JP5832938B2 - Image processing apparatus, method, and program - Google Patents

Description

  The present invention relates to an image processing apparatus, method, and program for obtaining a corresponding positional relationship between two three-dimensional images of the same subject to be compared.

  In image diagnosis, a change in a medical condition can be confirmed by comparing an image of a subject acquired in the latest examination with an image of the same subject obtained in a previous examination. At this time, alignment between the images is performed in order to facilitate comparison between the two images. Specifically, as shown in Patent Document 1 and Non-Patent Document 1, a control point that divides the space at a predetermined interval is set in one image, and the position of this control point is displaced and deformed. When the degree of similarity between the first image and the other image is evaluated using a predetermined evaluation scale, the displacement amount of the control point that maximizes the evaluation value is determined, and the displacement amount of the control point at this time is used. To align the images.

JP-T-2006-506153

J. Masumoto et al., “A similarity measure for nonrigid volume registration using known joint distribution of targeted tissue: Application to dynamic CT data of the liver” Medical Image Analysis 7 (2003), pp. 553-564

  By the way, conventionally, since only a necessary minimum range including a specific part that is attracting attention in diagnosis is photographed, one evaluation specialized for the specific part is also performed in the alignment processing for the obtained image. It was sufficient to use a scale. However, in recent years, with the advancement of exposure reduction technology, it has become possible to image a wider range including a large number of sites in a single imaging, and to perform multiple diagnoses for multiple sites in a single examination. became. Accordingly, there has been a demand for a technique capable of aligning each of a large number of parts with sufficient accuracy in the alignment process for the obtained image.

  Note that Patent Document 1 and Non-Patent Document 1 both propose a method of performing alignment by applying one evaluation scale specialized for a specific part of interest for the entire image given to both. However, these methods cannot answer the above demand.

  In view of the circumstances described above, an object of the present invention is to provide an image processing apparatus, method, and program capable of obtaining a more accurate corresponding positional relationship between two three-dimensional images of the same subject to be compared. Is.

  An image processing apparatus according to the present invention is an image processing apparatus for obtaining a corresponding positional relationship between two three-dimensional images of the same subject to be compared, and for each of a plurality of parts constituting the subject. A storage unit that associates and stores a deformation method used when deforming and an evaluation scale used when evaluating the degree of similarity between two images, and each of the tomographic images forming the two three-dimensional images is a subject. A part information acquisition unit that acquires information about whether a tomographic image represents a part of the image, and a set of partial images that are a set of tomographic images representing the same part in two three-dimensional images using the acquired information A partial image specifying unit that specifies a plurality of images, and for each set of specified partial images, a deformation method that is stored in association with a part represented in each partial image is used to store one of the three-dimensional images. Partial image The partial image is deformed by moving each of the pixels by a predetermined movement amount, and the degree of similarity between the deformed partial image and the partial image of the other three-dimensional image is represented in the partial image. An evaluation value obtained by repeatedly performing the process of evaluating using the evaluation scale stored in association with the image while changing the movement amount given to each pixel in the deformation, and evaluating the similarity with respect to the change in the movement amount A movement amount calculation unit for obtaining a movement amount given to each pixel in the deformation when the change amount of the image becomes a predetermined value or less, and a corresponding positional relationship between two three-dimensional images using the obtained movement amount. A positional relationship calculation unit to be obtained (first image processing apparatus).

  Here, “deforming a partial image” means changing the position of at least some of the pixels in the partial image, and is not limited to the deformation accompanied by a change in the external shape of the partial image, only the translation or Deformation by rotation alone is also included.

  As specific examples of “parts”, when the subject is a human body, various anatomies such as organs such as lungs / kidneys / hearts / colon, blood vessels / trachea / bronchi, skin, bones such as ribs / vertebral bodies / skulls, etc. In addition to biological structures, the head, neck, chest, abdomen, pelvis, legs, and other complex parts including two adjacent parts of each part, such as the head and neck and the chest abdomen .

  In the first image processing apparatus, the movement amount calculation unit represents each pixel of a partial image including a set of tomographic images representing a plurality of parts in the one three-dimensional image. For each of the parts, the movement amount is obtained using the deformation method and the evaluation scale associated with the part, and the positional relationship calculation unit is a value obtained by averaging the movement amounts obtained for the respective parts. May be used as the obtained movement amount to obtain a corresponding positional relationship between two three-dimensional images.

  In addition, the positional relationship calculation unit is configured to calculate a movement amount calculation unit for each pixel in a portion where the partial images obtained by using different deformation methods and / or different evaluation scales in the one three-dimensional image are adjacent to each other. The movement amount of the pixel obtained by the above is averaged using the movement amount in each of two or more pixels located in the vicinity of the pixel, and the averaged value is used as the obtained movement amount. The corresponding positional relationship between the three-dimensional images may be obtained.

  In addition, the positional relationship calculation unit obtains an estimated value of the movement amount of each pixel in a portion other than the specified partial image in the one three-dimensional image using the movement amount obtained by the movement amount calculation unit, A corresponding positional relationship between two three-dimensional images may be obtained using the obtained estimated value and the obtained movement amount.

  The storage unit stores any of non-rigid registration, rigid registration, and surface registration as a deformation method for each of a plurality of parts constituting the subject, and stores correlation coefficients, mutual information amounts, and pixel values. Any one of the differences may be stored as an evaluation scale. Note that the storage unit is not limited to the deforming unit and the evaluation scale exemplified here, and may store another deforming unit or evaluation scale suitable for each part in association with the part. For example, Mark Holden et al., “Voxel Similarity Measures for 3-D Serial MR Brain Image Registration”, IEEE Transactions on Medical Imaging, vol. 19, no. 2, February 2000, pp. 94-102 May be stored in association with each other.

  In the one three-dimensional image, the movement amount calculation unit selects each pixel of the partial image including a set of tomographic images representing a plurality of parts from the plurality of parts represented in the partial image. The amount of movement may be obtained using a deformation method and an evaluation scale stored in association with any one part.

  The image processing apparatus of the present invention is an image processing apparatus for obtaining a corresponding positional relationship between two three-dimensional images of the same subject to be compared, and for each of a plurality of parts constituting the subject. A storage unit that associates and stores a deformation technique used when deforming an image and an evaluation scale used when evaluating the similarity between two images, and the same part is represented in two three-dimensional images. A part image specifying unit that specifies a plurality of sets of part images, and a deformation method that is stored in association with the part represented in each part image for each part image specified The part image is deformed by moving each pixel in the part image of the three-dimensional image by a predetermined movement amount, and the degree of similarity between the part image after deformation and the part image of the other three-dimensional image It is shown in the part image Evaluation by using the evaluation scale stored in association with the position repeatedly while changing the amount of movement given to each pixel in the deformation, and evaluating the similarity to the change in the amount of movement A movement amount calculation unit that obtains a movement amount of each pixel of the part image in the deformation in which an evaluation value given to each pixel in the deformation when the change amount of the value is equal to or less than a predetermined value is equal to or greater than a predetermined value; Using the obtained amount of movement, an estimated value of the amount of movement of each pixel in a portion other than the specified part image in the one three-dimensional image is obtained, and using the obtained estimated value and the obtained amount of movement And a positional relationship calculation unit that obtains a corresponding positional relationship between two three-dimensional images (second image processing apparatus).

  Here, the positional relationship calculation unit calculates a movement amount for each pixel in a portion where the movement amounts are obtained by using different deformation methods and / or different evaluation scales in the one three-dimensional image. The movement amount of the pixel obtained by the unit is averaged using the movement amounts in each of two or more pixels located in the vicinity of the pixel, and the averaged value is used as the obtained movement amount. A corresponding positional relationship between two three-dimensional images may be obtained.

  The positional relationship calculation unit calculates an estimated value of the amount of movement of each pixel in a portion other than the specified part image in the one three-dimensional image, and the amount of movement in each of two or more pixels located in the vicinity of the pixel. May be obtained by interpolation using.

  The first and second image processing methods of the present invention are methods in which processes performed by the respective units of the first and second image processing apparatuses are executed by at least one computer.

  The first and second image processing programs of the present invention are programs that cause at least one computer to execute the processes performed by the respective units of the first and second image display devices. This program is recorded on a recording medium such as a CD-ROM or DVD, or recorded in a state where it can be downloaded to a storage attached to a server computer or a network storage, and provided to the user.

  According to the first image processing apparatus, method, and program of the present invention, when a corresponding positional relationship is obtained between two three-dimensional images of the same subject to be compared, each of a plurality of parts constituting the subject is obtained. On the other hand, the deformation method used when deforming the image and the evaluation scale used when evaluating the similarity between the two images are stored in association with each other, and each of the tomographic images constituting the two three-dimensional images is stored. Information on which part of the subject is a tomographic image is acquired, and using the acquired information, a plurality of sets of partial images composed of a set of tomographic images representing the same part in two three-dimensional images are obtained. For each identified partial image of each set, each pixel in the partial image of one three-dimensional image is determined using a deformation method stored in association with the portion represented in those partial images. With a certain amount of movement The partial image is deformed by moving, and the degree of similarity between the deformed partial image and the partial image of the other three-dimensional image is stored in association with the part represented in the partial image. The process of evaluating using a scale is repeatedly performed while changing the amount of movement given to each pixel in the deformation, and the amount of change in the evaluation value that evaluates the similarity to the change in the amount of movement is less than or equal to a predetermined value. Since the movement amount given to each pixel in the deformation in the case is obtained and the corresponding positional relationship between two three-dimensional images is obtained using the obtained movement amount, it is suitable for each of a plurality of parts. By using a deformation method and an evaluation scale, it is possible to obtain a corresponding positional relationship between image parts representing the part, and to obtain a more accurate corresponding positional relationship between two three-dimensional images as a whole. Kill.

  In addition, in the one three-dimensional image, for each pixel of the partial image formed of a set of tomographic images representing a plurality of parts, a deformation method associated with the part for each part represented in the partial image Then, the movement amount is obtained using an evaluation scale, and the corresponding positional relationship between two three-dimensional images is obtained using a value obtained by averaging the movement amounts obtained for each part as the obtained movement amount. In this case, it is possible to obtain a corresponding positional relationship that is averagely suitable for all of the plurality of parts represented in the partial image.

  Further, in the one three-dimensional image, for each pixel in a portion where the partial images obtained by using different deformation methods and / or different evaluation scales are adjacent to each other, the pixel of the pixel obtained by the movement amount calculation unit is obtained. The movement amount is averaged using the movement amount in each of two or more pixels located in the vicinity of the pixel, and the averaged value is used as the obtained movement amount to correspond between the two three-dimensional images. When the relationship is obtained, the corresponding positional relationship between the partial images for which the movement amount is obtained using different deformation methods and / or different evaluation scales can be smoothly connected.

  Further, in the one three-dimensional image, for each pixel of the partial image including a set of tomographic images representing a plurality of parts, any one part selected from the plurality of parts represented in the partial image When the movement amount is obtained by using the deformation method and the evaluation scale stored in association with each other, it is possible to obtain an accurate corresponding positional relationship with respect to one selected part.

  According to the second image processing apparatus, method, and program of the present invention, when the corresponding positional relationship is obtained between two three-dimensional images of the same subject to be compared, each of a plurality of parts constituting the subject is obtained. On the other hand, the deformation method used when deforming the image and the evaluation scale used when evaluating the similarity between the two images are stored in association with each other, and the same part is represented in the two three-dimensional images. One of the three-dimensional images is identified by using a deformation method stored in association with the part represented in each part image for each part image of the part part identified. The part image is deformed by moving each pixel in the part image by a predetermined movement amount, and the degree of similarity between the part image after deformation and the part image of the other three-dimensional image is displayed in the part image. Corresponding to the part that is A process of evaluating using a stored evaluation scale is repeated while changing the amount of movement given to each pixel in the deformation, and the amount of change in the evaluation value that evaluates the similarity to the change in the amount of movement is The amount of movement given to each pixel in the deformation when it is less than or equal to a predetermined value is obtained, and the obtained amount of movement is used to determine the amount of each pixel in the portion other than the specified part image in the one three-dimensional image. Since the estimated value of the movement amount is obtained, and the corresponding positional relationship between the two three-dimensional images is obtained using the obtained estimated value and the obtained movement amount, the deformation suitable for each of the plurality of parts Using a technique and an evaluation scale, it is possible to obtain the corresponding positional relationship between image parts representing the part, and to obtain a more accurate corresponding positional relationship between two three-dimensional images. It can be.

  Note that in the one three-dimensional image, for each pixel in a portion where the part images obtained by using different deformation methods and / or different evaluation scales are adjacent to each other, the pixel of the pixel obtained by the movement amount calculation unit is obtained. The movement amount is averaged using the movement amount in each of two or more pixels located in the vicinity of the pixel, and the averaged value is used as the obtained movement amount to correspond between the two three-dimensional images. When the relationship is obtained, the corresponding positional relationship between the part images for which the movement amount is obtained using different deformation methods and / or different evaluation scales can be smoothly connected.

1 is a schematic block diagram of an image generation apparatus according to a first embodiment. The figure for demonstrating the process by the image generation apparatus in 1st Embodiment. The flowchart which shows operation | movement of the image generation apparatus in 1st Embodiment. Schematic block diagram of an image generation apparatus according to the second embodiment The figure for demonstrating the process by the image generation apparatus in 2nd Embodiment. The flowchart which shows operation | movement of the image generation apparatus in 2nd Embodiment. The figure which shows the example of the table used for 1st Embodiment The figure which shows the example of the table used for 2nd Embodiment

  Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram illustrating a schematic configuration of the image processing apparatus 1. The configuration of the image processing apparatus 1 as shown in FIG. 1 is realized by executing an image processing program read into the auxiliary storage device on the computer. At this time, the image processing program is stored in a storage medium such as a CD-ROM or distributed via a network such as the Internet and installed in a computer.

  The image processing program includes an image acquisition process, a part information acquisition process, a partial image specifying process, a movement amount calculation process, a positional relationship calculation process, an image generation process, and a display control as processes to be executed by the CPU of the computer. In accordance with this rule, the CPU executes each of the above processes, so that the computer moves the image acquisition unit 12, the part information acquisition unit 13, the partial image specification unit 14, and the movement described later. It functions as an amount calculation unit 15, a positional relationship calculation unit 16, an image generation unit 17, and a display control unit 18.

  Further, the image processing apparatus 1 includes a storage device 11 such as a hard disk drive, and this storage device 11 corresponds to a storage unit of the present invention. Further, a display device 3 such as a display and an input device 4 such as a mouse and a keyboard are connected to the image processing apparatus 1.

  In the storage device 11, an image Va (hereinafter referred to as “first image Va”) obtained by photographing a predetermined subject using an imaging device such as CT, MRI, PET, SPECT, or an ultrasonic image in the latest examination. In addition, an image Vb (hereinafter, referred to as “second image Vb”) obtained by photographing the same subject by any one of the above-described photographing apparatuses in a past examination is stored. The first image Va and the second image Vb may be images taken with the same modality or may be images taken with different modalities.

  Each of the first image Va and the second image Vb is configured as an aggregate of image data of axial discontinuity images (tomographic images) with a predetermined slice interval and slice thickness. Of the tomographic images constituting the first image Va and the second image Vb, for each of all tomographic images except for some tomographic images whose parts are not recognized, the tomographic image is any part of the subject. The part information indicating whether the tomographic image represents is stored in association with each other. FIG. 2 is an image diagram showing part information stored in association with each of the first image Va, the second image Vb, and the tomographic images constituting the images. As shown in FIG. 2, in the tomographic images constituting the first image Va and the second image Vb, “Head” representing the head, “Neck” representing the neck, “Thorax” representing the chest, and the liver are represented. Site information such as “Live” and “Abdoman” representing the lower abdomen are associated with each other.

  The part information is acquired in advance by performing part recognition processing on each tomographic image. As the part recognition process, for example, a technique described in Japanese Patent Laid-Open No. 2008-259682 can be used. The method described in Japanese Patent Laid-Open No. 2008-259682 normalizes a plurality of input tomographic images, calculates a large number of feature amounts from the normalized tomographic images, and calculates each normalized tomographic image. The feature amount is input to the discriminator obtained by the AdaBoost method, the score for each part representing the part-likeness is calculated, and the body part of the human body is obtained using the calculated part score as an input and using dynamic programming. This is a method of determining the site represented in each tomographic image so that the arrangement order of the images is maintained. Also, a method using color template matching (see, for example, Japanese Patent Application Laid-Open No. 2002-253539), a method using a unique image of each part (for example, see Japanese Patent Application Laid-Open No. 2003-10166), or the like can be used.

  Further, in the storage device 11, for each of a plurality of parts constituting the subject, a deformation method used when deforming the image is associated with an evaluation scale used when evaluating the similarity between the two images. Table is stored. FIG. 7 is a diagram showing an example of a table used in this embodiment. As shown in FIG. 7, in the table T1, a deformation method and an evaluation scale are associated with each of a plurality of parts. Specifically, rigid registration and correlation coefficient in the head, surface registration and mutual information in the neck, non-rigid registration and the square of the pixel value difference in the chest, Is associated with non-rigid registration and mutual information, the lower abdomen is associated with surface registration and correlation coefficient, and the lower limb is associated with surface registration and the least square error between surfaces.

  Here, the non-rigid registration is based on the premise that the target image to be deformed according to the reference image is non-rigid, that is, the distance between any two points in the target image is variable. The rigid registration is based on the premise that the target image is a rigid body, that is, the distance between any two points in the target image is always constant. This is a method for changing the position of each pixel constituting the target image under the assumption that the target image is constant. Surface registration is a point sequence on the reference image corresponding to a point sequence on the target image having a surface shape. This is a method of changing the position of each pixel constituting the target image so that the residual is minimized.

Further, the correlation coefficient (NCC: peason product-moment corss correlation) is a value calculated by the following formula (1), and the mutual information (MI) is calculated by the following formula (2). It is a value, and the evaluation value of similarity becomes large, so that those evaluation scales are large. In the following formulas (1) to (3), a region Xo represents a subset of No pixel positions constituting the reference image X. Here, the reference image X is deformed according to the reference image X. The target image X ′ is an image existing in the space after deformation. That is,
It is. Here, T is a deformation function. F (Xo) represents the pixel value of the pixel located at the pixel position Xo in the reference image X, and g (Xo) represents the pixel value of the pixel in the modified target image X ′ corresponding to that pixel. , F (Xo) and G (Xo) represent a pixel value group in a portion where the reference image X and the target image X ′ after deformation overlap.

  A pixel value difference (MSD: mean square difference of intensities) is a value calculated by the following equation (3), and the smaller the value, the larger the similarity evaluation value. In the following equation (3), the difference image D (Xo) represents a difference in pixel values in a portion where the reference image X and the target image X ′ after deformation overlap. That is, d (Xo) = f (Xo) −g (Xo). P (fa) represents the probability that the pixel value fa appears in F (Xo), P (gb) represents the probability that the pixel value gb appears in G (Xo), and P (fa, gb) This represents the probability that the pixel value fa appears in F (Xo) and the pixel value gb appears in G (Xo).

  The image acquisition unit 12 acquires the first image Va and the second image Vb stored in the storage device 11, and the part information acquisition unit 13 each of the tomographic images constituting the first image Va and the second image Vb. The site | part information memorize | stored in the memory | storage device 11 in response to is acquired.

  The partial image specifying unit 14 uses the part information acquired by the part information acquiring unit 13 to specify a plurality of sets of partial images including a set of tomographic images representing the same part in the first image Va and the second image Vb. To do. Specifically, as shown in FIG. 2, in the partial image Ra1 and the second image Vb, which include all tomographic images in the first image Va, in which “Head” is included in the associated part information. The partial image Ra2 and the second image consisting of all the tomographic images in the first image Va, in which “Neck” is included in the part information associated with the set of the partial images Rb1 consisting of all the tomographic images. A set of partial images Rb2 composed of all tomographic images in Vb,..., A portion composed of all tomographic images in the first image Va in which “Adoman” is included in the associated part information The image Ra5 and a set of partial images Rb5 including all tomographic images in the second image Vb are specified.

  The movement amount calculation unit 15 refers to a table T1 as shown in FIG. 7 for each set of partial images Rai and Rbi (i = 1 to 5) specified by the partial image specification unit 14 and sets the set of the partial images. The deformation method stored in association with the part represented in the partial image and the evaluation scale are specified, and each pixel in the partial image Rai of the first image Va is set to a predetermined value using the specified deformation method. A process of deforming the part image Rai by moving it by a movement amount, and evaluating the degree of similarity between the deformed partial image Rai ′ and the partial image Rbi of the second image Vb using the specified evaluation scale. In this modification, the amount of change m (Xo) given to each pixel Xo (Xo∈Rai) is repeatedly changed, and the amount of change in the evaluation value that evaluates the degree of similarity with respect to the change in the amount of movement becomes a predetermined value or less. In case of said deformation Serial calculating the moving amount m (Xo) given to each pixel. Here, the amount of movement is m, a vector amount having the direction and distance of movement, and the transformed partial image Rai ′ is formed by displacing each pixel Xo of the partial image Rai by the amount of movement m (Xo). This is an image. In the following description, a set of movement amounts m (Xo) for each pixel Xo in the partial image Rai is referred to as a movement amount m (Rai).

  Specifically, the movement amount calculation unit 15 changes the evaluation value change amount with respect to the change in the movement amount m (Rai) to a predetermined value while changing the movement amount m (Rai) in the direction in which the evaluation value increases. By repeatedly performing the comparison process, the movement amount m (Rai) when the change amount of the evaluation value is equal to or less than the predetermined value is specified. Here, as the predetermined value, an arbitrary value that can be regarded as a sufficiently large change amount of the evaluation value with respect to the change of the movement amount m (Rai) may be set.

  Instead of this, for example, an evaluation value obtained by evaluating the similarity between the partial image Rai ′ and the partial image Rbi under the setting of the movement amount m (Rai) while changing the movement amount m (Rai) is a predetermined value. The amount of movement m (Rai) in the case where the evaluation value of the similarity is equal to or greater than a predetermined value may be specified by repeatedly performing the process of comparing with. Here, as the predetermined value, an arbitrary value that can be regarded as a sufficiently large evaluation value of the similarity may be set. Further, for example, evaluation values obtained by evaluating the degree of similarity between the partial image Rai ′ and the partial image Rbi are calculated for a plurality of different movement amounts m (Rai), and the evaluation value is the largest among the calculated evaluation values. May be specified, and the movement amount m (Rai) corresponding to the specified maximum evaluation value may be specified.

  In addition, the movement amount calculation unit 15 applies, for each part represented in the partial image, each part of the partial image including a set of tomographic images representing a plurality of parts in the first image Va. The amount of movement is obtained using the associated deformation technique and evaluation scale. That is, as shown in FIG. 2, a partial image in which two parts “Thorax” and “Live” are included in the associated part information, that is, an image of a part where the partial image Ra3 and the partial image Ra4 overlap. Specifies a movement amount as a part image representing the chest and a movement amount as a part image representing the liver, respectively.

  The positional relationship calculation unit 16 uses the movement amount m (Rai) obtained for each set of partial images Rai and Rbi (i = 1 to 5) by the movement amount calculation unit 15 to perform the first image Va and the second image. The corresponding positional relationship of all pixels between Vb is obtained. Specifically, a vector field that gives a movement amount to each of all the pixels in the first image Va is obtained. This vector field indicates that each pixel has a corresponding relationship with a pixel on the second image Vb at a position separated by the amount of movement given to that pixel.

  First, a movement amount m (Xo) determined by the movement amount m (Rai) specified by the movement amount calculation unit 15 is given to each pixel Xo of the partial image Rai (i = 1 to 5), and the first image Two or more pixels located in the vicinity of the estimated value of the movement amount of each pixel in a portion other than the partial image Rai (i = 1 to 5) in Va (for example, the portion Raf shown in FIG. 2). Are obtained by interpolation using the amount of movement given to each of them, and the obtained estimated value is given as the amount of movement of the pixel. Various spatial interpolation techniques such as linear interpolation and spline interpolation can be used for this interpolation processing. In particular, as the spline interpolation, for example, the method described in “Scattered Data Interpolation with Multilevel B-Splines”, IEEE Transactions on Visualization and Computer Graphics, vol. 3, no. 3, July-September 1997 by Seungyong Lee et al. be able to.

  At this time, in the first image Va, a movement amount is set for each pixel constituting a partial image composed of a set of tomographic images representing a plurality of parts, that is, an image of a portion where the partial image Ra3 and the partial image Ra4 overlap in this embodiment. The calculation unit 15 obtains a value obtained by averaging the movement amounts specified for each of the plurality of parts represented in the partial image, and gives the obtained average value as the movement amount of the pixel.

  The positional relationship calculation unit 16 gives the movement amount to each of all the pixels in the first image Va as described above, and further calculates the movement amount using a different deformation method and / or a different evaluation scale. For each pixel in the portion where the obtained partial images are adjacent to each other, the movement amount of the given pixel is averaged using the movement amount in each of two or more pixels located in the vicinity of the pixel, and the averaged This value is given again as the movement amount of the pixel.

  The image generation unit 17 generates a deformed image Va ′ by deforming the first image Va using the corresponding positional relationship obtained by the positional relationship calculation unit 16. The display control unit 18 causes the display device 3 to display the image Va ′ and the second image Vb generated by the image generation unit 17 so that they can be compared.

  FIG. 2 is a flowchart showing the operation of the image generation apparatus 1. As illustrated, first, the image acquisition unit 12 acquires the first image Va and the second image Vb stored in the storage device 11 (S1), and the part information acquisition unit 13 acquires the first image Va and the first image Vb. The site | part information memorize | stored in the memory | storage device 11 is acquired corresponding to each of the tomographic image which comprises 2 image Vb (S2). Next, the partial image specifying unit 14 specifies a plurality of sets of partial images including sets of tomographic images representing the same part in the first image Va and the second image Vb using the part information acquired in step S2. (S3). Here, the total number of the set of identified partial images is n.

  Next, the movement amount calculation unit 15 sets the count i to an initial value (S4), refers to the table T1, and stores the deformation method stored in association with the parts represented in the partial images Rai and Rbi. And an evaluation scale are specified, and an arbitrary movement amount m (Rai) is set for the partial image Rai using the specified deformation method (S5). Subsequently, the movement amount calculation unit 15 deforms the partial image Rai by the movement amount m (Rai) set in step S5 (S6), and the partial image Rai ′ after the deformation and the partial image Rbi of the second image Vb The evaluation value of the similarity is evaluated using the evaluation scale specified in step S5 (S7), and whether the change amount of the evaluation value with respect to the change of the movement amount m (Rai) at that time is equal to or smaller than a predetermined value is determined. Judgment is made (S8). If step S8 is negative, the movement amount m (Rai) is changed (S9), and the process returns to step S6. On the other hand, if step S8 is affirmed, it is further determined whether or not the count i has reached n (S10). If step S10 is negative, 1 is added to i (S11), and the process returns to step S5.

  On the other hand, when step S10 is affirmed, the positional relationship calculation unit 16 calculates the movement amount m (Rai) obtained by the movement amount calculation unit 15 for each partial image Rai and Rbi (i = 1 to n). Using this, the corresponding positional relationship of all the pixels between the first image Va and the second image Vb is obtained (S12). Next, the image generation unit 17 deforms the first image Va using the corresponding positional relationship obtained by the positional relationship calculation unit 16 to generate a deformed image Va ′ (S13), and the display control unit 18 However, the image Va ′ and the second image Vb generated by the image generation unit 17 are displayed on the display device 3 so that they can be compared (S14), and the process ends.

  As described above, according to the above embodiment, when the corresponding positional relationship is obtained between two three-dimensional images of the same subject to be compared, an image is obtained for each of a plurality of parts constituting the subject. The deformation method used when deforming and the evaluation scale used when evaluating the similarity between two images are stored in association with each other, and the same part specified in the two three-dimensional images is represented. For each set of partial images, a partial image of one three-dimensional image is converted into a partial image of the other three-dimensional image by using a deformation method and an evaluation scale stored in association with the parts represented in the partial images. Since the movement amount of each pixel at the time of deformation is obtained and the corresponding positional relationship between the two three-dimensional images is obtained using the obtained movement amount, the deformation suitable for each of a plurality of parts is obtained. Using methods and rating scales It is possible to obtain the corresponding positional relationship between the image portion between representing the site, it is possible to determine the exact correspondence positional relationship by overall between the two three-dimensional images.

  In the above embodiment, the part information acquisition unit 13 acquires the information from the storage device 11 on the assumption that the part information is acquired in advance in a separate process and stored in the storage device 11. However, the part information acquisition unit 13 directly executes the part recognition process, and each tomographic image forming the first image Va and the second image Vb represents any part of the subject. It is also possible to acquire information on whether or not.

  Further, in the above embodiment, for each pixel of the partial image composed of a set of tomographic images representing a plurality of parts, the movement amount calculation unit 15 applies the part to each part represented in the partial image. The movement amount is obtained using the deformation method and the evaluation scale that are associated with each other, and the positional relationship calculation unit 16 obtains the corresponding positional relationship between the images by using the average value of the plurality of obtained movement amounts. Although the movement amount calculation unit 15 has been described with reference to a case where the movement amount is calculated, the movement amount calculation unit 15 stores the deformation method and evaluation stored in association with any one part selected from a plurality of parts represented in the partial image. The movement amount is obtained using a scale, and the positional relationship calculation unit 16 may obtain a corresponding positional relationship between images using the obtained one movement amount.

  At this time, in the selection of any one part, when priority is given to those parts, the part with the higher priority may be automatically selected, or the most recent by the user It is also possible to determine one part that is presumed to receive the most attention based on the input operation and to select the determined part. As specific examples of the most recent input operation by such a user, an operation for designating an arbitrary point in an area representing any one part, or an operation for displaying an image mainly representing any one part Etc.

  Moreover, although the case where the image generation apparatus 1 is provided with the image generation part 17 and the display control part 18 was demonstrated in the said embodiment, those structures are not necessarily required and it is good to provide as needed. .

  The second embodiment of the present invention will be described below with reference to the drawings. FIG. 4 is a block diagram illustrating a schematic configuration of the image processing apparatus 100. The configuration of the image processing apparatus 100 as shown in FIG. 4 is realized by executing an image processing program read into the auxiliary storage device on the computer. At this time, the image processing program is stored in a storage medium such as a CD-ROM or distributed via a network such as the Internet and installed in a computer.

  This image processing program defines an image acquisition process, a part image specifying process, a movement amount calculating process, a positional relationship calculating process, an image generating process, and a display control process as processes to be executed by the CPU of the computer. In accordance with this rule, when the CPU executes each of the above processes, the computer acquires an image acquisition unit 112, a part image specifying unit 114, a movement amount calculation unit 115, and a positional relationship calculation unit 116, which will be described later. , Function as an image generation unit 117 and a display control unit 118.

  The image processing apparatus 100 also includes a storage device 111 such as a hard disk drive, and this storage device 111 corresponds to the storage unit of the present invention. Further, a display device 3 such as a display and an input device 4 such as a mouse and a keyboard are connected to the image processing apparatus 100.

  In the storage device 111, an image Vc (hereinafter, referred to as “first image Vc”) obtained by photographing a predetermined subject using an imaging device such as CT, MRI, PET, SPECT, or an ultrasonic image in the latest examination. In addition, an image Vd (hereinafter referred to as “second image Vd”) obtained by photographing the same subject by any of the above-described photographing apparatuses in the past examination is stored. The first image Vc and the second image Vd may be images taken with the same modality, or may be images taken with different modalities.

  Also, in the storage device 111, a deformation technique used when deforming an image is associated with each of a plurality of parts constituting the subject and an evaluation scale used when evaluating the similarity between the two images. Table is stored. FIG. 8 is a diagram showing an example of a table used in this embodiment. As shown in FIG. 8, in the table T2, a deformation method and an evaluation scale are associated with each of a plurality of parts. Specifically, the head has rigid registration and correlation coefficients, the lung has non-rigid registration and the square of the pixel value difference, the skin has surface registration and the least square error between the surfaces, the aorta Is the tree structure matching and the distance between corresponding points, the liver is non-rigid registration and mutual information, the heart is non-rigid registration and mutual information, and bones (individually divided) The rigid registration is associated with the square of the difference between the prime values, and the large intestine is associated with the structure matching and the distance between corresponding points.

  Here, in the structure matching, a graph structure (a graph structure having a tree structure in the case of tree structure matching) is extracted from each of the target image and the reference image, and corresponding points on the extracted graph structures are extracted from each other. This is a method of changing the position of each pixel constituting the target image so as to overlap. Refer to the description of the first embodiment for each deformation technique and evaluation scale exemplified in the other table T2.

  The image acquisition unit 112 acquires the first image Vc and the second image Vd stored in the storage device 111. The partial image specifying unit 114 specifies a plurality of sets of part images in which the same part is represented in the first image Vc and the second image Vd acquired by the image acquisition unit 112. For example, for each part constituting the subject, an area in which the part is expressed is extracted from each of the first image Vc and the second image Vd, and the extracted area is specified as a set of one part image. . Here, the region in which a specific part is represented refers to a region that includes the entire region representing the part and does not include any other region. It should be noted that, in the process of extracting the region where each part is represented from the target three-dimensional image (the first image Vc or the second image Vd), according to the part so that an optimum extraction result can be obtained. The method may be changed as appropriate.

  For example, when the region is the liver, the region can be extracted by the method described below. First, an arbitrary point in the liver region is set according to the user's operation of a pointing device or the like with respect to the cross-sectional image showing the liver (hereinafter, this point is referred to as “user set point”). Next, using a discriminator acquired by machine learning such as the AdaBoost method, a corner portion of the contour of the liver region is detected as a reference point. Further, identification obtained by machine learning such as the AdaBoost method for each point (voxel) in a three-dimensional region (hereinafter referred to as “processing target region”) having a size including the liver around the user set point. An evaluation value indicating whether the point is on the outline of the liver is calculated using a container. Then, each point on the outer periphery of the processing target region is determined in advance as a point in the background region outside the liver region, and the user set point and the reference point are determined in advance as points in the liver region, and further processing is performed. The liver region is extracted from the three-dimensional image of the target by applying the graph cut method using the evaluation value of each point in the target region.

  Further, when the part is a blood vessel, the region can be extracted by the method described below. First, a linear structure is searched by calculating eigenvalues of a 3 × 3 Hessian matrix for each local region in the target three-dimensional image. In the region including the linear structure, one of the three eigenvalues of the Hessian matrix is a value close to 0, and the other two are relatively large values. The eigenvector corresponding to the eigenvalue whose value is close to 0 indicates the principal axis direction of the linear structure. The area specifying unit 13 uses this relationship to determine the likelihood of a linear structure based on the eigenvalues of the Hessian matrix for each local area, and for the local area where the linear structure is identified, the center point is a candidate. Detect as a point. Then, the candidate points detected by the search are connected based on a predetermined algorithm. Thereby, a tree structure composed of candidate points and blood vessel branches (edges) connecting the candidate points is constructed. The coordinate information of the detected candidate points and vector information indicating the direction of the blood vessel branch are stored in the memory together with the candidate point and the identifier of the blood vessel branch. Subsequently, for each detected candidate point, the outline of the blood vessel (outer wall of the blood vessel) is identified in the cross section perpendicular to the blood vessel path based on the surrounding pixel values. The shape is identified using a known segmentation technique represented by Graph-Cuts. Through the above processing, a blood vessel region is extracted from the target three-dimensional image. The blood vessel region can also be extracted by using a threshold method, a Region Growing method, a Level Set method, and other various image processes.

  In general, as processing for recognizing a predetermined structure in a three-dimensional image, machine learning using an AdaBoost, a support vector machine (SVM), an adaptive vector machine (RVM), an artificial neural network (ANN), or the like is used. The resulting discriminating means can be used.

  FIG. 5 is an image diagram showing the first image Vc and the second image Vd. In the example shown in FIG. 5, in the first image Vc and the second image Vd, a set of regions Rc1 and Rd1 in which the head is represented, a set of regions Rc2 and Rd2 in which the lungs are represented, and the liver are represented. A set of the regions Rc3 and Rd3 that are displayed, and a set of the regions Rc4 and Rd4 that indicate the skin are specified.

  The movement amount calculation unit 115 refers to the table T2 as shown in FIG. 8 for each group of site images Rci and Rdi (i = 1 to 4) specified by the site image specifying unit 114, and sets the set of the group images. The deformation method stored in association with the region represented in the region image and the evaluation scale are specified, and each pixel in the region image Rci of the first image Vc is set to a predetermined value using the specified deformation method. The part image Rci is deformed by being moved by the movement amount, and the degree of similarity between the part image Rci ′ after the deformation and the part image Rdi of the second image Vd is evaluated using the specified evaluation scale. In this modification, the amount of change m (Xo) given to each pixel Xo (Xo∈Rci) is changed repeatedly, and the amount of change in the evaluation value that evaluates the degree of similarity with respect to the change in amount of movement is less than or equal to a predetermined value In case of deformation That the calculating the moving amount m given to each pixel (Xo). Here, the movement amount is m, a vector amount having a movement direction and a distance, and the deformed region image Rci ′ is formed by displacing each pixel Xo of the region image Rci by the movement amount m (Xo). This is an image. In the following description, each pixel Xo in the part image Rci is referred to as a movement amount m (Rci).

  Specifically, the movement amount calculation unit 115 changes the evaluation value change amount with respect to the change in the movement amount m (Rci) to a predetermined value while changing the movement amount m (Rci) in the direction in which the evaluation value increases. By repeatedly performing the comparison process, the movement amount m (Rci) when the change amount of the evaluation value is equal to or less than the predetermined value is specified. Here, as the predetermined value, an arbitrary value that can be regarded as a sufficiently large change amount of the evaluation value with respect to the change of the movement amount m (Rci) may be set.

  Instead of this, for example, an evaluation value obtained by evaluating the similarity between the partial image Rci ′ and the partial image Rdi under the setting of the movement amount m (Rci) while changing the movement amount m (Rci) is a predetermined value. The amount of movement m (Rci) when the evaluation value of the similarity is equal to or greater than a predetermined value may be specified by repeatedly performing the process of comparing with the above. Here, as the predetermined value, an arbitrary value that can be regarded as a sufficiently large evaluation value of the similarity may be set. In addition, for example, evaluation values obtained by evaluating the similarity between the part image Rci ′ and the part image Rdi are set for a plurality of different movement amounts m (Rci), and the evaluation value is the largest among the calculated evaluation values. The movement amount m (Rci) corresponding to the specified maximum evaluation value may be specified.

  The positional relationship calculation unit 116 uses the movement amount calculation unit 115 to calculate the first image Vc and the second image using the movement amount m (Rci) obtained for each group of part images Rci and Rdi (i = 1 to 4). The corresponding positional relationship of all pixels between Vd is obtained. Specifically, a vector field that gives a movement amount to each of all the pixels in the first image Vc is obtained. This vector field indicates that each pixel has a correspondence relationship with a pixel on the second image Vd at a position separated by the amount of movement given to that pixel.

  First, the movement amount m (Xo) determined by the movement amount m (Rci) specified by the movement amount calculation unit 115 is given to each pixel Xo of the part image Rci (i = 1 to 4), and the first image Two or more pixels located in the vicinity of the estimated value of the movement amount of each pixel in a part other than the part image Rci (i = 1 to 4) in Vc (for example, part Rcf shown in FIG. 5). Are obtained by interpolation using the amount of movement given to each of them, and the obtained estimated value is given as the amount of movement of the pixel. Various spatial interpolation techniques such as linear interpolation and spline interpolation can be used for this interpolation processing. In particular, as the spline interpolation, for example, the method described in “Scattered Data Interpolation with Multilevel B-Splines”, IEEE Transactions on Visualization and Computer Graphics, vol. 3, no. 3, July-September 1997 by Seungyong Lee et al. be able to.

  The positional relationship calculation unit 116 gives the movement amount to each of all the pixels in the first image Vc as described above, and further calculates the movement amount using a different deformation method and / or a different evaluation scale. For each pixel in a portion where the obtained part images are adjacent to each other, the movement amount of the given pixel is averaged using the movement amount in each of two or more pixels located in the vicinity of the pixel, and the averaged This value is given again as the movement amount of the pixel.

  The image generation unit 117 uses the corresponding positional relationship obtained by the positional relationship calculation unit 116 to deform the first image Vc and generate a deformed image Vc ′. The display control unit 118 causes the display device 3 to display the image Vc ′ and the second image Vd generated by the image generation unit 117 so that they can be compared.

  FIG. 6 is a flowchart showing the operation of the image generation apparatus 100. As shown in the figure, first, the image acquisition unit 112 acquires the first image Vc and the second image Vd stored in the storage device 111 (S101), and the part image specifying unit 114 detects the first image Vc and the first image Vc. A plurality of sets of part images representing the same part in the two images Vd are specified (S102). Here, it is assumed that the total number of sets of identified part images is n.

  Next, the movement amount calculation unit 115 sets the count i to an initial value (S103), refers to the table T2, and stores the deformation method stored in association with the parts represented in the part images Rci and Rdi. An evaluation scale is specified, and an arbitrary movement amount m (Rci) is set for the part image Rci using the specified deformation method (S104). Subsequently, the movement amount calculation unit 115 deforms the part image Rci by the movement amount m (Rci) set in step S105 (S105), the part image Rci ′ after the deformation, and the part image Rdi of the second image Vd. The evaluation value of the similarity is evaluated using the evaluation scale specified in step S104 (S106), and whether the change amount of the evaluation value with respect to the change of the movement amount m (Rci) at that time is equal to or less than a predetermined value is determined. Judgment is made (S107). If step S107 is negative, the movement amount m (Rci) is changed (S108), and the process returns to step S105. On the other hand, if step S107 is affirmed, it is further determined whether or not the count i has reached n (S109). If step S109 is negative, 1 is added to i (S110), and the process returns to step S104.

  On the other hand, when step S109 is affirmed, the positional relationship calculation unit 116 calculates the movement amount m (Rci) obtained by the movement amount calculation unit 115 for each group of part images Rci and Rdi (i = 1 to n). Using this, the corresponding positional relationship of all the pixels between the first image Vc and the second image Vd is obtained (S111). Next, the image generation unit 117 deforms the first image Vc using the corresponding positional relationship obtained by the positional relationship calculation unit 116 to generate a deformed image Vc ′ (S112), and the display control unit 118. However, the image Vc ′ and the second image Vd generated by the image generation unit 117 are displayed on the display device 3 so that they can be compared (S113), and the process ends.

  As described above, according to the above embodiment, when the corresponding positional relationship is obtained between two three-dimensional images of the same subject to be compared, an image is obtained for each of a plurality of parts constituting the subject. The deformation method used when deforming and the evaluation scale used when evaluating the similarity between two images are stored in association with each other, and the same part specified in the two three-dimensional images is represented. For each set of part images, a part image of one three-dimensional image is converted into a part image of the other three-dimensional image using a deformation method and an evaluation scale stored in association with the part represented in those part images. Since the movement amount of each pixel at the time of deformation is obtained and the corresponding positional relationship between the two three-dimensional images is obtained using the obtained movement amount, the deformation suitable for each of a plurality of parts is obtained. Using methods and rating scales It is possible to obtain the corresponding positional relationship between the image portion between representing the site, it is possible to determine the exact correspondence positional relationship by overall between the two three-dimensional images.

  In the first embodiment, the plurality of parts constituting the subject used for the processing of the part information acquisition unit 13, the partial image specification unit 14, and the movement amount calculation unit 15 are the head, neck, chest, The liver and the lower abdomen are four parts. In the second embodiment, a plurality of parts constituting the subject used in the processing of the part image specifying unit 114 and the movement amount calculating unit 115 are the head, lungs, Although the case where there are four parts of the liver and skin has been illustrated, in any of the embodiments, the part used for the treatment can be appropriately selected as necessary. For example, in the region where the bone and liver are in contact with each other, there may be a case where there is actually a discontinuous deformation. Can be prevented.

  In the above embodiment, the case where the image generation apparatus 100 includes the image generation unit 117 and the display control unit 118 has been described. However, these configurations are not necessarily required, and may be provided as necessary. .

1,100 Image generation device 3 Display device 4 Input device 11, 111 Storage device 12, 112 Image acquisition unit 13 Part information acquisition unit 14, 114 Partial image specification unit 15, 115 Movement amount calculation unit 16, 116 Position relation calculation unit 17 , 117 Image generation unit 18, 118 Display control unit Va, Vc First image Vb, Vd First image Ra1-Ra5 Partial image Rb1-Rb5 Partial image Rc1-Rc4 Partial image Rd1-Rd4 Partial image

Claims (14)

An image processing apparatus for obtaining a corresponding positional relationship between two three-dimensional images of the same subject to be compared,
A storage unit that stores a deformation technique used when deforming an image and an evaluation scale used when evaluating the similarity between two images in association with each of a plurality of parts constituting the subject,
A site information acquisition unit that acquires information on which tomographic image represents which site of the subject each of the tomographic images constituting the two three-dimensional images;
Using the acquired information, a partial image specifying unit for specifying a plurality of sets of partial images composed of a set of tomographic images representing the same part in the two three-dimensional images;
For each identified partial image of each set, each pixel in the partial image of one of the three-dimensional images is assigned a predetermined value using a deformation method stored in association with the portion represented in the partial image. The partial image is deformed by being moved by the movement amount, and the similarity between the deformed partial image and the partial image of the other three-dimensional image is stored in association with the part represented in the partial image. The process of evaluating using the evaluated evaluation scale is repeated while changing the amount of movement given to each pixel in the deformation, and the amount of change in the evaluation value that evaluates the similarity to the change in the amount of movement is predetermined. A movement amount calculation unit for obtaining a movement amount given to each pixel in the deformation when the value is equal to or less than a value;
A positional relationship calculation unit for determining a corresponding positional relationship between the two three-dimensional images using the determined movement amount ;
In the one three-dimensional image, the movement amount calculation unit adds each pixel of a partial image including a set of tomographic images representing a plurality of parts to each part represented in the partial image. The amount of movement is calculated using the associated deformation method and evaluation scale,
The positional relationship calculation unit calculates a corresponding positional relationship between the two three-dimensional images by using a value obtained by averaging the movement amounts obtained for each part as the obtained movement amount. An image processing apparatus. The positional relationship calculation unit calculates the movement amount for each pixel in a part where the partial images obtained by using the different deformation methods and / or different evaluation scales in the one three-dimensional image are adjacent to each other. The movement amount of the pixel obtained by the unit is averaged using the movement amount in each of two or more pixels located in the vicinity of the pixel, and the averaged value is used as the obtained movement amount. the image processing apparatus according to claim 1, characterized in that to obtain the corresponding positional relationship between the two three-dimensional images. The positional relationship calculation unit obtains an estimated value of the movement amount of each pixel in a portion other than the specified partial image in the one three-dimensional image using the movement amount obtained by the movement amount calculation unit, the image processing apparatus according to claim 1, wherein a is intended to determine the correspondence position relationship between the two three-dimensional images using the estimated value and the movement amount obtained the determined said. An image processing apparatus for obtaining a corresponding positional relationship between two three-dimensional images of the same subject to be compared,
A storage unit that stores a deformation technique used when deforming an image and an evaluation scale used when evaluating the similarity between two images in association with each of a plurality of parts constituting the subject,
A part image specifying unit for specifying a plurality of sets of part images in which the same part is represented in the two three-dimensional images;
For each identified part image of each set, each pixel in the part image of one of the three-dimensional images is determined in advance using a deformation method stored in association with the part represented in those part images. The part image is deformed by being moved by the movement amount, and the similarity between the part image after the deformation and the part image of the other three-dimensional image is stored in association with the part represented in the part image. The process of evaluating using the evaluated evaluation scale is repeated while changing the amount of movement given to each pixel in the deformation, and the amount of change in the evaluation value that evaluates the similarity to the change in the amount of movement is predetermined. A movement amount calculation unit for obtaining a movement amount given to each pixel in the deformation when the value is equal to or less than a value;
Using the obtained movement amount, an estimated value of the movement amount of each pixel in a portion other than the specified part image in the one three-dimensional image is obtained, and the obtained estimated value and the obtained movement An image processing apparatus comprising: a positional relationship calculation unit that obtains a corresponding positional relationship between the two three-dimensional images using a quantity. In the one three-dimensional image, the positional relationship calculation unit calculates the movement amount for each pixel in a portion where the part images obtained by using different deformation methods and / or different evaluation scales are adjacent to each other. The movement amount of the pixel obtained by the unit is averaged using the movement amount in each of two or more pixels located in the vicinity of the pixel, and the averaged value is used as the obtained movement amount. 5. The image processing apparatus according to claim 4, wherein a corresponding positional relationship between the two three-dimensional images is obtained. The positional relationship calculation unit moves the estimated value of the movement amount of each pixel in a portion other than the specified part image in the one three-dimensional image in each of two or more pixels located in the vicinity of the pixel. 6. The image processing apparatus according to claim 4 , wherein the image processing apparatus is obtained by interpolation using a quantity. An image processing method for obtaining a corresponding positional relationship between two three-dimensional images of the same subject to be compared,
For each of the plurality of parts constituting the subject, a deformation method used when deforming the image and an evaluation scale used when evaluating the similarity between the two images are stored in association with each other.
Obtaining information as to which tomographic image represents which part of the subject each of the tomographic images constituting the two three-dimensional images;
Using the acquired information, in the two three-dimensional images, specify a plurality of sets of partial images consisting of a set of tomographic images representing the same part,
For each identified partial image of each set, each pixel in the partial image of one of the three-dimensional images is assigned a predetermined value using a deformation method stored in association with the portion represented in the partial image. The partial image is deformed by being moved by the movement amount, and the similarity between the deformed partial image and the partial image of the other three-dimensional image is stored in association with the part represented in the partial image. The process of evaluating using the evaluated evaluation scale is repeated while changing the amount of movement given to each pixel in the deformation, and the amount of change in the evaluation value that evaluates the similarity to the change in the amount of movement is predetermined. Find the amount of movement given to each pixel in the deformation when it is less than or equal to the value,
An image processing method characterized by obtaining a corresponding positional relationship between the two three-dimensional images using the obtained movement amount ,
In the one three-dimensional image, the process for obtaining the movement amount is performed for each part of the partial image formed of a set of tomographic images representing a plurality of parts, for each part represented in the partial image. Is a process for obtaining the amount of movement using a deformation method and an evaluation scale associated with
The process for obtaining the positional relationship is a process for obtaining a corresponding positional relationship between the two three-dimensional images by using a value obtained by averaging the movement amounts obtained for each part as the obtained movement amount. A featured image processing method . The processing for obtaining the positional relationship is performed for each pixel in a portion where the partial images obtained by using the different deformation methods and / or different evaluation scales in the one three-dimensional image are adjacent to each other. Further, the movement amount of the pixel is averaged using the movement amount in each of two or more pixels located in the vicinity of the pixel, and the average value is used as the obtained movement amount. The image processing method according to claim 7 , wherein the image processing method is a process for obtaining a corresponding positional relationship between two-dimensional images. The processing for obtaining the positional relationship obtains an estimated value of the amount of movement of each pixel in a portion other than the specified partial image in the one three-dimensional image using the obtained amount of movement. 9. The image processing method according to claim 7, wherein the image processing method is a process of obtaining a corresponding positional relationship between the two three-dimensional images using an estimated value and the obtained movement amount. An image processing method for obtaining a corresponding positional relationship between two three-dimensional images of the same subject to be compared,
For each of the plurality of parts constituting the subject, a deformation method used when deforming the image and an evaluation scale used when evaluating the similarity between the two images are stored in association with each other.
In the two three-dimensional images, specify a plurality of sets of part images in which the same part is represented,
For each identified part image of each set, each pixel in the part image of one of the three-dimensional images is determined in advance using a deformation method stored in association with the part represented in those part images. The part image is deformed by being moved by the movement amount, and the similarity between the part image after the deformation and the part image of the other three-dimensional image is stored in association with the part represented in the part image. The process of evaluating using the evaluated evaluation scale is repeated while changing the amount of movement given to each pixel in the deformation, and the amount of change in the evaluation value that evaluates the similarity to the change in the amount of movement is predetermined. Find the amount of movement given to each pixel in the deformation when it is less than or equal to the value,
Using the obtained movement amount, an estimated value of the movement amount of each pixel in a portion other than the specified part image in the one three-dimensional image is obtained, and the obtained estimated value and the obtained movement An image processing method characterized in that a corresponding positional relationship between the two three-dimensional images is obtained using a quantity. In the processing for obtaining the positional relationship, in the one three-dimensional image, for each pixel in a portion where the part images obtained by using different deformation methods and / or different evaluation scales are adjacent to each other, the above-described calculation is performed. Further, the movement amount of the pixel is averaged using the movement amount in each of two or more pixels located in the vicinity of the pixel, and the average value is used as the obtained movement amount. The image processing method according to claim 10 , wherein the image processing method is a process of obtaining a corresponding positional relationship between two-dimensional images. In the processing for obtaining the positional relationship, an estimated value of the amount of movement of each pixel in a portion other than the specified part image in the one three-dimensional image is obtained in each of two or more pixels located in the vicinity of the pixel. 12. The image processing method according to claim 10, wherein the image processing method is obtained by interpolation using a movement amount. A program for causing a computer to function as an image processing device for obtaining a corresponding positional relationship between two three-dimensional images of the same subject to be compared,
The computer,
A storage unit that stores a deformation technique used when deforming an image and an evaluation scale used when evaluating the similarity of two images in association with each of a plurality of parts constituting the subject,
A site information acquisition unit that acquires information about which of the tomographic images each of the tomographic images represents each of the tomographic images constituting the two three-dimensional images;
Using the acquired information, in the two three-dimensional images, a partial image specifying unit that specifies a plurality of sets of partial images composed of a set of tomographic images representing the same part,
For each identified partial image of each set, each pixel in the partial image of one of the three-dimensional images is assigned a predetermined value using a deformation method stored in association with the portion represented in the partial image. The partial image is deformed by being moved by the movement amount, and the similarity between the deformed partial image and the partial image of the other three-dimensional image is stored in association with the part represented in the partial image. The process of evaluating using the evaluated evaluation scale is repeated while changing the amount of movement given to each pixel in the deformation, and the amount of change in the evaluation value that evaluates the similarity to the change in the amount of movement is predetermined. A movement amount calculation unit for obtaining a movement amount given to each pixel in the deformation when the value is equal to or less than a value;
An image processing program for functioning as a positional relationship calculation unit for determining a corresponding positional relationship between the two three-dimensional images using the determined movement amount ,
In the one three-dimensional image, the movement amount calculation unit adds each pixel of a partial image including a set of tomographic images representing a plurality of parts to each part represented in the partial image. The amount of movement is calculated using the associated deformation method and evaluation scale,
The positional relationship calculation unit calculates a corresponding positional relationship between the two three-dimensional images by using a value obtained by averaging the movement amounts obtained for each part as the obtained movement amount. An image processing program . A program for causing a computer to function as an image processing device for obtaining a corresponding positional relationship between two three-dimensional images of the same subject to be compared,
The computer,
A storage unit that stores a deformation technique used when deforming an image and an evaluation scale used when evaluating the similarity of two images in association with each of a plurality of parts constituting the subject,
A part image specifying unit for specifying a plurality of sets of part images in which the same part is represented in the two three-dimensional images;
For each identified part image of each set, each pixel in the part image of one of the three-dimensional images is determined in advance using a deformation method stored in association with the part represented in those part images. The part image is deformed by being moved by the movement amount, and the similarity between the part image after the deformation and the part image of the other three-dimensional image is stored in association with the part represented in the part image. The process of evaluating using the evaluated evaluation scale is repeated while changing the amount of movement given to each pixel in the deformation, and the amount of change in the evaluation value that evaluates the similarity to the change in the amount of movement is predetermined. A movement amount calculation unit for obtaining a movement amount given to each pixel in the deformation when the value is equal to or less than a value;
Using the obtained movement amount, an estimated value of the movement amount of each pixel in a portion other than the specified part image in the one three-dimensional image is obtained, and the obtained estimated value and the obtained movement The image processing program for functioning as a positional relationship calculation part which calculates | requires the corresponding positional relationship between said two three-dimensional images using quantity.