculating a second intensity value frequency distribution, which is information regarding a distribution of the number of pixels having a predetermined intensity value, for the second region; a probability difference calculation step of calculating a probability difference frequency distribution by calculating a difference value between the first intensity value frequency distribution and the second intensity value frequency distribution for each of the predetermined intensity values; a tissue decomposition step of generating a plurality of second images based on the first image, the second images being decomposed according to a predetermined criterion; a detection range setting step of applying a detection range setting algorithm to each of the plurality of second images and setting a detection range of the intensity value of the first image for each of the second images based on the difference value; a pixel selection step of selecting pixels to be highlighted for each of the second images based on the detection range set for each of the second images; an output step of outputting to an output unit a third image in which the pixels selected in the pixel selection step are highlighted; An image processing method that performs

Images