JP7052103B2 - Radiography equipment, radiography system, radiography method, and program - Google Patents

Description

The present invention relates to a radiographic apparatus, a radiological imaging system, a radiographic imaging method, and a program for specifying the contour of a predetermined target structure of a subject in an image.

In recent years, digital radiography equipment has become widespread in the medical field. When performing various examinations and diagnoses by radiography, it is common to create a lot of diagnostic value by applying various image processing to the obtained digital images. In particular, the technique of automatically extracting the contour of a specific target structure from the subject of an image is related to many other image processing techniques and has become an important technical issue.

As a typical example, a technique has been proposed in which the contour of the lung field is extracted from the acquired image when the frontal chest image is taken using a radiography apparatus, and the shape and position of the lung field are automatically recognized. The shape and position of the lung field in the frontal chest image is used for a wide range of diagnostic support purposes, such as calculation of cardiothoracic ratio (= heart width / chest width) and use for automatic recognition of lung field nodules. Therefore, high accuracy is required for these contour extractions.

In response to such demands, a technique for automatically extracting contours of a predetermined structure from an image has been proposed. As a typical example, the active shape model of Non-Patent Document 1 is known. This method is a technique that performs statistical analysis of multiple sample images by pre-learning, models the contour of the target structure to be extracted, and estimates the shape of the target structure based on the trained model. Is. For example, in Patent Document 1, two models, a contour shape model and a texture model representing local features (pixel values, etc.) around the contour, are set as contour trained models, and the contour of the lung field is extracted. A method has been proposed.

Japanese Unexamined Patent Publication No. 2004-008419

T. F. Cootes et, al. "Active shop models", computer vision and image understanding vol. 61, no1, Jan 1995

However, the above-mentioned prior art has the following problems in improving the accuracy of contour extraction.

The first is erroneous detection of contour candidates in local search. In the local search, a predetermined search range is set around the contour candidate, and the most similar contour candidate is searched from the search range. If the target structure is not included in the set search range, inappropriate contour candidates having characteristics similar to the texture model may be extracted from the search range, which may lead to misidentification of the contour candidates. In particular, when the search range is set to a region largely deviated from the region where the contour of the desired target structure can exist, inappropriate contour candidates are often extracted at a position significantly deviated from the original contour.

The second is improper deformation of contour candidates when the characteristics of the contour shape are maintained. As described above, when some of the contour candidates include inappropriate contour candidates, when trying to maintain the characteristics of the contour shape, the entire shape of the contour candidates is affected by the inappropriate contour candidates, and the contour candidates are affected. May be improperly deformed.

When local search and maintenance of contour shape features are repeated with these inappropriate contour candidates extracted, the search for the contour of the target structure is repeated while being affected by the inappropriate contour candidates, resulting in a result. Sometimes the desired contour could not be extracted.

The present invention has been made in view of the above problems, and an object of the present invention is to provide a radiation apparatus, a radiography system, a radiography method, and a program capable of appropriately specifying (or extracting) the contour of a target structure. And.

The radiography apparatus of the present invention is a radiography apparatus for specifying the contour of a predetermined target structure of a subject in an image, and is a contour search region in which the contour is searched based on the anatomical features of the structure of the subject. The area setting means for setting the contour candidate, the contour candidate setting means for setting the contour candidate of the target structure, and the contour candidate adjusted in order to make the contour candidate included in the contour search area approximate to the contour of the target structure. Provided with contour adjusting means

According to the present invention, the contour of the target structure can be appropriately specified (or extracted).

It is a block diagram which shows the basic structural example of the radiographing apparatus including the radiographing apparatus which concerns on embodiment of this invention. It is a block diagram which shows the example of the basic structure of the contour extraction circuit of the radiography apparatus which concerns on embodiment of this invention. It is a schematic diagram which shows the teacher data and a shape model in a learning circuit. It is a schematic diagram which shows the texture model in a learning circuit. It is a flowchart which shows the example of the processing flow of the contour extraction circuit. It is a flowchart which shows the example of the processing flow of the setting of a contour search area, the correction of a contour candidate, and the setting of a local area. It is a figure which shows the example of the chest front image including a structural area, a shielding area, and a direct irradiation area. (A) It is a figure which shows the example which adjusts the size of a contour candidate based on the anatomical feature of a target structure. (B) It is a figure which shows the example which adjusts the angle of a contour candidate based on the anatomical feature of a target structure. It is a figure which shows the example which separates and moves a part of contour candidates from contour candidates. It is a figure which shows the example which moves the whole contour candidate. (A) It is a figure which shows the example of the local region of a contour candidate. (B) It is a figure which shows the example which selects a new contour point in the local region of a contour candidate. It is a flowchart which shows the example of the processing flow of the smoothing process of a contour candidate.

Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings. FIG. 1 will describe a configuration example of a radiographic imaging system including a radiographic imaging apparatus that specifies the contour of a predetermined target structure of a subject in an image (for example, a radiographic image).

In this embodiment, an active shape model based on two models, a contour shape model and a texture model representing local features (pixel values, etc.) around the contour, is used, but other methods for extracting the contour of the target structure are also used. Applicable.

FIG. 1A is a block diagram showing an example of a basic configuration of a radiography system including the radiography apparatus of the present embodiment. FIG. 1B is a block diagram showing an example of a basic configuration of a contour extraction circuit of the radiography apparatus of the present embodiment.

The radiation photographing system 100 includes a radiation generator 101 that generates radiation, a bed 103 on which the subject 102 is arranged, a radiation detection device 104 that detects radiation and outputs image data according to the radiation that has passed through the subject 102. A control device 105 that controls the radiation generation timing and radiation generation conditions of the radiation generator 101, a data collection device 106 that collects various digital data, and an information processing device (radiation) that performs image processing and control of the entire device according to the user's instructions. It is equipped with a photographing device) 107.

The information processing apparatus 107 includes a preprocessing circuit 109, a learning circuit 110, a contour extraction circuit 111, an image processing apparatus 108 including a diagnostic image processing circuit 112, a CPU 114, a memory 115, an operation panel 116, and storage. The device 117 and the display device 118 are provided, and these are electrically connected via the CPU bus 113.

The memory 115 stores various data and the like necessary for processing by the CPU 114, and also includes a working work memory of the CPU 114. Further, the CPU 114 uses the memory 115 to control the operation of the entire device according to the user's instruction input to the operation panel 116.

The contour extraction circuit 111 includes an area setting unit 151, a contour candidate setting unit 152, and a contour adjustment unit 153.

In the present invention, radiation is not limited to commonly used X-rays, but is not limited to α-rays, β-rays, γ-rays, etc., which are beams produced by particles (including photons) emitted by radioactive decay. Beams with similar or higher energies (eg particle beams, cosmic rays, etc.) are also included. Hereinafter, a case where X-rays are used as radiation will be described as an example.

The radiological imaging system 100 starts an imaging sequence of the subject 102 according to a user's instruction via the operation panel 116. X-rays under predetermined conditions are generated from the radiation generator 101, and the X-rays that have passed through the subject 102 are irradiated to the radiation detection device 104. Here, the control device 105 controls the radiation generation device 101 with X-ray generation conditions such as voltage, current, and irradiation time, and can generate X-rays under predetermined conditions.

The image information output from the radiation detection device 104 is converted into an electric signal by the radiation detection device 104, and is collected as digital image data by the data collection device 106. The image data collected by the data collecting device 106 is transferred to the information processing device 107, and is transferred to the memory 115 via the CPU bus 113 under the control of the CPU 114. The image processing device 108 applies various image processing to the image data stored in the memory 115. The image processing device 108 extracts a desired contour of the target structure and creates an image suitable for diagnosis, stores the result in the storage device 117, and displays the result on the display device 118.

Next, detailed processing of the image processing apparatus 108 will be described. The image processing device 108 includes a preprocessing circuit 109, a learning circuit 110, a contour extraction circuit 111, and a diagnostic image processing circuit 112.

The preprocessing circuit 109 includes a circuit that preprocesses image data. Here, the pre-processing is to correct the characteristic variation caused by the characteristics of the radiation detection device 104 for the raw image data collected by the data acquisition device 106 so that the subsequent processing can be appropriately performed. Includes various correction processes.

The pretreatment circuit 109 is adapted to select an appropriate correction depending on the type of the radiation detection device 104. The correction includes dark correction, gain correction, defect correction, logarithmic conversion processing of image data, and the like. The dark correction is a correction for removing fixed pattern noise and the like of the radiation detection device 104. The gain correction is a correction for making the sensitivity of the radiation detection device 104 uniform in the imaging surface. The defect correction is a correction for interpolating the defective pixels included in the manufacturing process of the radiation detection device 104 from the surrounding pixels.

The diagnostic image processing circuit 112 receives the results of preprocessing and contour extraction processing described later, applies noise reduction processing, various enhancement processing, gradation conversion processing, etc. as diagnostic image processing, and is used for diagnosis. Perform processing.

Hereinafter, an example in which the lung field is the target structure in the front chest image of the X-ray simple radiography will be described. However, the present invention is not limited to the lung field, and can be applied to other target structures such as the heart.

Further, in the present embodiment, an active shape model is used, and an appropriate initial position is set as a contour candidate by the contour shape model. Then, the local feature amount around the contour candidate is compared with the texture model, a more probable contour candidate is calculated, and the contour candidate is updated to a more probable position. After that, the shape model deforms the updated contour candidates while maintaining the characteristics of the contour shape. The final contour candidate is identified by repeating the above local search and the search process by maintaining the contour shape feature.

The learning circuit 110 performs pre-learning processing regarding the characteristics of the contour of the target structure. This process is performed prior to X-ray photography, and a large number of training images are input in advance to create a statistical model (trained model) of features representing the contour of the target structure. The type of statistical model is not particularly limited, and the optimum model is used according to the type of contour of the target structure.

The learning circuit 110 learns a shape model that represents the statistical shape of the contour of the target structure. Further, the learning circuit 110 learns a texture model representing statistical local features of the contour of the target structure by the pixel value profile. In this embodiment, since the lung field in the front chest image is the target structure, a shape model based on the shape of the lung field contour and a texture model based on the profile information of the pixel values around the lung field contour are adopted.

First, a shape model of the contour of the lung field (target structure) will be described. As shown in FIG. 2A, teacher data in which the lung field contours of N learning images k (k: 1 to N) are manually specified are prepared. The lung field contour is designated using n contour points i (i: 1 to n).

Assuming that the x-coordinate of the i-th contour point in the k-th learning image is x- _ki and the y-coordinate is y- _ki , the contour point is expressed by the vector X _k as in the equation (1). The number of contour points n and the number of learning images N are not particularly limited, but it is preferable to set, for example, n = 100 or more and N = 1500 or more in the lung field contour in the chest front image.

Then, the vector of the contour point i is obtained for each of the N learning images k, and the vector represented by the equation (2) is defined as X.

As the shape model of the lung field contour, the one obtained by performing the principal component analysis of X and converting it as shown in the equation (3) is used.

Here, X _mean represents the average shape of X. P _S represents a principal component vector. b _S represents an eigenvalue. The principal component used can be freely selected, and for example, it is preferable that the principal component vector is selected so that the cumulative contribution rate is 90%.

Next, the texture model of the lung field contour will be described. As shown in FIG. 2B, for each of the contour points i of the lung field, the pixel value profile _vim is acquired in the perpendicular direction of the contour line. By setting the profile length to L (m: 1 to L) and calculating the pixel value profile v _im in the kth learning image for each contour point, the pixel value profile v _kim (i: 1 to n, m: 1). ~ L) is represented by the matrix V _k as in the equation (4).

When acquiring the pixel value profile, it is desirable that the image signal be normalized. For example, the image signal is normalized by converting each pixel value profile _vim from 0 to 1 using the maximum value and the minimum value of the acquired pixel value profile. Further, the image signal may be normalized by normalizing the entire trained image to 0 to 1 and then acquiring the pixel value profile _vim . It is preferable that the profile length L is set to, for example, 40 pixels or more.

Similar to the shape model, the pixel value profile of Eq. (4) is obtained for each of the N learning images, and the principal component analysis is performed. Let it be a texture model V.

Here, V _mean represents the average shape of V. _PA represents a principal component vector. b _A represents an eigenvalue. The principal component used can be freely selected, and for example, it is preferable that the principal component vector is selected so that the cumulative contribution rate is 90%.

As described above, the shape model and the texture model may be stored in the storage device 117 as a general-purpose model for the radiography apparatus and used by using the pre-calculated model. Further, the shape model and the texture model may be calculated by using the image data captured by each radiographic imaging device as the learning image k, depending on the usage status of the radiographic imaging device.

The contour extraction circuit 111 performs a process of specifying (or extracting) the contour of a predetermined target structure from the subject 102. The processing flow of the contour extraction circuit 111 will be described with reference to FIG.

In step S301, analysis preparation processing is performed on the preprocessed image obtained by the preprocessing circuit 109. The analysis preparation process is a process aimed at improving the contour extraction accuracy in the subsequent process, and a process suitable for the target structure is selected. The analysis preparation process includes a normalization process of an image signal, an edge enhancement process for emphasizing the contour to be extracted, a gradation conversion process, a noise reduction process, and a rotation process for making the calculated image direction constant. Further, the analysis preparation process includes an enlargement / reduction process for normalizing the subject size on the image when the pixel size of the radiation detection device 104 is different.

In step S302, the first contour candidate _Xt is set. The contour candidate setting unit 152 sets the contour candidate of the target structure. The contour candidate X _t includes n contour points having the same number of contour points as the shape model created in the learning circuit 110, and is represented by a vector as shown in the equation (6).

Here, the first contour candidate _Xt is the initial position when the contour is sequentially searched by the contour search loop processing of steps S303 to S308. The contour candidate _Xt may be set to an arbitrary position, but if the contour candidate Xt is set as the initial position as close as possible to the desired contour of the target structure, the contour extraction accuracy will be improved. It is preferable to set the obtained average shape X _mean of the lung field as the initial position. In this way, the contour candidate setting unit 152 sets the initial position of the contour candidate based on the shape model.

In steps S303 to S305, a process of calculating and evaluating a local feature amount that characterizes the contour of the lung field (target structure) from a predetermined search range (contour search region) and searching for the contour is performed. First, in step S303, the contour search region S in which the contour of the lung field is searched based on the local feature amount of the contour candidate of the lung field (target structure) is set. The area setting unit 151 sets the contour search area S in which the contour of the target structure is searched based on the anatomical features of the structure of the subject 102.

Specifically, the process in step S303 is a process including the flow shown in FIG. 4, and the process in step S303 will be described in detail below with reference to FIG.

In step S401, a process of calculating the anatomical feature of the subject 102 is performed. This process is performed independently of the statistical model (learned model) obtained by the learning circuit 110, and the contour of the target structure is assumed to exist based on the anatomical features of the subject 102. This is a process for the purpose of calculating the search area S. The region setting unit 151 is based on at least one of a structural region showing an anatomical feature of the structure of the subject 102, a shielded region where radiation is shielded, and a direct irradiation region where radiation is directly applied to the radiation detection device. The contour search area S is set.

Here, the anatomical features of the subject 102 will be described using the front chest image of FIG. 5 as a specific example. The front chest image is roughly divided into a shield region 411, a direct irradiation region 412, a shoulder joint / arm region (first structural region) 413, a spine region (second structural region) 414, and an abdominal region (third structural region). ) 415 and the lung field region (fourth structural region) 416.

The shielded area 411 is an area where radiation is shielded by a collimator. The direct irradiation area 412 is an area in which the radiation is directly applied to the radiation detection device 104 without passing through the subject 102. The shoulder joint / arm region 413 is a region including the anatomical features of the shoulder joint and the arm of the subject 102. The spinal region 414 is an region containing the anatomical features of the spine of the subject 102. The abdominal region 415 is an region containing the anatomical features of the abdomen of the subject 102. The lung field region 416 is a region containing the anatomical features of the lung field of the subject 102.

This process includes a shielding area recognition process (step S4011) for recognizing the shielding area 411, a direct irradiation area recognition process (step S4012) for recognizing the direct irradiation area 412, and a structural area (shoulder joint / arm area 413) of the subject 102. It includes a structure recognition process (step S4013) for recognizing the spinal region 414 and the abdominal region 415).

In the present embodiment, the lung field region 416 is estimated by removing the regions 411 to 415 recognized by steps S4011 to S4013 from the image data of the subject 102, and the lung field region 416 is set as the contour search region S. .. In this way, the region setting unit 151 identifies structural regions 413,414,415 of structures other than the lung field (target structure) based on the anatomical features of the structure other than the target structure, and structural regions 413,414. , 415 may be excluded from the area, and the contour search area S may be set.

A known technique can be applied to the process of recognizing the structural region based on the anatomical features of the subject 102. Further, all the recognition processes of steps S4011 to S4013 may be used, or the recognition process may be selected according to the target structure.

In step S402, a false detection determination process for the contour candidate _Xt is performed. This is a process for determining whether or not the contour candidate _Xt set in step S302 or step S308 has a structure likely to be a lung field (target structure) based on anatomical features. .. The result of the false positive determination process is confirmed in step S403, and if the result is non-conforming, the contour candidate _Xt is corrected in step S404. If the result is compliant, the process proceeds to step S405.

A preferable specific example of the false detection determination process of the contour candidate X _t is a determination process based on the size and angle of the contour candidate X _t . The size determination process (step S4021) based on the size of the contour candidate X _t and the angle determination process (step S4022) based on the angle of the contour candidate X _t will be described below.

In the size determination process of step _S4021 , the size of the lung field defined by the contour candidate Xt is calculated. As shown in FIG. 6A, left lung contour candidate XR _t = (xr _t1 , xr _t2 , ..., Xr nt _n , yr _t1 , yr _t2 , ..., yr _nt ) and right lung contour candidate XL. Left lung height WR, right lung height WL, left lung height WR, left lung height WL, left lung height WR, left lung height WL, left lung height WR, left lung height WL, left lung height WR, left lung height _WL , left lung height based _{on t} = ( _{xl t1} , _{xl t2} , ... The width HR and the width HL of the right lung are determined according to the formula (7).

Left and right lung heights Wr, WL and left and right lung widths Hr, HL, and their left-right ratios, multiple clinical data were compiled in advance, and left and right lung heights, left and right lung widths, and lung heights and lung heights. Set the range of variation in clinical data regarding the left-right ratio of width.

Based on the information on the variation range, it is determined whether or not the size of the lung field estimated from the contour candidate _Xt is out of the variation range. If the size of the lung field is out of the variation range, a process of expanding or contracting _Xt is performed in step S404 so that the lung field size is within the variation range.

In the lung field angle recognition process in step S4022, the angle θ of the lung field arrangement is calculated based on the left lung contour candidate XR _t and the right lung contour candidate XL _t . In this processing, as shown in FIG. 6B, the lung apex 421 and the costal lateral angle 422, which are anatomical features of the lung field, are recognized by image processing or the like, and a straight line connecting the lung apex 421 and the costal lateral angle 422 is connected. The angle θ between the y-axis and the y-axis is calculated. Similar to step S4021, a plurality of clinical data are aggregated in advance for the angle θ, and the range of variation of the clinical data regarding the angle of arrangement of the lung field is set.

Based on the information on the variation range, it is determined whether or not the angle of the lung field arrangement estimated from the contour candidate _Xt deviates from the variation range. If the angle of the lung field arrangement is out of the variation range, the process of rotating _Xt is performed in step S404 so that the angle of the lung field arrangement is within the variation range.

In this way, the contour adjusting unit 153 adjusts at least one of the size, angle, and position of the contour candidate _Xt based on the anatomical features of the target structure. By the above processing, if the anatomical features (size, angle, position, etc.) of the target structure deviate from the range of variation assumed as the target structure of the human body, the contour candidate _Xt is corrected to be the target. The search for the contour of the structure can be started from a more accurate position. As a result, the accuracy of extracting the final contour candidate can be improved.

In step S405, it is determined whether or not the contour candidate _Xt is included in the contour search area S calculated in step S401. If the contour candidate _Xt is not included in the contour search area S in step S405, the process proceeds to step S406.

In step S406, a process of moving a contour point of the contour candidate _Xt that is not included in the contour candidate area S to the contour candidate area S is performed. The contour adjustment unit 153 moves the contour candidate X _t so that the contour candidate X _t is included in the contour search region S when at least a part of the contour candidate X _t is not included in the contour search region S.

For example, as shown in FIG. 7, the contour points of the contour candidate X _t that are not included in the contour search area S are translated into the contour candidate area S, and the contour candidate X _t is set as the updated contour candidate X _t '. In this way, the contour adjusting unit 153 moves a part of the contour candidate X _t whose peripheral region of the contour candidate X _t or the contour candidate X _t is not included in the contour search region S separately from the contour candidate X _t . .. Further, as shown in FIG. 8, the entire contour points of the contour candidate X _t are translated into the contour candidate region S so that the contour points of the contour candidate X _t are included in the contour candidate region S, and the contour candidate X _t is translated. It may be an update contour candidate X _t '.

Further, the contour adjusting unit 153 includes a peripheral region when at least a part of a predetermined peripheral region (for example, a local region from which a local feature amount of the contour is extracted) of the contour candidate _Xt is not included in the contour search region S. The contour candidate _Xt may be moved so that the region is included in the contour search region S.

If the contour candidate _Xt is included in the contour search area S in step S405, the process proceeds to step S407.

In step S407, a local range (local area) D of each contour point is set in order to acquire a local feature amount (pixel value or the like) around the contour candidate _Xt . For each contour point of contour candidates X _t and X _t ', the profile length L _s is set in the perpendicular direction of the contour line, and the range of the profile length L _s included in the contour search area S is the local range of each contour point. Set as D. There is no limitation on the profile length L _s , but for example, in contour extraction of the lung field, it is preferable to set the profile length L _s to about twice the pixel value profile length L of the texture model.

In step S304 of FIG. 3, in the local range D set in step S303, a process of calculating the local features of the contour candidates X _t and X _t'is performed. As the local feature, the profile information of the pixel value is used. FIG. 9 is a diagram showing a local range D of the pixel value profile of the update contour candidate X _t '. As shown in FIG. 9A, a profile length L _s (number of profiles L _s ) local range D is set in the perpendicular direction of the contour line centering on the contour point (x _ki ', y _ki '), and each contour is defined. The pixel value profile is calculated in the local range D of the points. As a result, a pixel value profile of the number of contour points n × the number of profiles _Ls of the contour candidate X _t'is acquired.

In step S305, a new contour candidate X _e is extracted from each acquired pixel value profile for each contour point of the contour candidate X _t '. As shown in FIG. 9B, in the calculation of the contour candidate _Xe , the pixel value profile of the local range D of each contour point is compared with the pixel value profile v of the texture model, and points that are similar to each other are newly added. It is done by selecting it as a contour point.

The contour adjusting unit 153 adjusts the contour candidate by comparing the local feature (image profile v) of the contour of the target structure of the texture model with the local feature of the contour candidate X _t '. The contour adjustment unit 153 adjusts the contour candidate to the position where the similarity between the local feature of the contour of the target structure of the texture model (image profile v) and the local feature of the contour candidate X _t'is the highest (contour candidate X _e ). do. In this way, the contour adjusting unit 153 adjusts the contour candidate in order to approximate the contour candidate X _t'included in the contour search area S to the contour of the target structure.

The method for searching for similar points is not particularly limited. For example, as in the equation (8), the average shape _Vmaan and the principal component vector PA of the corresponding contour points of the texture model are used for the pixel value profile g (m) of the local range _D of the predetermined contour points. It is preferable to perform the KL conversion that has been performed and select the contour point i that minimizes the distance between the pixel value profile of the local range D and the pixel value profile of the texture model as a new contour point.

In step S306, the contour candidate _Xe is smoothed. The new contour candidate _Xe selects the contour points that are most similar to the texture model of the pixel value profile of the lung field contour. However, if the number of contour search loops in steps S303 to S308 is relatively small and the difference between the contour candidate _Xe and the actual contour is large, the shape of the contour candidate _Xe seems to be a lung field (target structure). It may collapse greatly from. Therefore, the smoothing process is a process of shaping the contour candidate _Xe into a shape like a lung field. A known technique can be applied to the smoothing treatment. As a preferred example, the process including the flow shown in FIG. 10 will be described.

In step S601, the coordinate transformation for moving the center of gravity G = (Gx, Gy) of the contour candidate X _e to the origin (0, 0) is performed according to the equation (9), and the contour shape X _e'is calculated. As a result, in steps S602 to S604, processing is performed in a state where the contour shape is targeted and the position information of the contour shape on the image is excluded.

In step S602, KL conversion is performed using the _mean shape _X main and the principal component vector PS in the shape model according to the equation (10), and the contour shape c in the transformed coordinate system is obtained.

In step S603, the contour shape c in the transformed coordinate system is subjected to threshold processing according to the equation (11) using the eigenvalue bS of the shape model and an arbitrary coefficient m, and the contour shape _c'in the transformed coordinate system is calculated. do. The coefficient m is an arbitrary coefficient that determines the degree of smoothness, and for example, a value of about m = 3 is specified. Note that j is the number of vectors when the principal component vector is selected so that the cumulative contribution rate becomes 90% when the shape model of the lung field (target structure) is created.

In step S604, the contour shape c'that has been threshold-processed in the conversion coordinate system is subjected to inverse KL conversion according to the equation (12) to calculate the smoothed contour candidate _Xc .

In step S605, the center of gravity = origin of the contour candidate _{X c} is moved to the center of gravity G = (Gx, Gy) before the coordinate return, and the smoothed contour candidate XS is calculated according to the equation (13).

In this way, the contour adjusting unit 153 smoothes the contour candidates based on the shape model. By the above processing, the contour candidate and the shape model can be integrated and updated to the contour candidate _XS in a state where the shape unique to the lung field is secured.

In step S307 of FIG. 3, the end determination of the contour search loop processing is performed. The end determination is determined based on whether or not the search loop process has been performed a predetermined number of times. When the number of loops of the search loop processing reaches a predetermined number, the contour search loop processing is terminated.

Further, in order to avoid performing the search loop processing more than necessary and shorten the processing time, the contour candidate X _S1 obtained in the previous contour search loop and the contour candidate X obtained in the current contour search loop are used. The difference of _S2 may be calculated, and it may be determined whether or not the difference is equal to or less than a predetermined threshold value. When the difference is equal to or less than a predetermined threshold value, it is determined that the updated contour candidate _XS is close to the actual contour, and the contour search loop processing is terminated.

Further, the end determination of the contour search loop processing may be performed by combining the determination based on the number of contour search loops and the determination based on the difference between the contour candidates. As a result of the end determination, if it is determined to continue the contour search loop processing, the process proceeds to step S308, and if it is determined to end the contour search loop processing, the process proceeds to step S309.

In step S308, in order to continue the contour search loop, the contour candidate XS calculated in step _S306 is reset as the contour candidate _Xt , and the process proceeds to step S303 to continue the contour search loop processing. Here, the contour candidate setting unit 152 resets the adjusted contour candidate XS as the updated contour candidate _{X t} , and the contour adjustment unit 153 approximates the updated contour candidate X _t to the contour of the target structure. In order to make the contour candidate readjust in steps S303 to S306.

In step S309, the contour search loop is terminated, and the contour candidate X _S is output as the final contour candidate X _F.

Through the above processing, the contour extraction circuit 111 can realize a function of extracting the contour of a predetermined target structure from the subject 102.

In addition, in order to improve the accuracy of the extraction result, steps S301 to S309 may be performed at a plurality of resolutions. In this case, the contour candidate setting unit 152 uses the contour candidate XS adjusted in the image (radiation image) of the first resolution as the updated contour candidate _{X t} , and has a second resolution different from the first resolution. Reset in the image. Then, the contour adjusting unit 153 re-adjusts the contour candidate in the image (radiation image) having the second resolution in order to approximate the updated contour candidate _Xt to the contour of the target structure.

Here, it is preferable that the second resolution is higher than the first resolution. The accuracy of contour extraction can be improved by performing contour search processing in order from low resolution to high resolution and setting the contour candidate X _F at low resolution as the first contour candidate X _t at high resolution in the subsequent stage.

The radiography apparatus of the present invention sets the contour search area S by using the anatomical features of the target structure in the contour extraction of the image data, corrects the size, angle, and position of the contour candidate, and the contour candidate is The contour candidate is moved so as to be included in the contour search area S. As a result, it becomes possible to appropriately manage the search range of the local feature, and it is possible to extract an appropriate contour candidate. In addition, even if inappropriate contour points that deviate from the contour of the target structure are extracted, the influence of the inappropriate contour points can be corrected by correcting the position of the contour points, and the desired contour can be extracted correctly. Can be done.

Although the preferred embodiments of the present invention have been described above, it goes without saying that the present invention is not limited to these examples, and various modifications and modifications can be made within the scope of the gist thereof. In particular, as a preferred example of the operation of the learning circuit 110 and the contour extraction circuit 111 of the present invention, an example in which the lung field contour in the chest front image of the X-ray simple radiography is used as the contour of the target structure has been described, but the present invention has been applied. The example is not limited to this. For example, the present invention is applicable not only to an X-ray photographed image but also to a CT image or a general camera image.

Further, the present invention may be achieved by reading the software program that realizes the functions of the above-described embodiment directly or remotely by the computer of the system or the device and executing the program code.

100 Radiation imaging system 101 Radiation generator 103 Sleeper 104 Radiation detection device 105 Control device 106 Data collection device 107 Information processing device 108 Image processing device 109 Preprocessing circuit 110 Learning circuit 111 Contour extraction circuit 112 Diagnostic image processing circuit 113 CPU bus 114 CPU
115 Memory 116 Operation panel 117 Storage device 118 Display device 151 Area setting unit 152 Contour candidate setting unit 153 Contour adjustment unit

Claims (15)

A medical image processing device that identifies the contour of a predetermined target structure of a subject in an image.
An area setting means for setting a contour search area in which the contour is searched based on the anatomical characteristics of the structure of the subject, and
Contour candidate setting means for setting contour candidates of the target structure, and
A contour adjusting means for adjusting the contour candidate in order to approximate the contour candidate included in the contour search region to the contour of the target structure, and a contour adjusting means.
Equipped with
The contour adjusting means includes the contour candidate or the peripheral region in the contour search region when at least a part of the contour candidate or a predetermined peripheral region of the contour candidate is not included in the contour search region. As described above, a medical image processing apparatus characterized by moving the contour candidate . The region setting means is based on at least one of a structural region showing the anatomical features of the structure of the subject, a shielded region where radiation is shielded, and a direct irradiation region where radiation is directly applied to the radiation detection device. The medical image processing apparatus according to claim 1, wherein the contour search area is set. The region setting means identifies a structural region of a structure other than the target structure based on the anatomical features of the structure other than the target structure, and sets the contour search region from the region excluding the structural region. The medical image processing apparatus according to claim 1 or 2. One of claims 1 to 3, wherein the contour adjusting means adjusts at least one of the size, angle, and position of the contour candidate based on the anatomical feature of the target structure. The medical image processing apparatus described in. The contour adjusting means according to claim 1 to 4 , wherein the contour candidate or a part of the contour candidate whose peripheral region is not included in the contour search region is separated from the contour candidate and moved. The medical image processing apparatus according to any one of the following items . A learning means for learning a texture model representing a statistical local feature of the contour of the target structure by using a pixel value profile is provided.
Any of claims 1 to 5 , wherein the contour adjusting means adjusts the contour candidate by comparing the local feature of the contour of the target structure of the texture model with the local feature of the contour candidate. The medical image processing apparatus according to item 1. The sixth aspect of the present invention is characterized in that the contour adjusting means adjusts the contour candidate to a position where the similarity between the local feature of the contour of the target structure of the texture model and the local feature of the contour candidate is the highest. The medical image processing device described. A learning means for learning a shape model representing the statistical shape of the contour of the target structure is provided.
The medical image processing apparatus according to any one of claims 1 to 7 , wherein the contour candidate setting means sets an initial position of the contour candidate based on the shape model. A learning means for learning a shape model representing the statistical shape of the contour of the target structure is provided.
The medical image processing apparatus according to any one of claims 1 to 8 , wherein the contour adjusting means smoothes the contour candidate based on the shape model. A medical image processing device that identifies the contour of a predetermined target structure of a subject in an image.
An area setting means for setting a contour search area in which the contour is searched based on the anatomical characteristics of the structure of the subject, and
Contour candidate setting means for setting contour candidates of the target structure, and
A contour adjusting means for adjusting the contour candidate in order to approximate the contour candidate included in the contour search region to the contour of the target structure, and a contour adjusting means.
Equipped with
The contour candidate setting means resets the adjusted contour candidate as the updated contour candidate, and resets the contour candidate.
The contour adjusting means readjusts the contour candidate in order to approximate the updated contour candidate to the contour of the target structure , and at least a part of the contour candidate or a predetermined peripheral region of the contour candidate. A medical image processing apparatus, characterized in that the contour candidate is moved so that the contour candidate or the peripheral region is included in the contour search region when the contour candidate is not included in the contour search region . A medical image processing device that identifies the contour of a predetermined target structure of a subject in an image.
An area setting means for setting a contour search area in which the contour is searched based on the anatomical characteristics of the structure of the subject, and
Contour candidate setting means for setting contour candidates of the target structure, and
A contour adjusting means for adjusting the contour candidate in order to approximate the contour candidate included in the contour search region to the contour of the target structure, and a contour adjusting means.
Equipped with
The contour candidate setting means resets the contour candidate adjusted in the image of the first resolution as the updated contour candidate in the image having a second resolution different from the first resolution.
The contour adjusting means readjusts the contour candidate in the image of the second resolution in order to approximate the updated contour candidate to the contour of the target structure , and the contour candidate or the contour candidate. When at least a part of the predetermined peripheral region of the above is not included in the contour search region, the contour candidate or the peripheral region is moved so as to be included in the contour search region. Medical image processing equipment. The contour adjusting means determines that the difference between the first contour candidate and the second contour candidate adjusted and reset by the first contour candidate is equal to or less than a predetermined value, or the contour candidate. The medical image processing apparatus according to any one of claims 1 to 11, wherein the contour adjustment is terminated when it is determined that the adjustment has been performed a predetermined number of times. A medical image processing system that identifies the contour of a predetermined target structure of a subject in an image.
Radiation generating means to generate radiation and
Radiation detection means for detecting the radiation and
An area setting means for setting a contour search area in which the contour is searched based on the anatomical characteristics of the structure of the subject, and
Contour candidate setting means for setting contour candidates of the target structure, and
A contour adjusting means for adjusting the contour candidate in order to approximate the contour candidate included in the contour search region to the contour of the target structure, and a contour adjusting means.
Equipped with
The contour adjusting means includes the contour candidate or the peripheral region in the contour search region when at least a part of the contour candidate or a predetermined peripheral region of the contour candidate is not included in the contour search region. As described above, a medical image processing system characterized by moving the contour candidate. A medical image processing method that specifies the contour of a predetermined target structure of a subject in an image.
A step of setting a contour search area in which the contour is searched based on the anatomical characteristics of the structure of the subject, and a step of setting the contour search region.
The process of setting contour candidates for the target structure and
A step of adjusting the contour candidate in order to approximate the contour candidate included in the contour search region to the contour of the target structure, and a step of adjusting the contour candidate.
Equipped with
In the step of adjusting the contour candidate, when at least a part of the contour candidate or a predetermined peripheral region of the contour candidate is not included in the contour search region, the contour candidate or the peripheral region is the contour search region. A medical image processing method, characterized in that the contour candidate is moved so as to be included in . A program for making a computer function as each means of the medical image processing apparatus according to any one of claims 1 to 12.

Images