JP6310118B2 - Image processing apparatus, treatment system, and image processing method - Google Patents

Description

  Embodiments described herein relate generally to a medical image processing apparatus, a treatment system, and a medical image processing method.

  An ambush irradiation method and a tracking irradiation method are known as radiotherapy methods when the affected part of a patient moves due to breathing, heartbeat, intestinal movement, and the like.

  In these irradiation methods, by periodically imaging the patient's body with fluoroscopy X-rays, a fluoroscopic image including an image of the affected part of the patient and a marker placed in the vicinity of the affected part is periodically generated, and a template and By performing the matching, the affected part and the marker are tracked, and the affected part is irradiated with the treatment beam. The template is registered when a user such as a doctor or a radiographer registers in a fluoroscopic image for creating a template generated by imaging a patient's body with fluoroscopic X-rays in advance at the time of treatment planning.

  Alternatively, the marker is tracked by applying a background subtraction method to a fluoroscopic image periodically imaged at the time of treatment, and the affected part is irradiated with the treatment beam.

US Pat. No. 6,307,914 US Patent Application Publication No. 2005/0059887

  As in the prior art described above, in the method in which the user registers an image of an affected area or marker in a fluoroscopic image for creating a template in the template, the template is registered by irradiating the subject with fluoroscopic radiation. It is necessary to generate a fluoroscopic image for creation, and the exposure amount of the subject increases.

  In particular, in order to cope with the difference in the shape of the affected part or marker image in the fluoroscopic image depending on the position of the affected part or marker in the subject, a plurality of fluoroscopic images for the template are taken at different times, and the fluoroscopy for each template is taken. When the image of the affected part or marker in the image is registered in the template, the exposure dose of the subject further increases.

  In addition, in the method of detecting a marker image by applying a background subtraction method to a fluoroscopic image captured during treatment, it is necessary to strongly irradiate fluoroscopic radiation in order to prevent erroneous detection of noise included in the fluoroscopic image. Yes, the exposure dose of the subject increases. In the background subtraction method, it is difficult to detect the affected area itself.

  The problem to be solved by the present invention is to provide a medical image processing apparatus, a treatment system, and a medical image processing method capable of reducing the exposure dose due to fluoroscopic radiation.

  The medical image processing apparatus according to the embodiment includes a calculation unit and a detection unit. The calculation unit is different from the two or more first pixels in the first region including two or more first pixels around the target pixel for each target pixel with respect to the fluoroscopic image seen through the inside of the subject. A second region including one or more second pixels, the pixel values of the two or more first pixels are close to each other, and the pixel values of the two or more first pixels and the one or more second pixels The likelihood that the value increases as the distance from the pixel value increases. The detection unit detects a position of an object inside the subject based on the likelihood of each of the target pixels.

The figure which shows the structural example of the treatment system of this embodiment. Explanatory drawing of the example method of calculating the likelihood of this embodiment. Explanatory drawing of the example of the matching method of the image of the object of this embodiment. Explanatory drawing of the example of the matching method of the image of the object of this embodiment. The timing chart figure which shows the irradiation timing example of the radiation and treatment beam of this embodiment. The flowchart figure which shows the process example of the tracking apparatus of this embodiment. Explanatory drawing of the example of the matching method of the image of the object of the modification 1. FIG. Explanatory drawing of the example of the matching method of the image of the object of the modification 1. FIG. Explanatory drawing of the example of the detection method of the position of the object using the posterior probability of the modification 4. FIG. Explanatory drawing of the example of the timing management method of the modification 10. FIG. The block diagram which shows the hardware structural example of the treatment system of this embodiment and each modification.

  Hereinafter, embodiments will be described in detail with reference to the accompanying drawings.

  FIG. 1 is a diagram illustrating an example of a configuration of a treatment system 1 of the present embodiment. As illustrated in FIG. 1, the treatment system 1 includes a bed 110, first irradiation units 112 and 113, radiation detection units 116 and 117 (an example of a first detection unit), and a tracking device 100 (a medical image processing device). An example), a control unit 123, and a second irradiation unit 120.

  In the treatment system 1 of the present embodiment, since the object inside the subject 111 sleeping on the bed 110 moves three-dimensionally due to the breathing, heartbeat, intestinal movement, etc. of the subject 111, the tracking device 100 Is used to track the object, and based on the tracked object, radiation treatment is performed on the affected part of the subject 111.

  In the present embodiment, a patient is assumed as the subject 111, but is not limited thereto. The object may be an affected part in the body of the subject 111 or may be a marker placed in the vicinity of the affected part. The marker only needs to be made of a material (for example, gold or platinum) that is difficult to transmit the radiations 114 and 115 irradiated from the first irradiation units 112 and 113. For example, the marker has a shape such as a sphere or a cylinder having a diameter of about 2 mm. can do. Radiotherapy includes treatment with X-rays, γ-rays, electron beams, proton beams, neutron beams, and heavy particle beams.

  In this embodiment, an example in which the tracking device 100 is used for radiation therapy will be described, but the application of the tracking device 100 is not limited to this. For example, the tracking device 100 can be used to track the position of a catheter or stent graft in catheter examination or stent graft treatment. Alternatively, it can also be used to detect a marker embedded in a calibration phantom at the time of calibration for obtaining a projection matrix of an imaging system for radiotherapy.

  The first irradiation units 112 and 113 irradiate the subject 111 with radiation 114 and 115, respectively. In the present embodiment, it is assumed that the first irradiation units 112 and 113 are movable. However, the present invention is not limited to this, and may be a fixed type. In the present embodiment, the number of first irradiation units is two, but is not limited thereto, and may be one or three or more.

  The radiations 114 and 115 are used for fluoroscopy inside the subject 111 and can be, for example, X-rays, but are not limited thereto.

  The radiation detection unit 116 detects the radiation 114 transmitted through the subject 111, generates a fluoroscopic image 118 through which the inside of the subject 111 is seen, and outputs the fluoroscopic image 118 together with the timing signal 131 to the tracking device 100. The radiation detection unit 117 detects the radiation 115 transmitted through the subject 111, generates a fluoroscopic image 119 seen through the inside of the subject 111, and outputs it to the tracking device 100 together with the timing signal 132. It is assumed that the fluoroscopic image 118 and the fluoroscopic image 119 have different fluoroscopic directions inside the subject 111.

  The radiation detection units 116 and 117 can be realized by, for example, a combination of an image intensifier and a camera, a combination of a color image intensifier and a camera, a flat panel detector, or the like. In the present embodiment, it is assumed that the radiation detection units 116 and 117 are movable. However, the present invention is not limited to this and may be fixed. In addition, in this embodiment, although the number of radiation detection parts is set to 2, it is not limited to this, Any number may be sufficient if it is the same number as the number of 1st irradiation parts.

  The tracking device 100 tracks an object inside the subject 111, and includes a memory 101 (an example of a storage unit), a calculation unit 102, a detection unit 103 (an example of a second detection unit), and a management unit 104. Are provided. The tracking device 100 can be realized by, for example, a control device including a field programmable gate array (FPGA), a central processing unit (CPU), a random access memory (RAM), and a read only memory (ROM).

  In the memory 101, the fluoroscopic image 118 is written from the radiation detection unit 116 at the timing when the timing signal 131 is notified to the management unit 104, and from the radiation detection unit 117 at the timing when the timing signal 132 is notified to the management unit 104. A fluoroscopic image 119 is written.

  A timing signal 131 for notifying the writing timing of the fluoroscopic image 118 to the memory 101 by the radiation detection unit 116 is sent from the radiation detection unit 116 to the management unit 104, and the memory 101 by the radiation detection unit 117 is sent from the radiation detection unit 117. A timing signal 132 is sent to notify the writing timing of the fluoroscopic image 119.

  The management unit 104 manages the timing of writing the perspective images 118 and 119 to the memory 101 based on the timing signals 131 and 132. In addition, the management unit 104 performs address management of the storage area of the memory 101. In addition, the management unit 104 sends a timing signal 138 to the calculation unit 102 in order to manage the timing of sending information 139 from the memory 101 to the calculation unit 102. Information 139 is perspective images 118 and 119.

  The calculation unit 102 acquires information 139 from the memory 101 based on the timing signal 138. The information 139 is the perspective images 118 and 119 as described above. For each target pixel, the calculation unit 102 calculates, for each target pixel, a first region including two or more first pixels around the target pixel, and one or more second pixels different from the two or more first pixels. The second area to be included is set.

  The shape of the first region is preferably a shape that depends on the shape of the object. Specifically, the shape of the first region is the shape of the object image shown in the fluoroscopic image. For example, if the shape of the object is a sphere, the shape of the object image shown in the fluoroscopic image is a circle, so the shape of the first region is a circle.

  In the present embodiment, the calculation unit 102 acquires object information related to the shape and size of an object and a projection matrix that projects three-dimensional coordinates onto the perspective image, and uses the acquired object information and projection matrix to generate a perspective image. The shape and size of the image of the captured object are calculated, and the first region is set with the shape and size of the image. The calculation unit 102 can determine the shape or type of the object based on, for example, information designating a mode input by the user in advance.

  For example, when the object is an affected part of the subject 111, the shape and size of the affected part can be specified from a CT image taken at the time of treatment planning or a CT image taken at the time of diagnosis before the time of treatment planning. However, it is assumed that a three-dimensional contour of the affected area is input to the CT image by a user such as a doctor. The treatment plan is a plan for calculating where and how many treatment beams are delivered.

  For this reason, when the object is an affected part, the calculation unit 102 acquires a CT image from the treatment planning apparatus (not shown in FIG. 1), and acquires a projection matrix of a fluoroscopic image from the control unit 123 described later. The shape and size of the image of the affected area shown in the fluoroscopic image may be calculated using the CT image and the projection matrix, and the first region may be set based on the shape and size of the image.

  In addition, for example, when the object is a marker placed near the affected part of the subject 111, the shape and size of the marker are known, so a user such as a doctor or a radiographer can input from an input device (not shown in FIG. 1). What is necessary is just to input the shape and size of a marker.

  Therefore, when the object is a marker, the calculation unit 102 acquires the shape and size of the marker from the input device, acquires the projection matrix of the perspective image from the control unit 123 described later, and acquires the shape and size of the acquired marker. And the projection matrix are used to calculate the shape and size of the marker image shown in the fluoroscopic image, and the first region may be set based on the shape and size of the image.

  Note that, even when the object is a marker, the calculation unit 102 calculates the shape and size of the marker image that appears in the fluoroscopic image using a CT image or the like, as in the case where the object is the affected part. The first area may be set by size.

  In addition, when the image of the object is dominant with respect to the first area, such as an area around the first area, the calculation unit 102 determines a region where the image of the object is not dominant, that is, other than the image of the object. A dominant region may be set as the second region. Although it is preferable that the first region and the second region are adjacent to each other, the present invention is not limited to this, and even if the first region and the second region are separated, the first region and the second region may overlap. it can.

  Then, for each pixel of interest, the calculation unit 102 has pixel values of two or more first pixels constituting the first region close to each other and one or more pixel values of the two or more first pixels and one or more constituting the second region. The likelihood that the value becomes larger as the pixel value of the second pixel is further away is calculated. The likelihood can be, for example, a correlation ratio between the pixel value of two or more first pixels and the pixel value of one or more second pixels, but is not limited thereto.

  When the shape of the object image shown in the fluoroscopic image differs depending on the direction in which the object is seen through, the calculation unit 102 adjusts the first area and the second area for each target pixel according to the shape of the object image that can be taken. Are set, a likelihood is calculated for each of the set first region and second region, and the likelihood having the largest value may be used as the likelihood of the target pixel.

  In addition, when the shape of the object image shown in the fluoroscopic image differs depending on the direction of seeing through the object, the calculation unit 102 may set the dominant area as the first area for any object image. .

  In this embodiment, the calculation unit 102 sets the first region and the second region for each pixel of interest for each of the perspective images 118 and 119, calculates the likelihood 135, and outputs the likelihood 135 to the detection unit 103. . In the present embodiment, each pixel of the perspective images 118 and 119 is the target pixel. That is, in the present embodiment, the calculation unit 102 calculates the likelihood 135 for each pixel of the perspective images 118 and 119 and outputs the likelihood 135 to the detection unit 103.

  FIG. 2 is an explanatory diagram illustrating an example of a likelihood calculation method according to this embodiment. In the example illustrated in FIG. 2, the likelihood calculation method will be described using the target pixel 200 of the perspective image 118 as an example, but the likelihood calculation method is also applied to other target pixels of the perspective image 118 and the target pixel of the perspective image 119. Is the same. In the example shown in FIG. 2, it is assumed that the shape of the object is a sphere.

  In the example illustrated in FIG. 2, since the shape of the object is a sphere and the shape of the image of the object shown in the fluoroscopic image 118 is a circle, the calculation unit 102 has a circular shape centered on the target pixel 200 with respect to the fluoroscopic image 118. The first area 201 is set, and the inside of the rectangle including the first area 201 and the outside of the first area 201 are set as the second area 202. In this case, the pixels constituting the first area 201 are the first pixels, and the pixels constituting the second area 202 are the second pixels.

In the case of the example illustrated in FIG. 2, the likelihood of the pixel of interest 200 is represented by σ _b ² / σ _t ² . σ _t indicates the standard deviation of the values of all the pixels included in the union set of the first region 201 and the second region 202 (the union set of the first pixel and the second pixel), and σ _b is expressed by Equation (1). expressed.

Here, n ₁ represents the number of pixels of the first pixel constituting the first region 201, n ₂ represents the number of pixels of the second pixel constituting the second region 202, and n _t represents the first region. 201 shows the number of pixels included in the union of 201 and the second region 202 (the union of the first pixel and the second pixel). m ₁ represents an average of the pixel values of the first pixels constituting the first region 201, m ₂ represents the average of the pixel values of the second pixel constituting the second region 202, m _t is The average value of the pixel values included in the union set of the first region 201 and the second region 202 (the union set of the first pixel and the second pixel) is shown.

That is, σ _b ² / σ _t ² is a correlation ratio between a set of pixel values of the first pixels constituting the first area 201 and a set of pixel values of the second pixels constituting the second area 202.

Σ _b ² / σ _t ² takes a value from 0 to 1, and the pixel values of the first pixels constituting the first region 201 are close to each other, and the pixel values of the first pixel constituting the first region 201 And the value of the second pixel constituting the second region 202 increases as the distance between them increases (the value approaches 1).

Therefore, when the image of the object is dominant with respect to the first region 201 and the image other than the object is dominant with respect to the second region 202, the value of σ _b ² / σ _t ² becomes large. On the other hand, if the image of the same composition becomes dominant in both the first region 201 and the second region 202, the value of σ _b ² / σ _t ² decreases (the value approaches 0).

In particular, in the present embodiment, since the first area is set with the shape and size of the object image (circle in the example shown in FIG. 2) in the perspective image, the object image and the first area 201 match. In this case, the value of σ _b ² / σ _t ² is significantly increased as compared with the case where the object image and the first region 201 do not match.

If σ _w ² = σ _t ² + σ _b ² is defined, Equation (2) is satisfied, and therefore σ _b ² / σ _t ² is also calculated by 1 / ((σ _w ² / σ _b ² ) +1). it can. Since the denominator σ _t ² of σ _b ² / σ _t ² can be 0, σ _b ² / (σ _t ² + ε) may be set as the likelihood of the target pixel 200. Note that ε is a small positive value. Since the denominator σ _b ² of σ _w ² / σ _b ² can be 0, 1 / ((σ _w ² / (σ _b ² + ε)) + 1) may be set as the likelihood of the target pixel 200.

In the example shown in FIG. 2, the likelihood (σ _b ² / σ _t ² ) of the target pixel 200 is set as the set of pixel values of the first pixel that forms the first region 201 and the second that forms the second region 202. Although the correlation ratio with the set of pixel values of the pixels is used, if the pixel values of the first pixel are closer to each other and the pixel values of the first pixel and the second pixel are farther apart, The correlation ratio may not be used.

  For example, the pixel values of the first pixels constituting the first region 201 are close to each other, the pixel values of the first pixels indicate that the radiation transmittance is low, and the pixel values of the first pixel and the second pixel The likelihood that the value increases as the distance between and increases.

However, in a fluoroscopic image, the density may be reversed, and the pixel value may be small or large when the radiation transmittance is low, so the value is large when the radiation transmittance is low. It is necessary to design the likelihood according to the specifications of the fluoroscopic image. Therefore, in the case of a transmission image in which the pixel value is small when the radiation transmittance is low, the likelihood when m ₁ is less than a predetermined threshold is set to 0 or a small positive value ε, and m ₁ is equal to or greater than the predetermined threshold. In this case, the likelihood may be σ _b ² / σ _t ² .

  The detection unit 103 detects the position of the object inside the subject 111 based on the likelihood of each target pixel of the fluoroscopic image calculated by the calculation unit 102. Specifically, the detection unit 103 detects that the image of the object is located at a target pixel having a likelihood that the value satisfies a predetermined condition.

  In the present embodiment, the detection unit 103 acquires a projection matrix for projecting the three-dimensional coordinates onto the perspective image 118 and a projection matrix for projecting the three-dimensional coordinates onto the perspective image 119, and each of the target pixels for each of the perspective images 118 and 119. Among the likelihoods 135, a target pixel having a likelihood that the magnitude of the value satisfies a predetermined condition is detected.

  Thereby, the detection unit 103 detects that the image of the object is positioned on the detected target pixel for each of the perspective images 118 and 119. Note that the projection matrices of the perspective images 118 and 119 can be acquired from the control unit 123 described later. Further, the predetermined condition may be that the value is the largest, the value exceeds the threshold, or the value exceeds the threshold and is the largest.

  Then, the detection unit 103 detects the three-dimensional position of the object using the projection matrix of each of the perspective images 118 and 119 and the target pixel detected from each of the perspective images 118 and 119. For example, the detection unit 103 calculates the three-dimensional position of the object from the pixel of interest (image of the object) detected from each of the perspective images 118 and 119 by epipolar geometry using the projection matrix of each of the perspective images 118 and 119, Position information 122 indicating the calculated three-dimensional position is output to the control unit 123.

  When there are a plurality of objects inside the subject 111, the detection unit 103 acquires, for each of the perspective images 118 and 119, region information indicating a third region where each of the plurality of object images is expected to exist. .

  For example, in the above-described CT image, if a third region for each object is input by a user such as a doctor, the detection unit 103 may transmit the fluoroscopic image 118, from the above-described treatment planning apparatus (not shown in FIG. 1). A CT image corresponding to each 119 is acquired as region information.

  For example, if a third region for each object is input to the fluoroscopic images 118 and 119 by a user such as a doctor or a radiographer, the detection unit 103 acquires the fluoroscopic images 118 and 119 from the calculation unit 102. Thus, the area information is acquired.

  In this case, with respect to the perspective images 118 and 119 displayed on the display device (not shown in FIG. 1), the user designates a point by clicking the mouse or touching the touch panel, and the region around the designated point is changed to the first region. Three regions may be used. In this way, even if the perspective images 118 and 119 are reproduced as moving images, the user can easily input the third region.

  Next, for each of the perspective images 118 and 119, the detection unit 103 detects a plurality of target pixels having a likelihood that the magnitude of the value satisfies a predetermined condition from the likelihood 135 of each target pixel. The predetermined condition is that the rank of the value size falls within the upper object number, the value exceeds the threshold value, or the value exceeds the threshold value, and the rank of the value size falls within the upper object number. I can do it.

  Next, the detection unit 103 uses the region information of the perspective image 118 to detect in which third region each of the plurality of target pixels detected from the perspective image 118 exists, and uses the region information of the perspective image 119. By detecting in which third region each of the plurality of pixels of interest detected from the perspective image 119 exists, the plurality of pixels of interest (images of a plurality of objects) detected in the respective perspective images 118 and 119 are seen through. The images 118 and 119 are associated with each other.

  Next, the detection unit 103 detects a plurality of pixels of interest (images of a plurality of objects) detected from the projection matrixes of the perspective images 118 and 119 and the perspective images 118 and 119, respectively, and correlated between the perspective images 118 and 119. Used to detect the three-dimensional position of a plurality of objects.

  3A and 3B are explanatory diagrams illustrating an example of an object image association method according to the present embodiment. In the example shown in FIGS. 3A and 3B, the number of objects is three. As shown in FIG. 3A, the target image 301, 302, 303 in which the three object images are located is detected in the perspective image 118. As shown in FIG. 3B, in the perspective image 119, target pixels 311, 312, and 313 in which images of three objects are located are detected.

  In this case, the detection unit 103 detects in which third region each of the target pixels 301, 302, and 303 detected from the perspective image 118 using the region information of the perspective image 118, and It is detected in which third region each of the target pixels 311, 312, and 313 detected from the perspective image 119 exists using the region information.

  For example, the region information of the perspective images 118 and 119 indicates the third region with ID = 1 to 3, respectively. Here, the target pixel 301 is in the third region with ID = 1 of the perspective image 118. The target pixel 302 exists in the third region with ID = 2 of the perspective image 118, the target pixel 303 exists in the third region with ID = 3 of the perspective image 118, and the target pixel 311 exists in the perspective image The target pixel 312 exists in the third region of ID = 2 of the fluoroscopic image 119, and the target pixel 313 exists in the third region of ID = 3 of the fluoroscopic image 119. It shall exist. Note that, in the perspective images 118 and 119, the third regions having the same ID are the same region.

  Therefore, the detection unit 103 associates the target pixel 301 and the target pixel 311, the target pixel 302 and the target pixel 312, and the target pixel 303 and the target pixel 313 between the perspective images 118 and 119.

  Thereby, even when there are a plurality of objects, images of the same object can be associated between the perspective images 118 and 119, and the three-dimensional position of the object can be calculated.

  The control unit 123 outputs a control signal 128 to the second irradiation unit 120 based on the position information 122 from the detection unit 103, and controls irradiation of the treatment beam 121 to the object (affected site) by the second irradiation unit 120. .

  The control signal 128 is a signal for controlling the irradiation timing of the treatment beam 121 by the second irradiation unit 120 in the case of the ambush irradiation method, and is position information of the object (affected part) in the case of the tracking irradiation method.

  In addition, the control unit 123 outputs timing signals 124, 125, 126, and 127 to the first irradiation units 112 and 113 and the radiation detection units 116 and 117, respectively, and the first irradiation units 112 and 113, and the radiation detection unit 116, respectively. 117, the operation timing is controlled.

  In addition, the control unit 123 controls the first irradiation units 112 and 113 and the radiation detection units 116 and 117 to perform test generation (imaging) of the perspective images 118 and 119, thereby projecting the projection matrix of the perspective images 118 and 119. Is generated and held in advance.

  The control unit 123 can be realized by, for example, a control device including a CPU, a RAM, a ROM, and the like. When the tracking device 100 is also realized by a control device, the tracking device 100 and the control unit 123 may be realized by the same control device or may be realized by different control devices.

  The second irradiation unit 120 irradiates the treatment beam 121 to the object (affected part) based on the three-dimensional position of the object detected by the detection unit 103. Specifically, the second irradiation unit 120 irradiates the treatment beam 121 to the object (affected part) based on the control signal 128 from the control unit 123.

  In the present embodiment, it is assumed that the treatment beam 121 is a heavy particle beam, but is not limited thereto, and may be an X-ray, a γ-ray, an electron beam, a proton beam, a neutron beam, or the like. .

  FIG. 4 is a timing chart showing an example of irradiation timings of the radiations 114 and 115 and the treatment beam 121 according to this embodiment.

  In the present embodiment, when the radiations 114 and 115 and the treatment beam 121 are irradiated simultaneously, the fluoroscopic images 118 and 119 become noisy due to the influence of the scattering of the radiation. 1st irradiation parts 112 and 113 and the 2nd irradiation part 120 are controlled so that may be irradiated independently.

  For example, when performing irradiation 30 times from each direction in 1 second, as shown in FIG. 4, the control unit 123 irradiates the radiation 114, 115 and the treatment beam 121 independently every 1/30 seconds. The first irradiation units 112 and 113 and the second irradiation unit 120 are controlled. Times 401 and 402 indicate 1/30 seconds.

  FIG. 5 is a flowchart illustrating an example of processing of the tracking device 100 according to the present embodiment.

  First, the fluoroscopic image 118 is written from the radiation detection unit 116 and the fluoroscopic image 119 is written from the radiation detection unit 117 to the memory 101 (step S101).

  Subsequently, the calculation unit 102 reads information 139 (the perspective images 118 and 119) from the memory 101, sets the first region and the second region for each target pixel for each of the perspective images 118 and 119, and the likelihood. The degree 135 is calculated (step S103).

  Subsequently, for each of the perspective images 118 and 119, the detection unit 103 detects a target pixel having a likelihood that the value satisfies a predetermined condition, and the projection matrix of each of the perspective images 118 and 119 and the perspective images 118 and 119, respectively. The three-dimensional position of the object is detected using the target pixel detected from (Step S105).

  When the tracking of the three-dimensional position of the object is continued (No in step S107), the process returns to step S101, and when the tracking of the three-dimensional position of the object is ended (Yes in step S107), the process is ended.

  As described above, according to the present embodiment, since the position of the object can be detected from the fluoroscopic image generated at the time of treatment, it is not necessary to perform fluoroscopy in advance for detecting the position of the object such as template registration. The exposure dose of the subject due to radiation can be reduced.

  In addition, when the correlation ratio is used for the likelihood as in this embodiment, since the likelihood is robust to noise, the object can be detected even if the intensity of the fluoroscopic radiation is weakened, and the subject due to the fluoroscopic radiation Can be reduced.

(Modification 1)
In the above embodiment, when there are a plurality of objects in the subject 111, the detection unit 103 associates the images of the same object between the perspective images 118 and 119 using the area information of the perspective images 118 and 119, respectively. However, in the first modification, an example in which images of the same object are associated between the perspective images 118 and 119 by epipolar geometry will be described.

  In this case, the detection unit 103 converts the plurality of target pixels (images of a plurality of objects) detected in the perspective images 118 and 119 by the epipolar geometry using the projection matrices of the perspective images 118 and 119, respectively. 119 may be associated with each other.

  6A and 6B are explanatory diagrams of an example of an object image association method according to the first modification. In the example shown in FIGS. 6A and 6B, the number of objects is three. As shown in FIG. 6A, the target image 301, 302, 303 in which the three object images are located is detected in the perspective image 118. As shown in FIG. 6B, in the perspective image 119, target pixels 311, 312, and 313 in which three object images are respectively detected are detected.

  In the example illustrated in FIGS. 6A and 6B, a method of associating the target pixel 302 of the perspective image 118 will be described. However, the target pixels 301 and 303 of the fluoroscopic image 118 can also be associated by the same method.

  The detection unit 103 calculates the epipolar line 602 of the perspective image 119 corresponding to the target pixel 302 using the projection matrix of each of the target pixel 302 and the perspective images 118 and 119. Then, the detection unit 103 calculates the projection length (perpendicular length) of each of the target pixels 311, 312, and 313 with respect to the epipolar line 602, and associates the target pixel 312 with the shortest projection length with the target pixel 302.

  Thereby, even when there are a plurality of objects, images of the same object can be associated between the perspective images 118 and 119, and the three-dimensional position of the object can be calculated.

  In the example shown in FIGS. 6A and 6B, since the difference between the minimum projection length and the second smallest projection length is large, there is little possibility of incorrect association, but the minimum projection length and the second smallest projection length. If the difference between is small, there is a possibility that the correspondence is incorrect. In this case, the association determination may be postponed to the next frame.

(Modification 2)
In the above embodiment, the likelihood is calculated using each pixel of the fluoroscopic image as the target pixel. However, in Modification 2, each pixel constituting the third region in which the presence of the image of the object is expected is the target pixel in the fluoroscopic image. An example of calculating the likelihood will be described.

  In this case, the calculation unit 102 acquires area information indicating a third area where an object image is expected to be present in the perspective image, and calculates the likelihood for each target pixel in the third area indicated by the area information. You can do it. The area information may be acquired by the same method as in the above embodiment.

  According to the modified example 2, since it is sufficient to calculate the likelihood with some pixels of the fluoroscopic image, the likelihood calculation time can be shortened. As a result, it is possible to increase the probability that the calculation of the likelihood is finished by the timing when the treatment beam should be irradiated, reduce the situation where the treatment beam cannot be irradiated without the calculation of the likelihood, and generate a fluoroscopic image (imaging) ) The number of times can be reduced, and the exposure amount of the subject by fluoroscopic radiation can be reduced. In addition, the time from the generation (imaging) of the fluoroscopic image to the treatment beam irradiation can be shortened.

  Note that the technique of Modification 2 is more effective as the perspective image frame rate is higher.

(Modification 3)
In the above-described embodiment, the likelihood is calculated using each pixel of the fluoroscopic image as a target pixel. However, in Modification 3, once the image of the object is detected in the fluoroscopic image, the likelihood of the object is determined based on a predetermined motion model. An example will be described in which a fourth region (a destination of an object image) where an image is expected is predicted, and likelihood is calculated using each pixel constituting the predicted fourth region as a target pixel.

  In this case, the calculation unit 102 predicts a fourth region in which an object image is expected to exist in the fluoroscopic image based on the predetermined motion model, and calculates the likelihood for each target pixel in the fourth region. You can do it. The predetermined motion model is, for example, a constant velocity motion model or a constant acceleration motion model.

  According to the modified example 3, since it is sufficient to calculate the likelihood with a part of the pixels of the fluoroscopic image, the likelihood calculation time can be shortened. As a result, it is possible to increase the probability that the calculation of the likelihood is finished by the timing when the treatment beam should be irradiated, reduce the situation where the treatment beam cannot be irradiated without the calculation of the likelihood, and generate a fluoroscopic image (imaging) ) The number of times can be reduced, and the exposure amount of the subject by fluoroscopic radiation can be reduced. In addition, the time from the generation (imaging) of the fluoroscopic image to the treatment beam irradiation can be shortened.

  Note that the technique of Modification 3 is more effective as the frame rate of the fluoroscopic image is higher.

(Modification 4)
In the modified example 3, the likelihood is calculated in the fourth region predicted based on the predetermined motion model. However, in the modified example 4, in addition to the likelihood of the fourth region, the prior probability based on the predetermined motion model is also used. An example of detecting the position of will be described.

  In this case, the detection unit 103 sets, in the fourth region, a prior probability that the value increases as the position of the object image moves based on the predetermined motion model, and based on the likelihood and the prior probability of each pixel of interest. Thus, the position of the object may be detected. For example, the detection unit 103 may calculate a posterior probability obtained by multiplying the likelihood of each pixel of interest and the prior probability, and detect the position of the object based on the calculated posterior probability.

  FIG. 7 is an explanatory diagram illustrating an example of a method for detecting the position of an object using the posterior probability according to the fourth modification. In the example illustrated in FIG. 7, the target pixel 701 of the perspective image 118 is detected as the position of the object image in a certain frame, and the object image is predicted to move in the direction of the arrow 702 based on a predetermined motion model in the next frame. The prior probability distribution 703 of each pixel constituting the fourth region is set in the fourth region. The distribution 703 may be set as a two-dimensional normal distribution, for example.

  The distribution 703 is a distribution that has the largest value at a position where the image of the object is predicted to move according to the predetermined motion model and decreases as the distance from the position is increased. However, even if the distance from the position is the same, the distribution 703 may be a distribution in which the value along the direction of the arrow 702 is larger than the position in the direction perpendicular to the direction of the arrow 702. .

  In addition, when an image of the object is detected in one of the perspective images 118 and 119, a position where the object image is predicted to move according to a predetermined motion model is set to a prior probability distribution set in the fourth region of the other. The value is the largest and the value decreases as the distance from the position increases. Even if the distance from the position is the same, the position along the direction of the arrow 702 is a position in the direction perpendicular to the direction of the arrow 702. It is good also as distribution where a value is larger than that and a value along the epipolar line is larger than a position in a direction perpendicular to the epipolar line.

  According to the fourth modification, it is possible to realize tracking that is robust against noise of a fluoroscopic image. As a result, since the tracking can be performed even if the radiation dose for fluoroscopy is lowered, the exposure dose of the subject by the fluoroscopy radiation can be reduced.

(Modification 5)
In the modified example 4, the position of the object is detected based on the posterior probability obtained by multiplying the likelihood of each pixel of interest and the prior probability, but the posterior probability of the next frame is determined by using the posterior probability of a certain frame as the prior probability of the next frame. And the position of the object may be detected based on the calculated posterior probability.

  According to the fifth modification, it is possible to realize tracking that is robust against noise in a fluoroscopic image. As a result, since the tracking can be performed even if the radiation dose for fluoroscopy is lowered, the exposure dose of the subject by the fluoroscopy radiation can be reduced.

(Modification 6)
In the third modification, the likelihood is calculated in the fourth region predicted based on the predetermined motion model and the position of the object is detected. However, the position where the posterior probability is maximized may be detected by a particle filter.

  According to the modified example 6, it is possible to realize tracking that is robust against noise of a fluoroscopic image. As a result, since the tracking can be performed even if the radiation dose for fluoroscopy is lowered, the exposure dose of the subject by the fluoroscopy radiation can be reduced.

(Modification 7)
In the fourth modification, when there are a plurality of objects in the subject 111, the kurtosis of the prior probability distribution 703 may be changed according to the distance between the objects.

  In this case, the calculation unit 102 predicts a fourth area where each of the plurality of object images is expected to exist in the fluoroscopic image based on the predetermined motion model, and the likelihood of each target pixel in each of the plurality of fourth areas. The degree may be calculated.

  In addition, for each image of the object, the detection unit 103 increases the value as the image moves closer to the position where the image moves based on the predetermined motion model, and increases the kurtosis when the distance between the plurality of images is equal to or less than the threshold value. The prior probability is set in the fourth region where the image is expected to exist, and the position of the object may be detected for each fourth region based on the likelihood of each pixel of interest and the prior probability.

  According to the modified example 7, when the distance between the plurality of images is equal to or less than the threshold value, the predicted position based on the predetermined motion model from the past frame has priority over the likelihood based on the appearance of the perspective image of the current frame. Therefore, even when a plurality of images intersect, the possibility of mistaking the images can be reduced. As a result, since the number of frames in which the image of the object cannot be tracked is reduced, the number of generation (imaging) of fluoroscopic images can be reduced, and the exposure dose of the subject due to fluoroscopic radiation can be reduced.

(Modification 8)
In the modified example 7, when the distance between the object images is equal to or smaller than the threshold value, the possibility of mistaking the object image by increasing the kurtosis of the prior probability distribution is reduced. An example in which the position of at least one of the first irradiation units 112 and 113 and the radiation detection units 116 and 117 is controlled so that the distance between the object images does not become the threshold value or less when the distance is equal to or less than the threshold value will be described. To do. However, in the modification 7, the 1st irradiation parts 112 and 113 and the radiation detection parts 116 and 117 need to be movable.

  In this case, the detection unit 103 sets, for each image of the object, a prior probability that the value increases as the image moves closer to the position based on the predetermined motion model in the fourth region where the image is expected to exist. For each fourth region, the position of the object is detected based on the likelihood and the prior probability of each pixel of interest.

  When the distance between the plurality of images is equal to or smaller than the threshold, the detection unit 103 generates the first irradiation units 112 and 113, and the radiation detection unit 116 that generate the fluoroscopic images so that the distance between the plurality of images exceeds the threshold. The position information of at least one of 117 may be calculated.

  For example, for each of the first irradiation units 112 and 113 and the radiation detection units 116 and 117, the detection unit 103 obtains projection matrices corresponding to several possible positions in the movable position, and the three-dimensional coordinates of each object and each From the projection matrix, the image coordinates of the image of each object in the fluoroscopic images 118 and 119 that are captured according to the positions of the first irradiation units 112 and 113 and the radiation detection units 116 and 117 are calculated. Then, the detection unit 103 may extract the position information of the first irradiation units 112 and 113 and the radiation detection units 116 and 117 where the distance between the object images exceeds the threshold (preferably, the maximum).

  Further, the detection unit 103 is a perspective image that is imaged according to the positions of the first irradiation units 112 and 113 and the radiation detection units 116 and 117 from the predicted position according to the predetermined motion model of the three-dimensional coordinates of each object and each projection matrix. The image coordinates of the image of each object at 118 and 119 may be calculated.

  The control unit 123 may control the position of at least one of the first irradiation units 112 and 113 and the radiation detection units 116 and 117 based on the position information calculated by the detection unit 103.

  According to the modified example 8, since a plurality of images do not intersect with each other, the images are not mixed up. As a result, since the number of frames in which the image of the object cannot be tracked is reduced, the number of generation (imaging) of fluoroscopic images can be reduced, and the exposure dose of the subject due to fluoroscopic radiation can be reduced.

(Modification 9)
In the modified example 3, the likelihood is calculated using each pixel constituting the fourth region predicted based on the predetermined motion model as the target pixel. However, the likelihood is calculated by further limiting the fourth region with an epipolar line. May be.

  For example, when an object image is detected from the perspective image 118, the object image also exists on the epipolar line of the perspective image 119. Even if the distortion of the fluoroscopic image cannot be corrected, an image of the object exists near the epipolar line.

  Therefore, the calculation unit 102 may calculate the likelihood for each target pixel in the fourth region predicted in the perspective image 119 and in the vicinity of the epipolar line.

  According to the modified example 9, since it is sufficient to calculate the likelihood with a part of the pixels of the fluoroscopic image, the likelihood calculation time can be shortened. As a result, it is possible to increase the probability that the calculation of the likelihood is finished by the timing when the treatment beam should be irradiated, reduce the situation where the treatment beam cannot be irradiated without the calculation of the likelihood, and generate a fluoroscopic image (imaging) ) The number of times can be reduced, and the exposure amount of the subject by fluoroscopic radiation can be reduced. In addition, the time from the generation (imaging) of the fluoroscopic image to the treatment beam irradiation can be shortened.

  Note that the method of Modification 9 is more effective as the frame rate of the fluoroscopic image is higher.

(Modification 10)
In the above embodiment, the perspective image is read from the memory 101 after the writing to the perspective image in the memory 101 is completed, and the calculation of the likelihood is started. However, in the modified example 10, the likelihood is calculated first. An example in which the calculation of the likelihood is started after the timing when all the pixels necessary for the calculation of the likelihood at the target pixel are written in the memory will be described.

  In this case, the pixels constituting the fluoroscopic image are written in the memory 101 in a predetermined order.

  In addition, the management unit 104 manages the timing at which all the pixels necessary for calculating the likelihood at the target pixel for which the first calculation of the likelihood is performed by the calculation unit 102 are written in the memory 101, and after that timing, Start reading.

  For example, in the case of the fluoroscopic image 118, the management unit 104 can manage up to which pixel value of the fluoroscopic image 118 has been received by the number of clocks after receiving the timing signal 131. Based on the number of clocks, the timing at which all the pixels necessary for calculating the likelihood at the target pixel for which the likelihood is first calculated by the calculation unit 102 is written in the memory 101 is managed.

  FIG. 8 is an explanatory diagram of an example of the timing management method of the tenth modification. In the example shown in FIG. 8, each pixel of the perspective image 118 is written in the memory 101 in the raster scan order from the upper left. In the example shown in FIG. 8, the likelihood is first calculated by the calculation unit 102 in the pixel of interest 800, and all the pixels necessary for calculating the likelihood in the pixel of interest 800 form the first region 801. This is a second pixel constituting one pixel and the second region 802.

  Therefore, the management unit 104 manages the timing at which all of the first pixel and the second pixel are written in the memory 101 by the number of clocks after receiving the timing signal 131, and is necessary for calculating the likelihood at the target pixel 800. The timing at which all the pixels are written in the memory 101 is managed.

  When the fluoroscopic image is read after the timing, the calculation unit 102 starts calculating the likelihood for each pixel of interest.

  According to the modified example 10, since the likelihood calculation start time can be shortened, the likelihood calculation end time can also be shortened, and the likelihood calculation time can be shortened. As a result, it is possible to increase the probability that the calculation of the likelihood is finished by the timing when the treatment beam should be irradiated, reduce the situation where the treatment beam cannot be irradiated without the calculation of the likelihood, and generate a fluoroscopic image (imaging) ) The number of times can be reduced, and the exposure amount of the subject by fluoroscopic radiation can be reduced. In addition, the time from the generation (imaging) of the fluoroscopic image to the treatment beam irradiation can be shortened.

(Modification 11)
In the above embodiment, all the pixels of the fluoroscopic image are written in the memory 101, but the pixels necessary for calculating the likelihood may be written in the memory 101.

  According to the eleventh modification, the transmission time of the fluoroscopic image can be shortened. As a result, it is possible to increase the probability that the calculation of the likelihood is finished by the timing when the treatment beam should be irradiated, reduce the situation where the treatment beam cannot be irradiated without the calculation of the likelihood, and generate a fluoroscopic image (imaging) ) The number of times can be reduced, and the exposure amount of the subject by fluoroscopic radiation can be reduced. In addition, the time from the generation (imaging) of the fluoroscopic image to the treatment beam irradiation can be shortened.

  Note that the technique of the modification 11 is more effective as the frame rate of the fluoroscopic image is higher.

(Modification 12)
In the above embodiment, the image of all the color components is the perspective image, but the image of the color component corresponding to the transmittance of the object may be the perspective image.

  For example, when the radiation detection units 116 and 117 are realized by a combination of a color image intensifier and a camera that captures a Bayer RAW image using a Bayer type color filter, a specific color component of the Bayer RAW image is used. May be the perspective images 118 and 119.

  In the combination of the color image intensifier and the Bayer type camera, although the transmittance is low, the sensitivity is the highest in the R (red) component, while the sensitivity is high, but the sensitivity is the highest in the B (blue) component. For this reason, when the object is a marker with low transmittance, the R component of the Bayer RAW image is set as the perspective images 118 and 119, and when the object is an affected part with low transmittance, the G component or B component of the Bayer RAW image is set as The perspective images 118 and 119 may be used.

  In addition, the Bayer RAW image may be demosaiced, and specific color components may be used as the perspective images 118 and 119. In addition, when not demosaicing, the number of pixels of the color component is small, so that the image processing time can be shortened.

  For example, when the radiation detection units 116 and 117 are realized by a combination of a color image intensifier and a camera that captures only a specific color component, the image of the specific color component may be used as the perspective images 118 and 119. Good.

  For example, when the object is a marker having a low transmittance, a camera using a color filter that transmits the R component may be used, and images including only the R component may be used as the perspective images 118 and 119. In addition, when the object is an affected part with low transmittance, a camera using a color filter that transmits the G component or the B component may be used, and images including only the G component or the B component may be used as the perspective images 118 and 119.

  Since the image is blurred by the demosaicing interpolation process, in this case, an image with less blur can be obtained than the image of only the color component of the image obtained by demosaicing the Bayer RAW image.

  According to the modified example 12, it is possible to reduce the probability that the tracking of the object fails. As a result, since the number of frames in which the image of the object cannot be tracked is reduced, the number of generation (imaging) of fluoroscopic images can be reduced, and the exposure dose of the subject due to fluoroscopic radiation can be reduced.

(Modification 13)
In the above embodiment, when the likelihood that the magnitude of the value satisfies the predetermined condition is not calculated, the detection unit 103 outputs a warning signal to the control unit 123, and the control unit 123 outputs a warning based on the warning signal. You may make it perform control which makes an apparatus (illustration omitted in FIG. 1) warn.

  In this case, the image of the object cannot be clearly seen in the perspective images 118 and 119, and there is a possibility that an erroneous detection of the position of the object may occur. For example, when a bone having low transmittance and an object overlap the lines of radiation 114 and 115, the image of the object cannot be clearly seen.

  According to the modified example 13, since a user such as a doctor or a radiologist is warned that such a state is present, the user can increase the intensity of the radiation 114, 115 or change the irradiation direction of the radiation 114, 115. Such a state can be resolved at an early stage. As a result, since the number of frames in which the image of the object cannot be tracked is reduced, the number of generation (imaging) of fluoroscopic images can be reduced, and the exposure dose of the subject due to fluoroscopic radiation can be reduced.

  When the likelihood that the magnitude of the value satisfies the predetermined condition is not calculated, the detection unit 103 outputs a control signal for increasing the intensity of the radiation to the control unit 123, and the control unit 123 is based on the control signal. Thus, the first irradiation units 112 and 113 may be controlled to increase the intensity of radiation.

  In this way, since the image of the object can be clearly seen in any frame of the perspective images 118 and 119, the state where the image of the object cannot be clearly seen can be eliminated at an early stage. As a result, since the number of frames in which the image of the object cannot be tracked is reduced, the number of generation (imaging) of fluoroscopic images can be reduced, and the exposure dose of the subject due to fluoroscopic radiation can be reduced.

(Modification 14)
In the above embodiment, the likelihood value may be generally increased around the image of the object. Therefore, if the likelihood is high in the vicinity, merging may be performed. This eliminates the mistake of originally detecting one object as a plurality.

(Modification 15)
In the above embodiment, the likelihood that takes a large value in the part that seems to be an object is calculated, and the part that has a large likelihood is detected as an object. May be detected as an object. The feature amount may be designed such that the smaller the pixel value of the first pixel is, and the smaller the pixel value of the first pixel and the pixel value of the second pixel are, the smaller the value is. For example, σ _t ² / σ _b ² or σ _w ² / σ _b ² can be used as the feature amount.

(Hardware configuration)
FIG. 9 is a block diagram illustrating an example of a hardware configuration of the treatment system 1 of the present embodiment and each modification. Up to this point, the treatment beam has been described on the assumption of a heavy particle beam, but FIG. 9 is a block diagram in the case of an X-ray. As illustrated in FIG. 9, the treatment system 1 includes a console 10, an imaging device 20, a bed device 30, a treatment plan device 40, and a radiation treatment device (linac: radiation that performs treatment by irradiating radiation based on treatment plan data. Treatment apparatus) 50. Although an example in which the imaging apparatus 20 is a CT apparatus will be described, the present invention is not limited to this.

  The console 10 corresponds to the tracking device 100 and the control unit 123, the bed device 30 corresponds to the bed 110, the treatment planning device 40 corresponds to the treatment planning device not shown in the above embodiment, and the radiation treatment device. 50 corresponds to the second irradiation unit 120.

  The imaging device 20, the bed device 30, and the radiation therapy device 50 are usually installed in an examination room. On the other hand, the console 10 is usually installed in a control room adjacent to the examination room. The treatment planning device 40 is installed outside the examination room and the control room. The treatment planning device 40 may be installed in the control room, or may be a device integrated with the console 10. Typical examples of the imaging apparatus 20 include an X-ray CT apparatus, an MRI (magnetic resonance imaging) apparatus, and an X-ray apparatus. Hereinafter, the case where the X-ray CT apparatus 20a is used as the imaging apparatus 20 will be described.

  As shown in FIG. 9, the console 10 of the treatment system 1 is configured based on a computer, and can communicate with a network such as a hospital basic LAN (local area network) not shown. The console 10 is mainly composed of basic hardware such as a CPU 11, a main memory 12, an image memory 13, an HDD (hard disc drive) 14, an input device 15, a display device 16, and a tracking device 100. The display device 16 also serves as the warning device described above. The CPU 11 is interconnected to each hardware component constituting the console 10 via a bus as a common signal transmission path. Note that the console 10 may include a recording medium drive.

  The CPU 11 is a control device having a configuration of an integrated circuit (LSI) in which an electronic circuit made of a semiconductor is enclosed in a package having a plurality of terminals. When a command is input by operating the input device 15 by an operator such as a doctor, the CPU 11 executes a program stored in the main memory 12. Alternatively, the CPU 11 is a program stored in the HDD 14, a program transferred from the network and installed in the HDD 14, or a program read from a recording medium mounted on a recording medium drive (not shown) and installed in the HDD 14. Are loaded into the main memory 12 and executed.

  The main memory 12 is a storage device having a configuration that combines elements such as a read only memory (ROM) and a random access memory (RAM). The main memory 12 stores IPL (initial program loading), BIOS (basic input / output system) and data, and is used for temporary storage of the work memory and data of the CPU 11.

  The image memory 13 is a storage device that stores fluoroscopic images, slice data as two-dimensional image data, treatment plan volume data as three-dimensional image data, and volume data just before treatment.

  The HDD 14 is a storage device having a configuration in which a metal disk coated or vapor-deposited with a magnetic material is incorporated in a non-detachable manner. The HDD 14 is a storage device that stores programs installed in the console 10 (including application programs as well as an OS (operating system)) and data. Also, it is possible to cause the OS to provide a GUI (graphical user interface) that allows a basic operation to be performed by the input device 15 by using a lot of graphics for displaying information on the display device 16 for an operator such as an operator. it can.

  The input device 15 is a pointing device that can be operated by an operator, and an input signal according to the operation is sent to the CPU 11.

  The display device 16 includes an image composition circuit (not shown), a video random access memory (VRAM), a display, and the like. The image synthesizing circuit generates synthesized data obtained by synthesizing character data of various parameters with image data. The VRAM develops the composite data as display image data to be displayed on the display. The display is composed of a liquid crystal display, a CRT (cathode ray tube) or the like, and sequentially displays display image data as a display image.

  The console 10 controls the operations of the X-ray CT apparatus 20a, the bed apparatus 30, and the radiation therapy apparatus 50. Further, the console 10 generates projection data by performing logarithmic conversion processing and correction processing (preprocessing) such as sensitivity correction on the raw data input from the DAS 24 of the X-ray CT apparatus 20a, and generates projection data based on the projection data. In addition, slice data as two-dimensional image data and volume data as three-dimensional image data are generated.

  The X-ray CT apparatus 20a of the treatment system 1 images a region including the treatment site in order to display image data of the region including the treatment site such as an affected part of the subject 111. The X-ray CT apparatus 20a includes an X-ray tube 21 as a radiation source, an aperture 22, an X-ray detector 23, a DAS (data acquisition system) 24, a rotating unit 25, a high voltage supply device 26, an aperture driving device 27, and a rotational drive. A device 28 and an imaging controller 29 are provided.

  The X-ray tube 21 generates a braking X-ray by causing an electron beam to collide with a metal target according to the tube voltage supplied from the high voltage supply device 26, and the X-ray is directed toward the X-ray detector 23. Irradiate. Fan beam X-rays and cone beam X-rays are formed by X-rays emitted from the X-ray tube 21.

  The diaphragm 22 adjusts the irradiation range of the X-rays emitted from the X-ray tube 21 by the diaphragm driving device 27. That is, the X-ray irradiation range can be changed by adjusting the aperture of the diaphragm 22 by the diaphragm driving device 27.

  The X-ray detector 23 is a matrix, that is, a two-dimensional array type X-ray detector (also referred to as a multi-slice detector) having a plurality of channels in the channel direction and a plurality of rows in the slice direction. is there. The X-ray detection element of the X-ray detector 23 detects X-rays emitted from the X-ray tube 21.

  The DAS 24 amplifies the transmission data signal detected by each X-ray detection element of the X-ray detector 23 and converts it into a digital signal. Output data from the DAS 24 is supplied to the console 10 via the imaging controller 29.

  The rotating unit 25 integrally holds the X-ray tube 21, the diaphragm 22, the X-ray detector 23, and the DAS 24. The rotating unit 25 can rotate around the subject 111 with the X-ray tube 21, the diaphragm 22, the X-ray detector 23, and the DAS 24 as one body in a state where the X-ray tube 21 and the X-ray detector 23 face each other. It is configured as follows. A direction parallel to the rotation center axis of the rotating unit 25 is defined as a z-axis direction, and a plane orthogonal to the z-axis direction is defined as an x-axis direction and a y-axis direction.

  The high voltage supply device 26 supplies power necessary for X-ray irradiation to the X-ray tube 21 under the control of the imaging controller 29.

  The diaphragm driving device 27 has a mechanism for adjusting the irradiation range of the diaphragm 22 in the X-ray slice direction under the control of the imaging controller 29.

  The rotation driving device 28 has a mechanism for rotating the rotating unit 25 so that the rotating unit 25 rotates around the hollow portion while maintaining the positional relationship under the control of the imaging controller 29.

  The imaging controller 29 includes a CPU and a memory. The imaging controller 29 controls the X-ray tube 21, the X-ray detector 23, the DAS 24, the high voltage supply device 26, the aperture driving device 27, the rotation driving device 28, etc. Run a scan.

  The couch device 30 of the treatment system 1 includes a couchtop 31, a couchtop drive device 32, and a couch controller 39.

  The top plate 31 can place the subject 111 thereon. The top plate driving device 32 is controlled by the bed controller 39 to move the top plate 31 up and down along the y-axis direction, the mechanism to move the top plate 31 back and forth along the z-axis direction, and the top plate 31. and a mechanism that rotates the y-axis direction as an axis.

  The bed controller 39 includes a CPU and a memory. The couch controller 39 controls the top board driving device 32 and the like, thereby executing a scan with the operation of the X-ray CT apparatus 20a. In addition, the bed controller 39 controls the top board driving device 32 and the like, thereby executing radiation therapy with the operation of the radiation therapy device 50.

  The treatment planning device 40 of the treatment system 1 is treatment planning data for performing radiation treatment by the radiation treatment device 50 based on slice data and volume data that are imaged using the X-ray CT device 20a and generated by the console 10. Is generated. Under the control of the console 10 based on the treatment plan data generated by the treatment planning device 40, the radiation treatment device 50 irradiates the treatment site of the subject 111 with the treatment beam. The treatment planning apparatus 40 is configured on the basis of a computer, and can communicate with a network such as a hospital backbone LAN (not shown). The treatment planning device 40 is mainly composed of basic hardware such as a CPU 41, a main memory 42, a treatment plan memory 43, an HDD 44, an input device 45, and a display device 46. The CPU 41 is interconnected to each hardware component constituting the treatment planning apparatus 40 via a bus as a common signal transmission path. The treatment planning device 40 may include a recording medium drive.

  The configuration of the CPU 41 is equivalent to the configuration of the CPU 11 of the console 10. When an instruction is input by operating the input device 45 by an operator, the CPU 41 executes a program stored in the main memory 42. Alternatively, the CPU 41 may be a program stored in the HDD 44, a program transferred from the network and installed in the HDD 44, or a program read from a recording medium installed in a recording medium drive (not shown) and installed in the HDD 44. Are loaded into the main memory 42 and executed.

  The configuration of the main memory 42 is the same as the configuration of the main memory 12 of the console 10. The main memory 42 stores IPL, BIOS, and data, and is used for temporary storage of the work memory and data of the CPU 41.

  The treatment plan memory 43 is a storage device that stores treatment plan data. The configuration of the HDD 44 is the same as the configuration of the HDD 14 of the console 10. The input device 45 is equivalent to the configuration of the input device 15 of the console 10. The display device 46 is equivalent to the configuration of the display device 16 of the console 10.

  The treatment planning apparatus 40 obtains the position of the treatment site and the shape of the treatment site of the subject 111 based on the image data generated by the X-ray CT apparatus 20a, and treats the treatment beam (X-ray, electron) to be irradiated on the treatment site. Line, neutron beam, proton beam, heavy particle beam, etc.), its energy, and irradiation field.

  The radiotherapy apparatus 50 of the treatment system 1 can generally generate MV class radiation. The radiation therapy apparatus 50 installs a diaphragm (collimator) at a radiation generation port, and realizes an irradiation shape and a dose distribution based on the treatment plan by the diaphragm. In recent years, a multi-leaf collimator (MLC) that can form a dose distribution corresponding to a complicated tumor shape by a plurality of movable leaves as a diaphragm is often used. The radiation therapy apparatus 50 adjusts the radiation dose by the irradiation field formed by the diaphragm, and eliminates or reduces the treatment site of the subject 111.

  The radiation therapy apparatus 50 includes a radiation source 51 as a radiation source, an aperture 52, an arm unit 55, a high voltage supply device 56, an aperture drive device 57, a rotation drive device 58, and a treatment controller 59.

  The radiation source 51 generates radiation according to the tube voltage supplied from the high voltage supply device 56.

  The diaphragm 52 adjusts the irradiation range of the radiation irradiated from the radiation source 51 by the diaphragm driving device 57. That is, by adjusting the aperture of the diaphragm 52 by the diaphragm driving device 57, the radiation irradiation range can be changed.

  The arm unit 55 holds the radiation source 51 and the diaphragm 52 together. The arm unit 55 is configured so that the radiation source 51 and the diaphragm 52 can be rotated around the subject 111 as a unit.

  The high voltage supply device 56 supplies the radiation source 51 with electric power necessary for irradiation with radiation under the control of the treatment controller 59.

  The diaphragm driving device 57 has a mechanism for adjusting the radiation range of the diaphragm 52 under the control of the treatment controller 59.

  The rotation driving device 58 has a mechanism for rotating the arm portion 55 so as to rotate around the connection portion between the arm portion 55 and the support portion under the control of the treatment controller 59.

  The treatment controller 59 includes a CPU and a memory. The treatment controller 59 controls the radiation source 51, the high voltage supply device 56, the diaphragm drive device 57, and the like according to the treatment plan data generated by the treatment planning device 40, thereby performing treatment with the operation of the bed device 30. Radiation is executed.

  As described above, according to the present embodiment and each modification, it is possible to reduce the exposure dose due to fluoroscopic radiation.

  Note that the present invention is not limited to the above-described embodiments as they are, and can be embodied by modifying the constituent elements without departing from the scope of the invention in the implementation stage. Various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the above embodiments. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, the constituent elements over different embodiments may be appropriately combined.

  For example, as long as each step in the flowchart of the above embodiment is not contrary to its nature, the execution order may be changed, a plurality of steps may be performed simultaneously, or may be performed in a different order for each execution.

DESCRIPTION OF SYMBOLS 1 Treatment system 100 Tracking apparatus 101 Memory 102 Calculation part 103 Detection part 104 Management part 110 Bed 112,113 1st irradiation part 116,117 Radiation detection part 123 Control part 120 2nd irradiation part

Claims (11)

A calculation unit that calculates the likelihood for each pixel of interest included in the fluoroscopic image seen through the inside of the subject;
A detection unit that detects a position of an object inside the subject based on the likelihood of each of the target pixels;
With
The pixel values of the two or more first pixels are included in the first region with respect to the first difference indicating the difference between the pixel values of the two or more first pixels included in the first region around the pixel of interest. The likelihood when the second difference indicating a difference from the pixel value of one or more second pixels that is not greater is larger than the likelihood when the second difference with respect to the first difference is small,
Image processing device. The first region is a shape that depends on the shape of the object.
The image processing apparatus according to claim 1. The calculation unit acquires area information indicating a third area where the object image is expected to exist in the perspective image, and calculates the likelihood for each target pixel in the third area indicated by the area information. calculate,
The image processing apparatus according to claim 1. The calculation unit predicts a fourth region where the object image is expected to exist in the fluoroscopic image based on a predetermined motion model, and calculates the likelihood for each target pixel in the fourth region. ,
The image processing apparatus according to claim 1. A storage unit in which the pixels constituting the fluoroscopic image are written in a predetermined order;
A management unit that manages the timing at which all the pixels necessary for the calculation of the likelihood in the target pixel for which the calculation of the likelihood is first performed by the calculation unit are written in the storage unit;
The calculation unit starts calculating the likelihood for each pixel of interest after the timing has elapsed.
The image processing apparatus of any one of Claims 1-4. A plurality of the objects exist inside the subject,
The calculation unit calculates the likelihood for each pixel of interest of each of a plurality of fluoroscopic images obtained by seeing through the inside of the subject in different directions,
The detection unit acquires region information indicating a third region in which each of the plurality of object images is expected to exist for each perspective image and a projection matrix that projects a three-dimensional coordinate onto the perspective image, and the perspective image A plurality of target pixels having a likelihood that a value satisfies a predetermined condition is detected for each of the plurality of target pixels detected from the fluoroscopic image using the region information of the fluoroscopic image for each of the fluoroscopic images. The plurality of pixels of interest detected in each of the plurality of fluoroscopic images, the projection matrix of each of the plurality of fluoroscopic images, Detecting a three-dimensional position of the plurality of objects using the plurality of pixels of interest detected from each of the plurality of perspective images and associated between the plurality of perspective images;
The image processing apparatus according to claim 1. A plurality of the objects exist inside the subject,
The calculation unit calculates the likelihood for each pixel of interest of each of a plurality of fluoroscopic images obtained by seeing through the inside of the subject in different directions,
The detection unit obtains a projection matrix for projecting three-dimensional coordinates to the fluoroscopic image for each fluoroscopic image, and detects a plurality of target pixels having a likelihood that the magnitude of a value satisfies a predetermined condition for each fluoroscopic image, The plurality of pixels of interest that are detected in each of the plurality of fluoroscopic images are associated with each other by the epipolar geometry using the projection matrix of each of the plurality of fluoroscopic images, and each of the plurality of fluoroscopic images is associated with each other. A projection matrix, detected from each of the plurality of perspective images, and using the plurality of pixels of interest associated with the plurality of perspective images to detect a three-dimensional position of the plurality of objects;
The image processing apparatus according to claim 1. The calculation unit obtains object information relating to the shape and size of the object and a projection matrix that projects three-dimensional coordinates onto the perspective image, and uses the object information and the projection matrix to reflect the object in the perspective image. Calculating the shape and size of the image of the object, and setting the first region with the shape and size of the image;
The image processing apparatus according to claim 1. A first irradiation unit for irradiating a subject with radiation;
A first detection unit that detects the radiation that has passed through the subject and generates a fluoroscopic image of the inside of the subject;
A calculation unit that calculates a likelihood for each target pixel included in the fluoroscopic image;
A second detection unit for detecting a position of an object inside the subject based on the likelihood of each of the target pixels;
A second irradiation unit configured to irradiate the object with a treatment beam based on the detected position;
With
The pixel values of the two or more first pixels are included in the first region with respect to the first difference indicating the difference between the pixel values of the two or more first pixels included in the first region around the pixel of interest. The likelihood when the second difference indicating a difference from the pixel value of one or more second pixels that is not greater is larger than the likelihood when the second difference with respect to the first difference is small,
Treatment system. A plurality of the objects exist inside the subject,
The calculation unit predicts a fourth area where each of the plurality of object images is expected to be present in the fluoroscopic image based on a predetermined motion model, and for each of the target pixels in each of the plurality of fourth areas. Calculating the likelihood,
The second detection unit sets, for each image, a prior probability that the value increases as the image moves closer to the position based on the predetermined motion model in the fourth region where the image is expected to exist. For each of the fourth regions, the position of the object is detected based on the likelihood and the prior probability of each of the target pixels, and when the distance between the plurality of images is equal to or less than a threshold, Calculating position information of at least one of the first irradiating unit and the first detecting unit in which the fluoroscopic image is generated such that the distance between images exceeds the threshold;
The treatment system according to claim 9, further comprising a control unit that controls a position of at least one of the first irradiation unit and the first detection unit based on the position information. A calculation step for calculating the likelihood for each target pixel included in the fluoroscopic image seen through the inside of the subject;
A detection step of detecting a position of an object inside the subject based on the likelihood of each of the target pixels;
Including
The pixel values of the two or more first pixels are included in the first region with respect to the first difference indicating the difference between the pixel values of the two or more first pixels included in the first region around the pixel of interest. The likelihood when the second difference indicating a difference from the pixel value of one or more second pixels that is not greater is larger than the likelihood when the second difference with respect to the first difference is small,
Image processing method.

Images