JP2005198887A - Method, apparatus and program for detecting anatomical structure, structure removal picture generation apparatus, and abnormal shadow detector - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To speedily detect the whole of the normal structure of a prescribed structure constituting a subject in a medical picture. <P>SOLUTION: A partial structure information extraction means 20 extracts partial structure information F part expressing the structure of a part of a prescribed structure in the medical picture P. A structure change vector setting means 40 sets a plurality of structure change vectors Ms based on the information F part. A structure model data creation means 50 creates structure model data M by adding the plurality of structure change vectors Ms to structure data Mb expressing the normal structure of the same kind of structure of the prescribed structure. A whole structure information obtaining means 90 obtains the structure model data M as whole structure information F all expressing the whole structure of the prescribed structure. The plurality of structure change vectors Ms is calculated so that a structure corresponding to the part of the structure expressed by the structure model data M may coincide with the structure of the part. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a method and apparatus for detecting the structure of an anatomical structure represented by a digital medical image, a program therefor, a structure removal image generation apparatus to which the above method is applied, and an abnormal shadow detection apparatus.

  Conventionally, in the medical field, a digital medical image (hereinafter simply referred to as a medical image) is image-analyzed using a computer for the purpose of reducing the load on an image reader such as a doctor in image diagnosis. (Computer Aided Diagnosis) processing is performed, and as such CAD processing, for example, abnormal shadow detection processing for automatically detecting an abnormal shadow in a medical image is known.

  The abnormal shadow detection process is a process of calculating an index value or the like representing the feature of the abnormal shadow on the image based on the medical image data and detecting the abnormal shadow based on the index value or the like. Therefore, an index value similar to an abnormal shadow may be calculated not only in an abnormal shadow but also in an area having image characteristics similar to the abnormal shadow. It may be detected by mistake. For example, in the case of diagnosis in a chest image, there are many cases where a place where a normal structure such as a rib overlaps is erroneously detected as a nodule shadow or a tumor shadow. The presence of such a misdetected normal location, so-called FP (False Positive), imposes a burden on the interpreter and may deteriorate the accuracy of diagnosis. Therefore, reduction of FP is strongly desired.

  For this reason, as described above, when a normal structure constituting a subject in a medical image used for image diagnosis affects the accuracy of abnormal shadow detection, such a structure is detected and the structure is detected. It is necessary to change the abnormal shadow detection process on the object, determine the abnormal shadow candidate detected on the structure with another index, or remove the structure from the medical image. Become.

Therefore, various methods for detecting the structure of such a structure or generating an image from which the structure is removed have been proposed. For example, there is a method for detecting a rib part in a medical image. Examples of such a method include a method for detecting a rib shape by an edge extraction filter (see, for example, Non-Patent Document 1), a parabola detection method by Hough transform, and the like. In addition, as a method of removing structures that hinder the detection of abnormal shadows from medical images, the ribs and lung field textures that match the chest images based on the chest images that are diagnosis targets are used for these structures. One image obtained by searching using a deformation model representing an object, and subtracting the obtained rib component from the chest image to remove the rib and showing only the soft part of the lung field A method for generating a chest image of the subject has been studied by the present applicant.
Peter de Souza, “Automatic Rib Detection in Chest Radiographs”, Computer Vision, Graphics and image Processing 23, 129-161 (1983)

  However, in the method of detecting the rib shape using the edge extraction filter described above, a structure can be detected in a region having a relatively simple structure such as the central part of the lung field. In a region where the structure is relatively complicated, such as the outside of the lung field where the ribs overlap, it is difficult to detect the structure. In addition, in the method of searching for the rib and lung soft part texture that matches the above chest image, and subtracting the rib component from the chest image to generate the rib removal image, the texture of the rib and lung soft part is searched for by hand search. There is a problem that the calculation time is enormous because it is necessary.

  In view of the above circumstances, the present invention provides a structure detection method and apparatus for an anatomical structure capable of detecting the entire structure of a predetermined structure constituting a subject in a medical image in a short time, and therefore And a structure removal image generation device that generates structure removal image data representing a medical image from which a predetermined structure has been removed, to which the structure detection method is applied, and an abnormality in the structure removal image It is an object of the present invention to provide an abnormal shadow detection apparatus that detects a shadow candidate.

  According to the first anatomical structure detection method of the present invention, image data carrying a medical image representing a subject composed of a plurality of anatomical structures is acquired, and the medical data is obtained based on the image data. In the image, partial structure information representing a structure of a part of a predetermined structure is extracted from the plurality of structures, and a structure representing the structure of the same kind of structure as the predetermined structure is extracted based on the partial structure information A structure model data is created by adding a plurality of structure change vectors for changing the structure to the data, and the structure model data is acquired as overall structure information representing the overall structure of the predetermined structure. A structure detection method for a physical structure, wherein the plurality of structure change vectors are set so that a structure corresponding to the part of the structure represented by the structure model data matches the structure of the part. Is a method which is characterized in.

  According to the second anatomical structure detection method of the present invention, image data carrying a medical image representing a subject composed of a plurality of anatomical structures is acquired, and the medical data is obtained based on the image data. In the image, partial structure information representing a partial structure of a predetermined structure is extracted from the plurality of structures, and partial structure information representing a partial structure of the same kind of structure as the predetermined structure is set. The partial structure information extracted is set for the first time as the partial structure information, and a part of the structure represented by the information previously set by the process is changed after the second time. And a plurality of structure change vectors for changing the structure are added to the structure data representing the structure of the same structure as the predetermined structure based on the partial structure information. The structure model data obtained by The plurality of structure change vectors are set so that the structure corresponding to the part of the structure matches the structure of the part, and the structure data is added to the structure data by adding the plurality of structure change vectors. Based on a process of creating model data and predetermined information obtained from the image data and including information indicating the structure of the predetermined structure, the structure represented by the structural model data and the predetermined in the medical image A process of calculating an index value representing the degree of difference from the structure of the structure, and a process of determining whether the difference is within a predetermined range based on the index value Is repeated until it is determined to be within a predetermined range, and the structural model data created last by the process of creating the structural model data is the whole representing the overall structure of the predetermined structure It is a method and acquiring as concrete information.

  According to the first anatomical structure detection apparatus of the present invention, there is provided an image data acquisition means for acquiring image data carrying a medical image representing a subject composed of a plurality of anatomical structures, Based on the partial structure information, a partial structure information extracting unit that extracts partial structure information representing a structure of a part of the predetermined structure among the plurality of structures in the medical image, and the predetermined structure based on the partial structure information A structure corresponding to the part of the structure represented by the structure model data obtained by adding a plurality of structure change vectors for changing the structure to the structure data representing the structure of the structure of the same kind as the object Structure change vector setting means for setting the plurality of structure change vectors so as to match the structure, and the structure model data by adding the plurality of structure change vectors to the structure data It is characterized by comprising structural model data creating means for creating and overall structure information obtaining means for obtaining the structural model data as overall structure information representing the overall structure of the predetermined structure. .

  According to the second anatomical structure detection apparatus of the present invention, there is provided an image data acquisition means for acquiring medical data representing a medical image representing a subject composed of a plurality of anatomical structures, and the image data On the basis of the above, in the medical image, partial structure information extracting means for extracting partial structure information representing a structure of a part of the predetermined structure among the plurality of structures, and a structure of the same kind as the predetermined structure Means for setting partial structure information representing the structure of a part, wherein the partial structure information extracted is set for the first time as the partial structure information, and set last time by the means after the second time. Partial structure information setting means for setting information representing a structure obtained by changing a part of the structure represented by the information, and structure data representing the structure of a structure of the same type as the predetermined structure based on the partial structure information To change the structure to A structure change in which the plurality of structure change vectors are set so that a structure corresponding to the part of the structure represented by the structure model data obtained by adding a number of structure change vectors matches the part of the structure. Information indicating the structure of the predetermined structure obtained from the image data, vector setting means, structural model data creating means for creating the structural model data by adding the plurality of structural change vectors to the structural data Index value calculating means for calculating an index value representing the degree of difference between the structure represented by the structural model data and the structure of the predetermined structure in the medical image based on predetermined information containing Based on the determination means for determining whether or not the difference is within a predetermined range, the setting of the partial structure information by the partial structure information setting means, and the structure change base. Setting the structural change vector by the torque setting means, creating the structural model data by the structural model data creating means, calculating the index value by the index value calculating means, and determining by the determining means, The repetitive control means that repeats until it is determined to be within a predetermined range, and the entire structural information that represents the overall structure of the predetermined structure, the structural model data created last by the structural model data creating means As a whole structure information acquisition means.

  As the “medical image”, for example, a radiation image, a CT image, an MRI image, and the like can be considered, and as the “subject”, the chest, abdomen, limbs, and the like of an animal or a human body can be considered.

  As the “anatomical structure”, for example, a bone part, a soft tissue, and the like can be considered.

  The “predetermined structure” may be a single type of structure or an aggregate composed of a plurality of types of structures.

  As the “part of the predetermined structure”, for example, when the medical image is a chest image and the predetermined structure is a rib, a lung field having a simple structure with little overlap between the ribs and other structures. The ribs in the central region can be considered. For example, when the medical image is a chest image and the predetermined structure is a lung field soft part, the lung field soft part from which the rib part has been removed can be considered. it can.

  “Predetermined information” does not need to be able to specify the structure of a predetermined structure only by the information, but includes information representing the structure of the predetermined structure and is a comparison target with the structure represented by the structural model data. Anything can be obtained. As such “predetermined information”, an image obtained by performing edge extraction filter processing on a medical image, an outline of a structure being emphasized, a medical image itself, and the like can be considered.

  In the structure detection apparatus for the first and second anatomical structures, the structure data is an average structure when a large number of different data representing structures of the same kind of structure as the predetermined structure are used as samples. The plurality of structural change vectors are proportional to respective eigenvectors obtained when principal component analysis is performed using a large number of different data representing the structure of the same kind of structure as the predetermined structure as teacher data. It may be a plurality of vectors.

  “Eigenvector” means a vector that can be considered mathematically linearly independent.

  As a method of “principal component analysis”, a method described in the literature [Matthew Turk, Alex Pentland, Eigenfaces for Recognition, Journal of Cognitive Neuroscience, vol. 3, num1, 1991] can be used.

  The structure may be at least one of a shape and a texture.

  The subject may be a chest, and the predetermined structure may be a soft part of a lung field or a bone part including a rib.

  The structure removal image generation apparatus of the present invention includes an image data acquisition unit that acquires image data carrying a medical image representing a subject composed of a plurality of anatomical structures, and the medical image based on the image data. And a partial structure information extracting means for extracting partial structure information representing a partial structure of a predetermined structure among the plurality of structures, and a partial structure information of the same kind of structure as the predetermined structure. Means for setting partial structure information, wherein the partial structure information extracted is set for the first time as the partial structure information, and the structure represented by the information previously set by the means is set for the second time and thereafter. A partial structure information setting means for setting information representing a structure whose part has been changed, and based on the partial structure information, the structure is changed to structure data representing a structure of the same kind of structure as the predetermined structure. Multiple structural changes for A structure change vector setting means for setting the plurality of structure change vectors so that a structure corresponding to the part of the structure represented by the structure model data obtained by adding vectors matches the part of the structure; , A structural model data creating means for creating the structural model data by adding the plurality of structural change vectors to the structural data, and a predetermined containing information indicating the structure of the predetermined structure obtained from the image data Based on the information, the index value calculating means for calculating the index value representing the degree of difference between the structure represented by the structural model data and the structure of the predetermined structure in the medical image, and based on the index value, Determining means for determining whether or not the difference is within a predetermined range; setting of partial structure information by the partial structure information setting means; The structural change vector is set by the structural model data creation by the structural model data creation means, the index value is calculated by the index value calculation means, and the determination by the determination means. Iterative control means that repeats until it is determined to be within the range, and the structural model data that was last created by the structural model data creation means is acquired as overall structure information that represents the overall structure of the predetermined structure. Structure-removed image data carrying a structure-removed image, which is a medical image representing the subject from which the predetermined structure has been removed, based on the whole structure information acquisition means, the image data, and the whole structure information And a structure removal image data generation means for generating the structure removal image data.

  The abnormal shadow detection apparatus of the present invention includes an image data acquisition unit that acquires medical data representing a medical image representing a subject composed of a plurality of anatomical structures, and the medical image based on the image data, Partial structure information extracting means for extracting partial structure information representing a structure of a part of a predetermined structure among the plurality of structures, and a partial structure representing a partial structure of a structure of the same type as the predetermined structure A means for setting information, wherein as the partial structure information, the extracted partial structure information is set for the first time, and after the second time, a part of the structure represented by the information previously set by the processing is set. Partial structure information setting means for setting information representing the changed structure, and based on the partial structure information, the structure is changed to structure data representing a normal structure of the same kind of structure as the predetermined structure. Multiple structural changes for A structure change vector setting means for setting the plurality of structure change vectors so that a structure corresponding to the part of the structure represented by the structure model data obtained by adding the spectra matches the part of the structure; , A structural model data creating means for creating the structural model data by adding the plurality of structural change vectors to the structural data, and a predetermined containing information indicating the structure of the predetermined structure obtained from the image data Based on the information, the index value calculating means for calculating the index value representing the degree of difference between the structure represented by the structural model data and the structure of the predetermined structure in the medical image, and based on the index value, Determination means for determining whether or not the difference is within a predetermined range; setting of partial structure information by the partial structure information setting means; and structure change vector setting means The structure change vector is set, the structure model data is created by the structure model data creation means, the index value is calculated by the index value calculation means, and the judgment is judged by the judgment means. Iterative control means that repeats until it is determined to be within the range, and the structural model data that was last created by the structural model data creation means is acquired as overall structure information that represents the overall structure of the predetermined structure. Structure-removed image data carrying a structure-removed image, which is a medical image representing the subject from which the predetermined structure has been removed, based on the whole structure information acquisition means, the image data, and the whole structure information A structure removal image data generating means to be generated, and a difference detection unit for detecting abnormal shadow candidates in the structure removal image based on the structure removal image data. It is characterized by comprising an ordinary shadow detecting means.

  According to a first program of the present invention, the computer uses the image data acquisition means for acquiring medical data representing a medical image representing a subject composed of a plurality of anatomical structures, and the medical data based on the image data. In the image, partial structure information extracting means for extracting partial structure information representing a structure of a part of the predetermined structure among the plurality of structures, and a structure of the same type as the predetermined structure based on the partial structure information The structure corresponding to the portion of the structure represented by the structure model data obtained by adding a plurality of structure change vectors for changing the structure to the structure data representing the structure of the object matches the structure of the portion. And a structure change vector setting means for setting the plurality of structure change vectors, and adding the plurality of structure change vectors to the structure data to obtain the structure model data. And structural model data creating means for forming, the structural model data, a program to function as a whole structural information acquisition means for acquiring the overall structure information indicating the overall structure of the predetermined structure.

  According to a second program of the present invention, the computer uses the image data acquisition means for acquiring the image data carrying the medical image representing the subject composed of a plurality of anatomical structures, and the medical data based on the image data. In the image, partial structure information extracting means for extracting partial structure information representing a partial structure of a predetermined structure among the plurality of structures, and representing a partial structure of the same kind of structure as the predetermined structure Means for setting partial structure information, wherein, as the partial structure information, the extracted partial structure information is set for the first time, and the structure information represented by the previous setting by the means is set for the second time or later. Partial structure information setting means for setting information representing a structure obtained by changing a part, and the structure data in the structure data representing the structure of the same kind of structure as the predetermined structure based on the partial structure information For changing A structure change in which the plurality of structure change vectors are set so that a structure corresponding to the part of the structure represented by the structure model data obtained by adding a number of structure change vectors matches the part of the structure. Information indicating the structure of the predetermined structure obtained from the image data, vector setting means, structural model data creating means for creating the structural model data by adding the plurality of structural change vectors to the structural data Index value calculating means for calculating an index value representing the degree of difference between the structure represented by the structural model data and the structure of the predetermined structure in the medical image based on predetermined information containing Based on the determination means for determining whether or not the difference is within a predetermined range, the setting of the partial structure information by the partial structure information setting means, and the structure change base. Setting the structural change vector by the torque setting means, creating the structural model data by the structural model data creating means, calculating the index value by the index value calculating means, and determining by the determining means, The repetitive control means that repeats until it is determined to be within a predetermined range, and the entire structural information that represents the overall structure of the predetermined structure, the structural model data created last by the structural model data creating means Is a program for functioning as overall structure information acquisition means.

  According to the first structure detection method and apparatus and the program therefor of the present invention, partial structure information representing a structure of a part of a predetermined structure in a medical image is extracted, and the predetermined structure is extracted based on the partial structure information. When the structure model data representing the structure of the same kind of structure is created by adding a plurality of structure change vectors to the structure data representing the structure of the structure of the same kind as the structure, The plurality of structure change vectors are set so that the structure corresponding to the part matches the structure of the part, and the created structure model data is used as overall structure information representing the overall structure of the predetermined structure. Since it is obtained, structural model data reflecting the structure of a part of a given structure can be obtained by a given calculation instead of groping, and the structure is complicated. The overall structure of the predetermined structure including a difficult detection part is the come of the edge extraction filter processing and the like, can be detected in a short time.

  According to the second structure detection method, apparatus, and program of the present invention, a structure of a part of a predetermined structure in a medical image is extracted, and a part of a structure of the same kind as the predetermined structure is represented. A process of setting partial structure information, the process of setting the partial structure while changing the extracted partial structure as an initial structure for each setting, and the same kind of structure as the predetermined structure The structure change is made so that the structure corresponding to the part of the structure represented by the structure model data obtained by adding a plurality of structure change vectors to the structure data representing the structure of the object matches the structure of part of the above part. In the process of setting the vector and the process of creating the structural model data by adding the structural change vector to the structural data, the structure represented by the created structural model data substantially matches the actual structure of the predetermined structure. The last obtained structural model data is acquired as information representing the overall structure of the predetermined structure, so that the partial structure of the predetermined structure actually extracted is reflected. The structure change vector can be searched using the structure model data as a starting point, and the entire structure of the predetermined structure can be expressed in a shorter time and with higher accuracy than when searching for the structure model data by hand. It can be expected to acquire the whole structure information.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing a schematic configuration according to an embodiment of the second structure detection apparatus of the present invention.

  The structure detection apparatus 1 shown in FIG. 1 has image data P carrying a medical image P representing a subject composed of a plurality of anatomical structures (hereinafter, the image and image data carrying the image are denoted by the same symbol). A partial structure for extracting partial structure information Fpart representing the structure of a part of a predetermined structure from among the plurality of structures in the medical image P based on the image data acquisition means 10 for acquiring the image data; Information extracting means 20 and means for setting partial structure information Mpart representing a partial structure of a structure of the same kind as the predetermined structure, and the first extraction is performed as the partial structure information Mpart. A partial structure information setting means 30 for setting the partial structure information Fpart set and setting information representing a structure obtained by changing a part of the structure represented by the previously set information after the second time, and a partial structure Based on the information Mpart Among the structures represented by the structure model data M obtained by adding a plurality of structure change vectors Ms for changing the structure to the structure data Mb representing the normal structure of the structure of the same type as the predetermined structure Structure change vector setting means 40 for setting a plurality of structure change vectors Ms so that the structure corresponding to “part” matches the “part of structure”, and the plurality of structure change vectors Ms as structure data Mb. The structural model data M is generated based on the predetermined information including the information indicating the structure of the predetermined structure obtained from the image data P and the structural model data generation means 50 for adding and generating the structural model data M. Index value calculation means 60 for calculating an index value d representing the degree of difference between the structure to be represented and the structure of the predetermined structure in the medical image P, and the difference is determined based on the index value d. The determination means 70 for determining whether or not it is within the range, the setting of the partial structure information Mpart by the partial structure information setting means 30, the setting of the structure change vector Ms by the structure change vector setting means 40, and the structure model data The creation unit 50 creates the structural model data M, the index value calculation unit 60 calculates the index value d, and the determination unit 70 determines that the difference is within a predetermined range. And repetitive control means 80 that repeats until the end, and structural model data M that is finally created by the structural model data creation means 50 is acquired as overall structure information Fall that represents the overall structure of the predetermined structure. 90.

  First, as a first embodiment, the above “medical image” is obtained by detecting a chest radiograph representing a chest including bones such as ribs and clavicles and a soft part of the lung (detecting radiation transmitted through the chest in the front-rear direction of the human body). A projected image of the chest), the “predetermined structure” as a rib, the “structure” as a shape, and the “structure of a portion of the predetermined structure” as a partial rib in the central lung field. Consider the case of shape. In other words, the overall shape of the rib is detected based on the partial shape of the rib in the center of the lung field where there is little overlap between the ribs and other structures and the accurate shape of the rib can be easily extracted. This is the case.

  Note that the overall shape of the rib is represented by predetermined n feature points that form the contour of the rib.

  In addition, an average shape vector Xb that represents the average shape of the ribs and an eigen shape that represents an eigen component (principal component) of the shape of the ribs by principal component analysis from rib shapes in chest images representing many normal breasts in advance. The vector Ai (i = 1, 2,...) Is calculated, and the average shape vector Xb is multiplied as the structure data Mb, and the eigen shape vector Ai is multiplied by the weight coefficient αi as the plurality of structure change vectors Ms. Let us use the vector αiAi.

  As the method of “principal component analysis”, the method described in the literature [Matthew Turk, Alex Pentland, Eigenfaces for Recognition, Journal of Cognitive Neuroscience, vol. 3, num1, 1991] is used.

  The operation of the structure detection apparatus 1 in this embodiment will be described below.

  FIG. 2 is a diagram illustrating a processing flow of the structure detection apparatus 1 in the present embodiment.

  The image data acquisition means 10 acquires image data P representing a chest radiographic image P generated by a CR (Computed Radiography) system or the like by external input (step S1). FIG. 3 is a diagram showing the input chest radiation image P. As shown in FIG.

  Based on the acquired image data P, the partial structure information extraction means 20 detects the rib shape at the center of the lung field in the chest radiographic image P as a partial rib shape by an edge extraction filter or the like, Partial structure information Fpart representing a simple shape is extracted and acquired (step S2). Here, the partial structure information Fpart is obtained from m feature points fxt (from t = 1) of n feature points fxj (j = 1, 2,..., N) forming the contour of the rib. m of values up to n). FIG. 4 is a diagram showing feature points fxt (points with a cross mark) on the chest radiation image P.

  The partial structure information setting unit 30 initially sets the partial structure information Fpart extracted by the partial structure information extraction unit 20, that is, the feature point fxt as the partial structure information Mpart (step S3).

  The structure change vector setting means 40 calculates and sets a plurality of structure change vectors Ms as follows (step S4).

The shape vector X representing the overall shape of the rib can be expressed as a formula (1) as a vector in which the coordinates (x, y) of the n feature points fxj are sequentially arranged.

Therefore, if the partial shape of the rib represented by the partial structure information Mpart constitutes a part of the entire shape of the rib in the chest radiographic image P, the rib shape vector Xinit in this initial state is It can be expressed as in equation (2).

Here, xp and yp are coordinates corresponding to each feature point represented by the partial structure information Mpart, and? Indicates coordinates corresponding to other unknown feature points. In order to make the shape vector Xinit easier to handle, a vector Zinit is defined as shown in Expression (3).

  Here, the matrix K is a matrix that takes the value 1 at the position of the feature point (diagonal component) that can be considered known and takes the value 0 at the position of the other unknown feature point. That is, the vector Zinit is a shape vector having a value at a position corresponding to a known feature point and having 0 (zero) at a position corresponding to an unknown feature point.

By the way, a shape vector X representing the shape of an arbitrary rib represents an average shape vector Xb representing an average shape of the rib calculated by principal component analysis from rib shapes in many chest images, and an eigen component of the shape of the rib. It can be expressed as a linear sum with the eigen shape vector Ai, and can be expressed as in Expression (4).

  Here, αi is a weighting coefficient for each eigen shape vector.

When the matrix K is multiplied to both sides of the equation (4), the equation (5) is obtained and the equation (5 ′) is obtained.

In general, even if {Ai} is linearly independent, {K · Ai} is not linearly independent, but the number of known feature points is smaller than the number of {K · Ai} vectors (value of i). If the number is sufficiently large, {K · Ai} may be roughly considered to be linearly independent. Therefore, the inner product of both sides of the equation (5 ′) and K · Ap (p = 1, 2,...) Is taken,

The weighting coefficient αp (p = 1, 2,...) Is obtained. This is a coefficient corresponding to the components of these vectors when Zinit-K · Xb is decomposed into each of the components of vectors K · A1, K · A2,. It means that the values of α1, α2,.

  Then, by multiplying the obtained coefficients α1, α2,... By the eigenshape vectors A1, A2,..., Weighted eigenshape vectors αiAi are obtained, which are set as the structure change vector Ms. can do.

  This is because the “part” of the rib shape represented by the structure model data M obtained by adding the weighted eigenshape vector αiAi as the structure modification vector Ms to the average shape vector Xb as the structure data Mb. Is equivalent to setting the structure change vector Ms so that the shape corresponding to “.” Matches the partial shape represented by the partial structure information Mpart.

  Then, the structural model data creation means 50 linearly combines the average shape vector Xb and the weighted eigen shape vector αiAi, and obtains the obtained vector as the structural model data M (step S5). That is, the shape vector X when the coefficients α1, α2,... Are substituted into the equation (4) can be used as the structural model data M.

  The index value calculation means 60 is based on predetermined information containing information indicating the shape of the ribs in the chest radiation image P obtained from the image data P, and the shape represented by the structural model data M and the ribs in the chest radiation image P. An index value d representing the degree of difference from the shape is calculated (step S6). Specifically, the following processing is performed.

  As the “predetermined information”, for example, an output image obtained by applying a Sobel filter sensitive to the edge to the chest radiation image P is used. The Sobel filter is a filter having sensitivity to an edge in a predetermined direction, and its output takes a value that is far from 0 (zero) in the vicinity of the edge, and takes a value close to 0 in other places. Therefore, in the output image of the Sobel filter, the position where the output value (pixel value) of the filter takes a large absolute value can be considered as the contour position of the anatomical structure including the ribs in the chest radiographic image P.

  Then, in the output image of this filter, for each position corresponding to the coordinates of the feature point represented by the structural model data M, the distance between the position and the contour position in the vicinity of the position is obtained, and the variance is determined as the structure. The index value d represents the degree of difference between the shape represented by the model data M and the shape of the rib in the chest radiation image P.

  If the calculated index value d is within a predetermined value, the determining unit 70 determines that the difference between the shape represented by the structural model data M and the rib shape in the chest radiation image P is within a predetermined range, When the index value d exceeds the predetermined value, it is determined that the difference is not within the predetermined range (step S7).

  In response to the determination by the determination means 70 that the difference is not within the predetermined range, the repetition control means 80 changes a part of the shape represented by the partial structure information Mpart to the partial structure information setting means 30. Is set as new partial structure information Mpart, and a series of processes from setting the partial structure information Mpart by the partial structure information setting unit 30 to determination by the determination unit 70 are repeatedly performed. In response to the determination by the determination means 70 that the difference is within the predetermined range, the repeated control is stopped. As a result, the structural change vector Ms is searched until structural model data M representing a shape approximate to the actual rib shape is found.

  Then, the overall structure information acquisition means 90 acquires the structure model data M finally created by the structure model data creation means 50 as overall structure information Fall representing the overall shape of the ribs in the chest radiation image P ( Step S8). FIG. 5 is a diagram showing the shape of the rib represented by the overall structure information Fall as feature points on the chest radiographic image P (points with black circle marks).

  Next, as a second embodiment, the “medical image” is a chest radiation image, the “predetermined structure” is a lung soft part, the “structure” is a texture, and the “part of the predetermined structure” Let's consider the case where the texture of the pulmonary soft part from which the rib region is removed is taken as “the structure of”. Consider the case of detecting the entire texture of a soft lung area in a chest radiograph. In other words, the purpose is to detect the entire texture of the lung soft part based on the partial texture of the lung soft part excluding the rib area, which is an area where the texture of the soft part of the lung soft field is difficult to understand. is there.

  Note that the texture shape must be standardized to a predetermined shape in order to enable the handling of texture comparison and synthesis, etc., but here, the shape of the texture of the lung soft part is changed to many normal breasts. The average shape of the lung field in the chest image representing.

  In addition, the entire texture of the lung field soft part is expressed by pixel values of n pixels forming the lung field soft part normalized to the average shape.

  In addition, in advance, in the chest image representing many normal breasts, the lung field soft part image representing each lung field soft part is warped by affine transformation or the like to this average shape, and from the texture of the lung field soft part image after warping, An average texture vector Gb representing an average texture in a lung field soft part normalized to an average shape by principal component analysis, and an inherent texture vector representing an inherent component (principal component) of a texture in a lung field soft part normalized to an average shape Bi (i = 1, 2,...), A vector obtained by multiplying the average texture vector Gb as the structure data Mb and the specific texture vector Bi as the plurality of structure change vectors Ms by the weight coefficient βi. Let us use βiBi.

  The operation of the structure detection apparatus 1 in this embodiment will be described below.

  FIG. 6 is a diagram showing a processing flow of the structure detection apparatus 1 in the present embodiment.

  The image data acquisition means 10 acquires image data P representing a chest radiation image P as shown in FIG. 3 generated by a CR (Computed Radiography) system or the like by an external input (step S11).

  In the first embodiment, the partial structure information acquisition unit 20 describes, based on the acquired image data P, the texture of the lung soft part from which the rib region in the chest radiation image P is removed as the texture of a part of the lung soft part. The partial structure information Fpart representing the partial texture is extracted and acquired based on the overall shape of the rib detected by the above method (step S12). Here, the partial structure information Fpart is set to m pixels fgt (from t = 1 to n) out of n pixels fgj (j = 1, 2,..., N) forming the entire lung soft part. Are acquired as pixel values at m). FIG. 7 is a diagram showing the texture of the soft lung area excluding the rib region in the chest radiographic image P.

  The partial structure information setting means 30 initially sets the partial structure information Fpart extracted by the partial structure information extraction means 20, that is, the pixel value in the pixel fgt as the partial structure information Mpart (step S13).

The structure change vector setting means 40 calculates and sets a plurality of structure change vectors Ms (step S14). Here, the texture vector G representing the entire texture of the lung field soft part is the pixel value Qj of each pixel fgj. Can be expressed as the following equation (7).

Therefore, if the partial texture of the lung field soft part represented by the partial structure information Mpart constitutes a part of the entire texture of the lung field soft part in the chest radiation image P, the lung field soft part in this initial state The texture vector Ginit can be expressed as shown in Equation (8).

Here, Qp is a pixel value in each pixel represented by the partial structure information Mpart, and? Indicates other unknown pixel values. In order to make this texture vector Ginit easier to handle, a vector Winit as shown in Expression (9) is defined.

  Here, the matrix U is a matrix that takes a value of 1 at pixel value positions (diagonal components) that can be considered known and a value of 0 at other unknown pixel value positions. That is, the vector Winit is a texture vector that has a value at a position corresponding to a known pixel value, and 0 (zero) at a position corresponding to an unknown pixel value.

By the way, the texture vector G that represents the texture of an arbitrary lung field soft part is calculated by principal component analysis from the texture of the lung field soft part in many chest images, and the average texture vector Gb that represents the average texture of the lung field soft part, It can be expressed by a linear sum with an eigentexture vector Bi that represents an eigencomponent of the texture of the field soft part, and can be expressed as in equation (10).

  Here, βi is a weighting coefficient for each unique texture vector.

When the matrix W is multiplied to both sides of the equation (10), the equation (11) is obtained and the equation (11 ′) is obtained.

In general, even if {Bi} is linearly independent, {U · Bi} is not linearly independent, but the known pixel value is smaller than the number of {U · Bi} vectors (i values). If the number is sufficiently large, roughly, {U · Bi} may be considered to be linearly independent. Therefore, the inner product of both sides of the equation (11 ′) and U · Bp (p = 1, 2,...) Is taken,

The weighting coefficient βp (p = 1, 2,...) Is obtained. This is a coefficient corresponding to the components of these vectors when Winit-U · Gb is decomposed into each of the components of vectors U · B1, U · B2,. It means that the values of β1, β2,.

  .. Are multiplied by the specific texture vectors B1, B2,... To obtain a weighted specific texture vector βiBi, which is set as the structure modification vector Ms. can do.

  This is because the average texture vector Gb as the structure data Mb and the weighted unique texture vector βiBi as the structure modification vector Ms are added to the above-mentioned “of the lung soft part texture represented by the structure model data M”. This corresponds to setting the structure change vector Ms so that the texture corresponding to “part” matches the part of texture represented by the partial structure information Mpart.

  Then, the structural model data creation means 50 linearly combines the average texture vector Gb and the weighted unique texture vector βiBi, and obtains the obtained vector as the structural model data M (step S15). That is, the texture vector G when the coefficients β1, β2,... Are substituted into the equation (10) can be used as the structural model data M.

  The index value calculation means 60 is based on predetermined information containing information indicating the texture of the lung soft region in the chest radiation image P obtained from the image data P, and the texture represented by the structural model data M and the chest radiation image P. An index value d representing the degree of difference from the texture of the lung field soft part is calculated (step S16). Specifically, the following processing is performed.

  As the “predetermined information”, for example, a lung field image representing a lung field whose shape is standardized in the chest radiation image P is used.

  Then, in the texture of the standardized lung field image and the texture represented by the structural model data M, a difference between the pixel values is obtained for each corresponding pixel, and the variance represented by the texture represented by the structural model data M and the chest radiation The index value d represents the degree of difference from the texture of the lung field soft part in the image P.

  If the calculated index value d is within a predetermined value, the determination unit 70 determines that the difference between the texture represented by the structural model data M and the texture of the soft lung area in the chest radiographic image P is within a predetermined range. When the index value d exceeds the predetermined value, it is determined that the difference is not within the predetermined range (step S17).

  In response to the determination by the determination unit 70 that the difference is not within the predetermined range, the repetition control unit 80 changes a part of the texture represented by the partial structure information Mpart to the partial structure information setting unit 30. The information representing the texture is set as new partial structure information Mpart, and a series of processing from setting of the partial structure information Mpart by the partial structure information setting unit 30 to determination by the determination unit 70 is repeatedly performed. In response to the determination by the determination means 70 that the difference is within the predetermined range, the repetitive control is stopped. As a result, the structural change vector Ms is searched until the structural model data M representing the texture approximated to the texture of the actual lung field soft part is found.

  Then, the entire structure information acquisition unit 90 acquires the structure model data M finally generated by the structure model data generation unit 50 as the entire structure information Fall representing the entire texture of the lung soft part in the chest radiation image P. (Step S18). FIG. 8 is a diagram showing the texture of the soft lung area represented by the overall structure information Fall obtained as described above.

  According to such a structure detection apparatus 1, a partial structure of a predetermined structure in a medical image is extracted, and partial structure information representing a partial structure of the same kind of structure as the predetermined structure is set. A process for setting the partial structure while changing the extracted partial structure as an initial structure for each setting, and a normal structure of the same type of structure as the predetermined structure. The structure change vector is set so that the structure corresponding to the part of the structure represented by the structure model data obtained by adding a plurality of structure change vectors to the structure data to be represented matches the structure of part of the part. Repeat the process and the process of adding the structure change vector to the structure data to create the structure model data until the structure represented by the created structure model data substantially matches the actual structure of the predetermined structure. In Since the obtained structural model data is acquired as information representing the overall structure of the predetermined structure, the structural model data reflecting the partial structure of the predetermined structure actually extracted is used as a starting point. In addition, the structure change vector can be searched, and overall structure information representing the entire structure of the predetermined structure can be obtained in a shorter time and with higher accuracy than when searching for the structure model data by hand. I can expect that.

  In the above embodiment, the structure change vector Ms is searched until the structure represented by the structure model data M approximates the actual structure of the predetermined structure in the medical image within a predetermined range. When priority is given to shortening the calculation time, or when it is not necessary to increase the accuracy of the overall structure information Fall obtained, the structure of a part of the extracted predetermined structure is searched without searching the structure change vector Ms. Based on the partial structure information Fpart to be represented, a structure corresponding to the “part” among the structures represented by the structure model data M obtained by adding a plurality of structure change vectors Ms to the structure data Mb is the “partial structure”. The structure model data M obtained by setting a plurality of structure change vectors Ms so as to match the above and adding the structure change vectors Ms to the structure data Mb is detected. The whole structure information Fall representing the entire structure of the predetermined structure may be acquired as it is. In this case, the first structure detection apparatus of the present invention can be used.

  FIG. 9 is a diagram showing a schematic configuration according to an embodiment of the first structure detection apparatus of the present invention.

  The structure detection apparatus 2 shown in FIG. 9 includes image data acquisition means 10, partial structure information extraction means 20, structure change vector setting means 40, structure model data creation means 50, and overall structure information acquisition means 90. ing.

  FIG. 10 is a diagram illustrating a processing flow of the structure detection apparatus 2.

  The image data acquisition means 10 acquires image data P carrying a medical image P representing a subject composed of a plurality of anatomical structures (step S21), and the partial structure information extraction means 20 is based on the image data P. Thus, in the medical image P, the partial structure information Fpart representing the structure of a part of the predetermined structure among the plurality of structures is extracted (step S22). The structure change vector setting means 40, based on the part structure information Fpart, outputs a plurality of structure change vectors Ms for changing the structure to structure data Mb representing the normal structure of the structure of the same type as the predetermined structure. A plurality of structure change vectors Ms are set so that the structure corresponding to the “part” in the structure represented by the structural model data M obtained by addition matches the “partial structure” (step S23). The structural model data creation means 50 creates the structural model data M by adding a plurality of structural change vectors Ms to the structural data Mb (step S24). Then, the overall structure information acquisition means 90 acquires the structure model data M created by the structure model data creation means 50 as overall structure information Fall representing the overall structure of the predetermined structure (step S25).

  According to the structure detection apparatus 2 as described above, partial structure information representing a partial structure of a predetermined structure in the medical image is extracted, and based on the partial structure information, a structure of the same type as the predetermined structure is extracted. A structure corresponding to the part of the structure represented by the structure model data when the structure model data representing the structure of the same kind of structure is created by adding a plurality of structure change vectors to the structure data representing the normal structure. The plurality of structural change vectors are set so that the partial structure matches the partial structure, and the created structural model data is acquired as overall structural information representing the overall structure of the predetermined structure. Structural model data that reflects the structure of a part of the structure can be obtained by a predetermined calculation instead of groping. The overall structure of the predetermined structure including a detection difficult part, can be detected in a short time.

  Further, based on the same configuration as the structure detection device 1 or the structure detection device 2, the structure is a medical image representing the subject from which the predetermined structure has been removed based on the image data P and the entire structure information Fall. The structure removal image generation device may be further provided with a structure removal image data generation means 100 for generating the structure removal image data P ′ carrying the removal image P ′.

  11 shows the structure removal image generation device 3 when the structure same as that of the structure detection device 1 is used as a base, and FIG. 12 shows the structure removal image generation device 4 when the same structure as the structure detection device 2 is used as a base. FIG. In any structure removal image generation device, the structure removal image data generation means 100 carries the image Ps of the predetermined structure based on the whole structure information Fall acquired by the whole structure information acquisition means 90. Generate structure image data Ps, subtract the image Ps of the predetermined structure from the medical image P based on the image data P and the structure image data Ps, and remove the structure from which the predetermined structure is removed The structure removal image data P ′ carrying the image P ′ is generated.

  For example, the structure removal image data generation means 100 controls the other means to obtain the overall shape of the ribs in the radiographic image P as the first overall structure information Fall (1), and then In the chest radiation image P, the texture of the region hollowed out by the overall shape of the rib is warped to a predetermined standardized shape and set as the second partial structure information Fpart (2). The texture is acquired as the second overall structure information Fall (2), and this texture is warped to the overall shape of the rib obtained earlier to obtain the rib image Pr. The rib image Pr is obtained from the chest radiation image P. Subtraction is performed to generate structure-removed image data P ′ carrying a rib-removed image P ′ with the ribs removed from the chest.

  In this case, the texture of the region cut out in the overall shape of the above rib is usually not a pure rib texture, but a texture in which the shadow of the rib overlaps with the blood vessel shadow. Therefore, the obtained rib image may include an error due to the existence of a blood vessel shadow or the like. In order to eliminate such errors as much as possible, the following method can be considered.

  The structure removal image data generation means 100 controls the other means to change the overall shape of the rib from the partial shape of the central lung field of the rib in the chest radiographic image P. Fall (1) is obtained, and then the entire texture of the lung soft part including the vascular shadow is obtained from the partial texture of the lung field excluding the rib region obtained based on the information. Obtained as overall structure information Fall (2). Then, the entire texture of the lung soft part is warped to the shape of the lung field in the chest radiation image P to obtain the lung field soft part image Ph, and further, the lung field soft part image Ph is subtracted from the chest radiation image P, A bone part image Pk in which the rib texture is emphasized is obtained. Since the lung field soft part image Ph includes blood vessel shadows and the like, the bone part image Pk obtained by subtracting this from the chest radiation image P does not include blood vessel shadows and the like. In this bone part image Pk, the texture of the region hollowed out by the overall shape of the rib obtained previously is warped to a predetermined standardized shape and set as the third partial structure information Fpart (3). The overall texture is acquired as the third overall structure information Fall (3), and this texture is warped to the overall shape of the previously obtained rib to obtain a rib image. This rib image is obtained as a chest radiation image P. The structure removal image data carrying the rib removal image from which the rib has been removed from the chest is generated.

  According to such structure-removed image generation apparatuses 3 and 4, a structure-removed image in which a predetermined structure that may be an obstacle in image diagnosis is removed from a single medical image in a short time. It is possible to improve the performance and efficiency of image diagnosis.

  Further, based on the same configuration as that of the structure detection device 1 or the structure detection device 2, in addition to the structure removal image data generation unit 100 described above, based on the structure removal image data P ′ generated by the unit, the structure An abnormal shadow detection device 110 that detects an abnormal shadow candidate R in the object removal image P ′ may be further provided to provide an abnormal shadow detection device.

  FIG. 13 is a diagram showing an abnormal shadow detection device 5 based on the same configuration as the structure detection device 1, and FIG. 14 is a diagram showing an abnormal shadow detection device 6 based on the same configuration as the structure detection device 2. .

  Here, as a method of detecting an abnormal shadow candidate in the “abnormal shadow detection means”, for example, a density gradient (or luminance gradient) in an image disclosed in JP-A-2002-330950 is used as a density gradient. Compared to this multiple structural element using an iris filter process that extracts the image portion with a high degree of concentration of this density gradient vector as a candidate and multiple structural elements according to the size of the abnormal shadow to be detected. A method using a morphological filter process for extracting an image portion whose density fluctuates in a spatially narrow range as a candidate can be considered.

  According to such abnormal shadow detection devices 5 and 6, it is possible to generate in a short time a structure-removed image in which a predetermined structure that can be an obstacle in image diagnosis is removed from one medical image, It is possible to improve the performance and efficiency in the abnormal shadow detection process.

The figure which shows the structure of the structure detection apparatus 1 by one Embodiment of the 2nd structure detection apparatus of this invention. The figure which shows the processing flow by 1st Example of the structure detection apparatus 1. FIG. The figure which shows the chest radiograph P The figure which shows the shape of the part of the rib extracted in the chest radiograph P by the feature point The figure which shows the whole shape of the rib which the created structural model data M represents by a feature point The figure which shows the processing flow by 2nd Example of the structure detection apparatus 1. FIG. The figure which shows the texture of the lung field soft part where the rib area was removed The figure which shows the whole texture of the lung soft part which the created structural model data M represents The figure which shows the structure of the structure detection apparatus 2 by one Embodiment of the 1st structure detection apparatus of this invention. The figure which shows the processing flow of the structure detection apparatus 2 The figure which shows the structure of the structure removal image generation apparatus 3 by one Embodiment of the structure removal image generation apparatus of this invention. The figure which shows the structure of the structure removal image generation apparatus 4 by one Embodiment of the structure removal image generation apparatus of this invention. The figure which shows the structure of the abnormal shadow detection apparatus 5 by one Embodiment of the abnormal shadow detection apparatus of this invention. The figure which shows the structure of the abnormal shadow detection apparatus 6 by one Embodiment of the abnormal shadow detection apparatus of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1, 2 Structure detection apparatus 3, 4 Structure removal image generation apparatus 5,6 Abnormal shadow detection apparatus 10 Image data acquisition means 20 Partial structure information extraction means 30 Partial structure information setting means 40 Structure change vector setting means 50 Structure model data Creation means 60 Index value calculation means 70 Determination means 80 Repetition control means 90 Overall structure information acquisition means 100 Structure removal image data generation means 110 Abnormal shadow detection means

Claims (11)

Acquire image data carrying a medical image representing a subject composed of a plurality of anatomical structures,
Based on the image data, in the medical image, partial structure information representing a structure of a part of a predetermined structure among the plurality of structures is extracted,
Based on the partial structure information, a structure model data is created by adding a plurality of structure change vectors for changing the structure to the structure data representing the structure of the same structure as the predetermined structure,
A method for detecting the structure of an anatomical structure, wherein the structure model data is acquired as overall structure information representing an overall structure of the predetermined structure,
A structure detection method for an anatomical structure, wherein the plurality of structure change vectors are set so that a structure corresponding to the part of the structure represented by the structure model data matches the structure of the part. Acquire image data carrying a medical image representing a subject composed of a plurality of anatomical structures,
Based on the image data, in the medical image, partial structure information representing a structure of a part of a predetermined structure among the plurality of structures is extracted,
A process of setting partial structure information representing a partial structure of a structure of the same type as the predetermined structure, and setting the partial structure information extracted at the first time as the partial structure information, A process of setting information representing a structure obtained by changing a part of the structure represented by the information previously set by the process after the second time;
A structure represented by structure model data obtained by adding a plurality of structure change vectors for changing the structure to structure data representing the structure of the same kind of structure as the predetermined structure based on the partial structure information A process for setting the plurality of structure change vectors so that a structure corresponding to the part matches the part of the structure;
Processing for adding the plurality of structural change vectors to the structural data to create the structural model data;
Based on predetermined information containing information indicating the structure of the predetermined structure obtained from the image data, the difference between the structure represented by the structural model data and the structure of the predetermined structure in the medical image A process of calculating an index value representing the degree,
The process of determining whether the difference is within a predetermined range based on the index value is repeated until it is determined by the determination that the difference is within the predetermined range,
Structure detection of an anatomical structure characterized in that the structure model data created last by the process of creating the structure model data is acquired as overall structure information representing the overall structure of the predetermined structure Method. Image data acquisition means for acquiring image data carrying a medical image representing a subject composed of a plurality of anatomical structures;
Partial structure information extraction means for extracting partial structure information representing a structure of a part of a predetermined structure among the plurality of structures based on the image data;
Based on the partial structure information, the structure model data obtained by adding a plurality of structure change vectors for changing the structure to the structure data representing the structure of the same kind of structure as the predetermined structure. A structure change vector setting means for setting the plurality of structure change vectors so that a structure corresponding to the portion matches the structure of the portion;
Structural model data creating means for creating the structural model data by adding the plurality of structural change vectors to the structural data;
An apparatus for detecting a structure of an anatomical structure, comprising: whole structure information obtaining means for obtaining the structure model data as whole structure information representing the whole structure of the predetermined structure. Image data acquisition means for acquiring image data carrying a medical image representing a subject composed of a plurality of anatomical structures;
Partial structure information extraction means for extracting partial structure information representing a structure of a part of a predetermined structure among the plurality of structures based on the image data;
A means for setting partial structure information representing a partial structure of a structure of the same type as the predetermined structure, the partial structure information being set as the partial structure information for the first time, A partial structure information setting means for setting information representing a structure obtained by changing a part of the structure represented by the information previously set by the means after the second time;
A structure represented by structure model data obtained by adding a plurality of structure change vectors for changing the structure to structure data representing the structure of the same kind of structure as the predetermined structure based on the partial structure information A structure change vector setting means for setting the plurality of structure change vectors so that a structure corresponding to the part matches the part of the structure;
Structural model data creating means for creating the structural model data by adding the plurality of structural change vectors to the structural data;
Based on predetermined information containing information indicating the structure of the predetermined structure obtained from the image data, the difference between the structure represented by the structural model data and the structure of the predetermined structure in the medical image Index value calculation means for calculating an index value representing the degree;
Determination means for determining whether the difference is within a predetermined range based on the index value;
Setting of partial structure information by the partial structure information setting means, setting of a structure change vector by the structure change vector setting means, creation of structure model data by the structure model data creation means, and index value calculation means Repetitive control means that repeats calculation of an index value and determination by the determination means until the determination determines that the difference is within a predetermined range;
An anatomy comprising: overall structure information acquisition means for acquiring the last structural model data created by the structural model data creation means as overall structure information representing the overall structure of the predetermined structure Structure detector for biological structures. The structure data represents an average structure when a number of different data representing the structure of a structure of the same type as the predetermined structure is used as a sample,
The plurality of structure change vectors are a plurality of vectors proportional to respective eigenvectors obtained when principal component analysis is performed using a large number of different data representing the structure of the same kind of structure as the predetermined structure as teacher data. The structure detection device for an anatomical structure according to claim 3 or 4.   6. The structure detection device for an anatomical structure according to claim 3, 4 or 5, wherein the structure is at least one of a shape and a texture. The subject is a chest;
The structure detection apparatus for an anatomical structure according to any one of claims 3 to 6, wherein the predetermined structure is a rib or a lung soft part. Image data acquisition means for acquiring image data carrying a medical image representing a subject composed of a plurality of anatomical structures;
Partial structure information extraction means for extracting partial structure information representing a structure of a part of a predetermined structure among the plurality of structures based on the image data;
A means for setting partial structure information representing a partial structure of a structure of the same type as the predetermined structure, the partial structure information being set as the partial structure information for the first time, A partial structure information setting means for setting information representing a structure obtained by changing a part of the structure represented by the information previously set by the means after the second time;
A structure represented by structure model data obtained by adding a plurality of structure change vectors for changing the structure to structure data representing the structure of the same kind of structure as the predetermined structure based on the partial structure information A structure change vector setting means for setting the plurality of structure change vectors so that a structure corresponding to the part matches the part of the structure;
Structural model data creating means for creating the structural model data by adding the plurality of structural change vectors to the structural data;
Based on predetermined information containing information indicating the structure of the predetermined structure obtained from the image data, the difference between the structure represented by the structural model data and the structure of the predetermined structure in the medical image Index value calculation means for calculating an index value representing the degree;
Determination means for determining whether the difference is within a predetermined range based on the index value;
Setting of partial structure information by the partial structure information setting means, setting of a structure change vector by the structure change vector setting means, creation of structure model data by the structure model data creation means, and index value calculation means Repetitive control means that repeats calculation of an index value and determination by the determination means until the determination determines that the difference is within a predetermined range;
Overall structure information acquisition means for acquiring the structure model data last created by the structural model data creation means as overall structure information representing the overall structure of the predetermined structure;
Structure-removed image data for generating structure-removed image data carrying a structure-removed image that is a medical image representing the subject from which the predetermined structure has been removed, based on the image data and the entire structure information A structure-removed image generation apparatus comprising: a generation unit. Image data acquisition means for acquiring image data carrying a medical image representing a subject composed of a plurality of anatomical structures;
Partial structure information extraction means for extracting partial structure information representing a structure of a part of a predetermined structure among the plurality of structures based on the image data;
A means for setting partial structure information representing a partial structure of a structure of the same type as the predetermined structure, the partial structure information being set as the partial structure information for the first time, A partial structure information setting means for setting information representing a structure obtained by changing a part of the structure represented by the information previously set by the means after the second time;
Based on the partial structure information, structural model data obtained by adding a plurality of structural change vectors for changing the structure to structural data representing a normal structure of a structure of the same type as the predetermined structure. A structure change vector setting means for setting the plurality of structure change vectors so that a structure corresponding to the part of the structure to be represented matches the part of the structure;
Structural model data creating means for creating the structural model data by adding the plurality of structural change vectors to the structural data;
Based on predetermined information containing information indicating the structure of the predetermined structure obtained from the image data, the difference between the structure represented by the structural model data and the structure of the predetermined structure in the medical image Index value calculation means for calculating an index value representing the degree;
Determination means for determining whether the difference is within a predetermined range based on the index value;
Setting of partial structure information by the partial structure information setting means, setting of a structure change vector by the structure change vector setting means, creation of structure model data by the structure model data creation means, and index value calculation means Repetitive control means that repeats calculation of an index value and determination by the determination means until the determination determines that the difference is within a predetermined range;
Overall structure information acquisition means for acquiring the structure model data last created by the structural model data creation means as overall structure information representing the overall structure of the predetermined structure;

Structure-removed image data for generating structure-removed image data carrying a structure-removed image that is a medical image representing the subject from which the predetermined structure has been removed, based on the image data and the entire structure information Generating means;
An abnormal shadow detection apparatus comprising: an abnormal shadow detection unit that detects an abnormal shadow candidate in the structure removal image based on the structure removal image data. Computer
Image data acquisition means for acquiring image data carrying a medical image representing a subject composed of a plurality of anatomical structures;
Partial structure information extraction means for extracting partial structure information representing a structure of a part of a predetermined structure among the plurality of structures based on the image data;
Based on the partial structure information, the structure model data obtained by adding a plurality of structure change vectors for changing the structure to the structure data representing the structure of the same kind of structure as the predetermined structure. A structure change vector setting means for setting the plurality of structure change vectors so that a structure corresponding to the portion matches the structure of the portion;
Structural model data creating means for creating the structural model data by adding the plurality of structural change vectors to the structural data;
A program for causing the structure model data to function as an overall structure information acquisition unit that acquires the entire structure information representing the overall structure of the predetermined structure. Computer
Image data acquisition means for acquiring image data carrying a medical image representing a subject composed of a plurality of anatomical structures;
Partial structure information extraction means for extracting partial structure information representing a structure of a part of a predetermined structure among the plurality of structures based on the image data;
A means for setting partial structure information representing a partial structure of a structure of the same type as the predetermined structure, the partial structure information being set as the partial structure information for the first time, A partial structure information setting means for setting information representing a structure obtained by changing a part of the structure represented by the information previously set by the means after the second time;
A structure represented by structure model data obtained by adding a plurality of structure change vectors for changing the structure to structure data representing the structure of the same kind of structure as the predetermined structure based on the partial structure information A structure change vector setting means for setting the plurality of structure change vectors so that a structure corresponding to the part matches the part of the structure;
Structural model data creating means for creating the structural model data by adding the plurality of structural change vectors to the structural data;
Based on predetermined information containing information indicating the structure of the predetermined structure obtained from the image data, the difference between the structure represented by the structural model data and the structure of the predetermined structure in the medical image Index value calculation means for calculating an index value representing the degree;
Determination means for determining whether the difference is within a predetermined range based on the index value;
Setting of partial structure information by the partial structure information setting means, setting of a structure change vector by the structure change vector setting means, creation of structure model data by the structure model data creation means, and index value calculation means Repetitive control means that repeats calculation of an index value and determination by the determination means until the determination determines that the difference is within a predetermined range;
A program for functioning as overall structure information acquisition means for acquiring the last structural model data created by the structural model data creation means as overall structure information representing the overall structure of the predetermined structure.

Images