JP2016221100A - Medical image processing apparatus, and treatment system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a medical image processing apparatus and treatment system that are capable of improving reliability.SOLUTION: A medical image processing apparatus of an embodiment includes an image acquisition unit, image feature derivation unit, and position estimation unit. The image acquisition unit acquires a subject's perspective image picked up by an image pickup apparatus. The image feature derivation unit derives an image feature on a target area, which is part or all of the perspective image acquired by the image acquisition unit. The position estimation unit, on the basis of correspondence information showing correspondence between the image feature and a region to be searched, estimates a position of the subject's region to be searched in the perspective image from the image feature derived by the image feature derivation unit.SELECTED DRAWING: Figure 1

Description

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

  In radiation therapy, first, a CT (Computed Tomography) image near the affected area is taken in advance at the planning stage, and a treatment plan is made. In the treatment stage, the patient is treated by irradiating the affected part with a treatment beam in accordance with the treatment plan. The affected area of a patient may move due to breathing, heartbeat, intestinal movement, and the like. As a treatment method corresponding to this, a gated irradiation method and a tracking irradiation method are known. By these irradiation methods, it is possible to reduce the amount of the treatment beam irradiated to the normal site other than the affected part.

  In addition, CT images of various respiratory phases may be taken immediately before treatment, separately from treatment planning. In this case, a CT image similar to the treatment planning CT image is selected from CT images of various respiratory phases. The treatment beam is automatically irradiated when the fluoroscopic image used for generating the selected CT image and the fluoroscopic image taken during the treatment substantially match the image information around the diaphragm.

  However, the conventional automation technology does not provide treatment support by tracking the position of the affected part itself in the treatment stage or the rehearsal stage, and thus may not have sufficient reliability.

JP 2008-154861 A

Ying Cui, Jennifer G Dy, Gregory C Sharp, Brain Alexander and Steve B Jiang, “Multiple template-based fluoroscopic tracking of lung tumor mass without implant fiducial markers,” Physics in Medicine and Biology, vol.52, no.20, pp .6229-6242, 2007.

  The problem to be solved by the present invention is to provide a medical image processing apparatus and a treatment system capable of improving reliability.

  The medical image processing apparatus according to the embodiment includes an image acquisition unit, an image feature derivation unit, and a position estimation unit. The image acquisition unit acquires a fluoroscopic image of the subject imaged by the imaging device. The image feature deriving unit derives image features for a target region that is a part or all of the fluoroscopic image acquired by the image acquiring unit. The position estimating unit estimates the position of the search part of the subject in the fluoroscopic image from the image feature derived by the image feature deriving part based on correspondence information indicating a correspondence relationship between the image feature and the position of the search part. To do.

1 is a diagram illustrating an example of a functional configuration of a medical image processing apparatus 100 according to a first embodiment. The figure which shows typically the content of the function which derives an affected part probability from the image feature of an object area | region. The figure which shows typically the other example of the content of the function which derives an affected part probability from the image feature of an object area | region. The figure which shows typically the other example of the content of the function which derives an affected part probability from the image feature of an object area | region. The figure which shows an example of the content of the process by the image feature derivation | leading-out part 112 and the position estimation part 114. FIG. 6 is a flowchart showing an example of the content of processing executed by the medical image processing apparatus 100 according to the first embodiment. 6 is a flowchart showing another example of the content of processing executed by the medical image processing apparatus 100 according to the first embodiment. The figure which shows an example of the content of the process by the image feature derivation | leading-out part 112 and the position estimation part 114. FIG. The figure which shows the function structural example of 100 A of medical image processing apparatuses which concern on 2nd Embodiment. 10 is a flowchart illustrating an example of a flow of processing executed by the medical image processing apparatus 100A according to the second embodiment. The figure which shows the display screen when an affected part position and an object area | region are designated by input operation by a user. The figure which shows a mode that the corresponding | compatible information 150 is constructed | assembled. The figure which shows the structural example of the treatment system 1 containing the treatment apparatus 10 and the medical image processing apparatus 100A. The flowchart which shows an example of the flow of the treatment performed using the treatment system. The flowchart which shows an example of the flow of a treatment plan. The flowchart which shows an example of the flow of a learning process. The figure which shows typically a mode that the affected part position input with respect to one CT image is expand | deployed to another CT image. The flowchart which shows the flow of the process performed in rehearsal. The figure which shows an example of the display screen of a rehearsal stage. The flowchart which shows an example of the flow of the process performed in a treatment stage. The figure which shows an example of the display screen of a treatment stage.

  Hereinafter, a medical image processing apparatus and a treatment system of an embodiment will be described with reference to the drawings.

<< Medical image processing device >>
<First Embodiment>
Hereinafter, the medical image processing apparatus 100 according to the first embodiment will be described. FIG. 1 is a diagram illustrating a functional configuration example of a medical image processing apparatus 100 according to the first embodiment. The medical image processing apparatus 100 includes, for example, an image acquisition unit 110, an image feature derivation unit 112, a position estimation unit 114, a correspondence information acquisition unit 116, and an output unit 118.

  Among these functional units, the image feature deriving unit 112, the position estimating unit 114, and the other functional units are, for example, programs stored in a storage device by a processor such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). It is a software function unit that functions when executed. That is, the medical image processing apparatus 100 can be realized by using, for example, a general-purpose computer apparatus as basic hardware. The image feature deriving unit 112, the position estimating unit 114, and other functional units can be realized by causing a processor mounted on the computer device to execute a program. At this time, the medical image processing apparatus 100 may be realized by installing the above program in a computer device in advance, or may be stored in a storage medium such as a CD-ROM or the above program via a network. You may implement | achieve by distributing and installing this program in a computer apparatus suitably. Some or all of these functional units may be hardware functional units such as an FPGA (Field Programmable Gate Array), an LSI (Large Scale Integration), and an ASIC (Application Specific Integrated Circuit).

  The image acquisition unit 110 acquires a fluoroscopic image as an input image from a treatment device (described later) connected by a network such as a LAN (Local Area Network), a WAN (Wide Area Network), or a serial communication line. The image acquisition unit 110 includes a communication interface for connecting to the network. The fluoroscopic image is, for example, a moving image obtained by imaging a subject with X-rays. A subject is an organism such as a person or an animal, and may be referred to as a person who receives treatment, that is, a “patient”.

  The image feature deriving unit 112 derives, for the target region in the fluoroscopic image acquired by the image acquiring unit 110, an image feature indicating the feature of the image in the target region. The target area may be a part of the fluoroscopic image or may be the entire fluoroscopic image. In addition, the position of the target area in the fluoroscopic image is specified in advance, and the image feature deriving unit 112 may derive the image feature for the target area, or the image feature deriving unit 112 may change the target area. Image features may be derived. In the former case, for example, the correspondence information 150 includes information for specifying the position of the target region.

  The image feature is, for example, information obtained by obtaining a luminance gradient for each pixel in the target region and arranging the obtained luminance gradient. The luminance gradient can be obtained by a technique such as a SOBEL filter. In addition, the image feature may be an array of pixel values of each pixel in the target area, or information obtained by arraying a histogram obtained with respect to the luminance gradient, pixel value, etc. of each pixel in the target area. There may be.

  As will be described later, the correspondence information is generated by learning processing using, for example, DRR (Digitally Reconstructed Radiograph) as a learning image. For this reason, it is preferable to obtain an image feature that does not change much due to a difference between a fluoroscopic image by X-rays and DRR. The direction of the above-described luminance gradient has a tendency not to change so much between the X-ray fluoroscopic image and the DRR, and is therefore preferably used in the medical image processing apparatus 100.

Hereinafter, the arrayed image features are treated as vector information, and the image features are represented as → x = (x ₁ , x ₂ ,..., X _n ). Note that “→” in front of a character indicates that the following character is a vector.

  The position estimation unit 114 estimates the affected part position of the subject from the image feature derived by the image feature deriving unit 112. The affected part position of the subject is an example of the “search position” of the subject. For example, the position estimation unit 114 estimates the affected part position of the subject based on the correspondence information 150 obtained by the correspondence information obtaining unit 116 and indicating the correspondence between the image feature and the affected part position. The generation process (learning process) of the correspondence information 150 will be described later.

  The correspondence information acquisition unit 116 acquires the correspondence information 150. The correspondence information 150 is read from a storage device included in the medical image processing apparatus 100, for example. In this case, the storage device is a ROM (Read Only Memory), a RAM (Random Access Memory), a HDD (Hard Disk Drive), a flash memory, or the like, and the correspondence information acquisition unit 116 is regarded as a function of the position estimation unit 114. It's okay. Further, the correspondence information 150 may be acquired via a network from an external device such as a treatment device, a data server, or a network storage described later. In this case, the correspondence information acquisition unit 116 includes a communication interface for connecting to a network. The correspondence information 150 may be obtained from a storage medium such as a CD (Compact Disc) or a DVD (Digital Versatile Disc). In this case, the correspondence information acquisition unit 116 includes a drive device to which a storage medium is mounted.

  The output unit 118 includes a display device such as an LCD (Liquid Crystal Display) or an organic EL (Electroluminescence). In this case, the output unit 118 displays the affected part position of the subject estimated by the position estimation unit 114 so as to be superimposed on the fluoroscopic image. The output unit 118 may include a communication interface for outputting the estimation result obtained by the position estimation unit 114 to an external device (for example, a treatment device). In this case, the output unit 118 has a configuration that is common in part or in whole with the image acquisition unit 110.

  Hereinafter, the estimation of the affected part position based on the image feature will be described in more detail. The correspondence information 150 includes, for example, a function f represented by the formula (1). In the equation, (u, v) is a position on the input image, and l (u, v) is an output value indicating the probability that the position (u, v) is an affected part. Hereinafter, the output value l (u, v) is referred to as an affected area probability l (u, v). The function f is, for example, a correlation coefficient (replaced with zero in the case of a negative value) or fitting between an array that is teacher data obtained in advance and an arrayed image feature derived by the image feature deriving unit 112 It is a function that outputs an index such as a rate, or a value obtained by discretizing these indices with a threshold value. The affected area probability l (u, v) may be expressed by, for example, a continuous value from 0 to 1, a binary value of 0 (not affected area) or 1 (affected area), or 0 The probability may be expressed by a discrete value or sign of three or more values such as (not affected), 0.5 (possibly affected), or 1 (is affected). The function is exemplified as an example of the derivation rule included in the correspondence information 150, and the function may be replaced with a derivation rule such as a map, table data, or algorithm having an equivalent input / output relationship. Hereinafter, this will be described on the assumption.

  f (→ x) = l (u, v) (1)

  FIG. 2 is a diagram schematically showing the content of the function f for deriving the affected area probability l (u, v) from the image feature → x of the target area OA. The left diagram in FIG. 2 shows a target area OA set in the input image IM. The right diagram of FIG. 2 shows how the affected area probability l (u, v) is obtained by applying the function f to the image feature → x derived from the target area OA. As shown in the figure, the function f, which is the correspondence information 150, indicates that an affected area probability l (at a specific position (u, v) on the input image IM when the image feature → x of the target area OA in the input image IM is input. u, v) is derived. The specific position (u, v) is, for example, the center (for example, the center of gravity) of the target area OA. Further, the function f may output the affected area probability of an area having a width instead of a single point.

  When the target area OA is not the entire input image IM, the function f derives an affected area probability l (u, v) depending on the position of the target area OA. In this case, the function f may be expressed by Expression (2). In the formula, p is position information in the input image IM of the target area OA.

  f (→ x, p) = l (u, v) (2)

The function f may derive one affected part probability l (u, v), or may output the affected part probability l (u, v) corresponding to a plurality of positions. FIG. 3 is a diagram schematically illustrating another example of the content of the function f for deriving the affected area probability l (u, v) from the image feature of the target area OA → x. The left diagram in FIG. 3 shows a target area OA set in the input image IM. The right figure of FIG. 3 shows the affected area probability corresponding to a plurality of positions (u ₁ , v ₁ ) to (u _k , v _k ) by applying the function f to the image feature derived from the target area OA → x. It shows how l (u, v) is obtained. As shown in the figure, the function f, which is the correspondence information 150, has a plurality of (k) positions (u ₁ , v) on the input image IM when the image feature → x of the target area OA in the input image IM is input. ₁ ) to (u _k , v _k ) corresponding to affected area probabilities l (u ₁ , v ₁ ) to l (u _k , v _k ) are derived.

  The function f may be a function for deriving an affected area probability l (u, v) at one or a plurality of positions of the input image IM based on the image feature → x for the entire input image IM. FIG. 4 is a diagram schematically showing another example of the content of the function f for deriving the affected area probability l (u, v) from the image feature of the target area OA → x. The left diagram of FIG. 4 shows a state where the input image IM matches the target area OA. The right diagram of FIG. 4 shows how the affected area probability l (u, v) is obtained by applying the function f to the target area OA, that is, the image feature derived from the input image IM → x. As shown in the figure, the function f, which is the correspondence information 150, indicates that the affected area probability at a specific position (u, v) on the input image IM when the image feature of the input image IM (= target area OA) → x is input. l (u, v) is derived.

  The position estimation unit 114 estimates the affected part position of the subject using the various correspondence information 150 exemplified above.

[Processing when the correspondence information includes one function]
When the correspondence information 150 includes one function f, the image feature derivation unit 112 and the position estimation unit 114, for example, change the position of the target area OA in the input image IM and change the affected part probability l at a plurality of positions in the input image IM. (U, v) is derived.

  FIG. 5 is a diagram illustrating an example of the contents of processing performed by the image feature deriving unit 112 and the position estimating unit 114. Here, it is assumed that the function f derives an affected area probability l (u, v) representing the probability with a continuous value from 0 to 1, for example. The function f is assumed to derive one affected area probability l (u, v) for one target area OA. The left diagram in FIG. 5 illustrates a state in which the target area OA is sequentially set in the input image IM. As shown in the left diagram of FIG. 5, the image feature deriving unit 112 derives the image feature → x for each target area OA while moving the target area OA within the scanning area SA in the input image IM. Then, the position estimation unit 114 derives an affected area probability l (u, v) corresponding to a plurality of positions based on the image feature → x for each target area OA. The right diagram in FIG. 5 is a diagram illustrating the distribution of the affected area probability l (u, v) derived by the position estimation unit 114. In such a distribution, it can be determined that the position where the affected area probability l (u, v) is high is the position where the probability that the affected area is actually the affected area of the subject is high.

  The position estimation unit 114 further estimates the affected part position based on the affected part probability l (u, v) obtained for each position. The position of the affected part of the subject is estimated by the process exemplified below.

(Estimation process A)
For example, the position estimation unit 114 estimates a position where the affected area probability l (u, v) is the maximum value as an affected area position (u _t , v _t ). In this case, the affected part position (u _t , v _t ) estimated by the position estimation unit 114 is expressed by, for example, Expression (3).

(Estimation process B)
The position estimation unit 114 may estimate the center of gravity of the distribution of the affected area probability l (u, v) as the affected area position (u _t , v _t ). In this case, the affected part position (u _t , v _t ) estimated by the position estimation unit 114 is expressed by, for example, Expression (4).

(Estimation process C)
The position estimating unit 114, a center of the local region sum of the affected probability l (u, v) is the largest in the local area (e.g. the center of gravity of a circle or ellipse center, etc.), the affected area position (u _t, v _t ). In this case, the affected part position (u _t , v _t ) estimated by the position estimation unit 114 is expressed by, for example, Expression (5).

  Note that the contents of the processes exemplified as (estimation process A) to (estimation process C) are merely examples, and the position estimation unit 114 may perform other processes having similar properties. The same shall apply hereinafter.

  FIG. 6 is a flowchart illustrating an example of the content of processing executed by the medical image processing apparatus 100 according to the first embodiment. This flowchart shows the flow of processing when the processing illustrated in FIG. 5 is performed.

  First, the correspondence information acquisition unit 116 acquires correspondence information about a subject whose target position is to be estimated from now on (step S200).

  Next, the image acquisition unit 110 acquires an image of one frame from an image input as a moving image (step S202), and the image feature deriving unit 112 initializes the position of the target area OA (step S204). As shown in FIG. 5, for example, the upper left area of the scanning area SA is set as the initial area of the target area OA.

  Next, the image feature deriving unit 112 derives the image feature → x for the target area OA (step S206), and the position estimating unit 114 derives the affected area probability l (u, v) based on the image feature → x. (Step S208). Then, the image feature deriving unit 112 determines whether or not the image feature deriving unit 112 and the position estimating unit 114 have completed the processing for all target areas OA that can be set in the scanning area SA (step S210). If the processing has not been completed for all target areas OA, the image feature deriving unit 112 changes the position of the target area OA (step S212), and returns the process to step S206. As shown in FIG. 5, the image feature deriving unit 112 sets the position of the target area OA in order by a predetermined width in the horizontal direction of the image, and when reaching the right end of the scanning area SA, the image feature deriving unit 112 has a predetermined width in the vertical direction of the image. Repeatedly setting the position of the target area OA at the left end position of the lowered scanning area SA. When the target area OA has reached the lowermost right position of the scanning area SA, it is determined in step S210 that “the process has been completed for all target areas OA”.

  When it is determined in step S210 that the processing has been completed for all target areas OA, the position estimation unit 114 estimates the affected part position from the distribution of the affected part probability l (u, v) derived in step S208 over a plurality of times. (Step S214). Specifically, the position estimation unit 114 estimates the position of the affected part of the subject by any one of the above-exemplified (estimation process A) to (estimation process C).

  Next, the output unit 118 outputs the affected part position of the subject estimated by the position estimation unit 114 (step S216). Then, when the supply of the moving image is stopped or an end operation is performed by the user, the process of this flowchart ends (step S218). The user is a person who uses the medical image processing apparatus 100 such as a doctor or a medical radiographer. While the supply of moving images continues, the processes in steps S202 to S216 are repeatedly executed until an end operation is performed.

  When the function f derives the affected area probability l (u, v) corresponding to a plurality of positions for one target area OA, the affected area probability l (u, v) corresponding to the plurality of positions is calculated. The affected part position of the subject may be estimated by performing any one of the above-exemplified (estimation process A) to (estimation process C). FIG. 7 is a flowchart illustrating another example of the content of processing executed by the medical image processing apparatus 100 according to the first embodiment. This flowchart shows a process flow when the function f derives the affected area probability l (u, v) corresponding to a plurality of positions for one target area OA.

  First, the correspondence information acquisition unit 116 acquires correspondence information about a subject whose target position is to be estimated (step S300).

  Next, the image acquisition unit 110 acquires an image of one frame from the image input as a moving image (step S302), and the image feature deriving unit 112 derives the image feature → x for the target area OA (step S304). ), The position estimation unit 114 derives an affected area probability l (u, v) corresponding to a plurality of positions based on the image feature → x (step S306). Then, the position estimation unit 114 estimates the affected part position from the distribution of the affected part probability l (u, v) derived in step S306 (step S308). Specifically, the position estimation unit 114 estimates the position of the affected part of the subject by any one of the above-exemplified (estimation process A) to (estimation process C).

  Next, the output unit 118 outputs the affected part position of the subject estimated by the position estimation unit 114 (step S310). Then, when the supply of moving images is stopped or an end operation is performed by the user, the processing of this flowchart ends (step S312). While the supply of moving images continues, the processes in steps S302 to S310 are repeatedly executed until an end operation is performed.

[Processing when the correspondence information includes multiple functions]
The correspondence information 150 acquired by the correspondence information acquisition unit 116 may include a plurality of functions f instead of including only one function f. When the correspondence information 150 includes a plurality of functions f, the image feature deriving unit 112 and the position estimating unit 114 perform the following processing, for example.

FIG. 8 is a diagram illustrating an example of the contents of processing performed by the image feature deriving unit 112 and the position estimating unit 114. Here, the two functions f ₁ and f ₂ respectively derive the affected area probabilities l ₁ (u, v) and l ₂ (u, v) representing the probabilities with continuous values from 0 to 1, for example. To do. Further, the functions f ₁ and f ₂ respectively derive a plurality of affected area probabilities l ₁ (u, v) and l ₂ (u, v) for one target area OA (see FIG. 3). ). The left diagram in FIG. 8 illustrates a state in which a plurality of target areas OA ₁ and OA ₂ are set in the input image IM. The right diagram in FIG. 8 shows the image feature derived from the target area OA ₁ → the affected area probability obtained by applying the function f ₁ to x ₁ and the image characteristic derived from the target area OA ₂ → the function to x ₂ . The distribution of the affected area probability l (u, v) obtained by superimposing the affected area probability obtained by applying f ₂ is shown.

The image feature deriving unit 112 derives the image feature → x ₁ for the target area OA ₁ for the function f ₁ that is set in advance in the input image IM, and also sets the image feature for the target area OA ₂ for the function f ₂ → x ₂ is derived. The position estimation unit 114 derives the affected area probability l ₁ (u, v) based on the image feature for the target area OA ₁ → x ₁ and also determines the affected area probability based on the image feature for the target area OA ₂ → x _2. l ₂ (u, v) is derived. Then, the position estimation unit 114 derives the affected part probability l (u, v) by superposing the affected part probability l ₁ (u, v) and the affected part probability l ₂ (u, v). Hereinafter, the overlay process will be exemplified.

(Superposition processing D)
For example, the position estimation unit 114 performs superimposition by obtaining the sum of the affected area probabilities l ₁ (u, v) to the affected area probabilities l _n (u, v) (n is an arbitrary natural number and is an identifier of a function). The affected part probability l (u, v) is derived. In this case, the overlapped disease part probability l (u, v) is expressed by, for example, Expression (6). Where δ is the Kronecker delta.

(Overlay process E)
The position estimating unit 114, the affected area probability _l 1 (u, v) from by obtaining the product of the affected area probability _l n (u, v), superimposed diseased probability l (u, v) be derived the Good. In this case, the overlapped disease part probability l (u, v) is expressed by, for example, Expression (7).

(Overlay process F)
Further, the position estimation unit 114 may derive an affected area probability l (u, v) obtained by superimposing the affected area probability l _n (u, v) from the affected area probability l ₁ (u, v) by kernel density estimation. . In this case, the overlapped disease part probability l (u, v) is expressed by, for example, Expression (8). In the equation, K is a kernel function, h is a bandwidth, and Z is a product of the number of samples and the bandwidth.

Furthermore, the position estimation unit 114 exemplifies the above for the affected part probability l (u, v) superimposed by any one of the above-exemplified (superimposition processing D) to (superimposition processing F) ( Any one of the estimation processes A) to (estimation process C) is performed to estimate the affected area position (u _t , v _t ).

  Note that the contents of the processes exemplified as (superimposition process D) to (superimposition process F) are merely examples, and the position estimation unit 114 may perform other processes having similar properties. The same shall apply hereinafter.

  Further, even when the correspondence information 150 acquired by the correspondence information acquisition unit 116 includes a plurality of functions f, the image feature deriving unit 112 and the position estimation unit 114 are within the scanning area SA as shown in FIG. Processing may be performed while changing the area OA. Also in this case, the position estimation unit 114 performs a process of superimposing the affected area probabilities derived by the plurality of functions f in the same manner as described above.

  According to the medical image processing apparatus 100 of the present embodiment described above, the image information is derived for the target area OA that is a part or all of the fluoroscopic image, and the correspondence information indicating the correspondence between the image feature and the position of the search part. Based on 150, the position of the affected part of the subject in the fluoroscopic image is estimated from the derived image feature, so that the position of the affected part can be estimated more accurately. As a result, it is possible to improve the reliability of treatment with radiation using an image.

<Second Embodiment>
Hereinafter, the medical image processing apparatus 100A according to the second embodiment will be described. FIG. 9 is a diagram illustrating a functional configuration example of the medical image processing apparatus 100A according to the second embodiment. The medical image processing apparatus 100A according to the second embodiment is different from the medical image processing apparatus 100 according to the first embodiment in that it further includes a correspondence information learning unit 120 and an input unit 122. Hereinafter, the same points as those in the first embodiment will be denoted by the same reference numerals, description thereof will be omitted, and differences will be mainly described.

  The correspondence information learning unit 120 learns the correspondence information 150 based on the learning image input to the medical image processing apparatus 100A. The learning image is, for example, a DRR generated from a 3DCT image for treatment planning. The angle of view of the DRR is calculated in advance so as to coincide with the fluoroscopic image that is the input image IM. Details will be described in the treatment system.

  The input unit 122 includes input devices such as a keyboard, a mouse, a touch panel, a touch pad, and a radio button.

  FIG. 10 is a flowchart illustrating an example of a flow of processing executed by the medical image processing apparatus 100A according to the second embodiment. First, the correspondence information learning unit 120 specifies the affected part position in the learning image (step S400). The affected part position is specified by receiving an input operation on the input unit 122 by the user while displaying a learning image on the display device, for example. The affected part position in the DRR may be specified in advance, or the correspondence information learning unit 120 may automatically specify the affected part position by analyzing the learning image.

  Next, the correspondence information learning unit 120 determines a target region from which an image feature is derived in the learning image (step S402). The target area is specified by accepting an input operation on the input unit 122 by the user while displaying a learning image on the display device, for example, similarly to the affected part position. The target area may be set so as to include the affected part, or may be set so as to include a position (for example, a bone or muscle) where the image feature appears notably although the affected part is not included. Further, the correspondence information learning unit 120 may automatically select a position having such characteristics.

  FIG. 11 is a diagram illustrating a display screen when an affected area position and a target area are designated by an input operation by a user. The left diagram in FIG. 11 is an affected area position designation screen, and the right diagram in FIG. 11 is a target area designation screen. First, the affected part position TP is designated by a mouse operation or a touch operation on the affected part position designation screen. Then, when the confirm button B1 is operated, a transition is made to the target area designation screen shown in the right diagram of FIG. On the target area designation screen, when two corners C1 and C2 on the opposite side are designated by, for example, a mouse operation or a slide operation on the touch panel, the target area OA is superimposed and displayed based on the designated corner. When the confirmation button B1 is operated, the specification of the affected part position TP and the target area OA is completed. When a rule such as “the affected part position is the center of the target area” is applied, if one corner is designated in the right diagram of FIG. 11, the other corners are calculated from the designated corner and the affected part position. Therefore, the designation of the target area OA is completed. Furthermore, when the size of the target area OA is fixed, the target area OA is automatically determined when the affected part position TP is designated, and therefore the right diagram of FIG. 11 is not displayed. When the button B2 is operated, the entire learning image is designated as the target area.

  Next, the correspondence information learning unit 120 derives an image feature in the target region (step S404). Such processing may be executed as processing of the correspondence information learning unit 120, or may be executed by calling the image feature deriving unit 112 as a subroutine or the like.

  Next, the correspondence information learning unit 120 constructs the correspondence information 150 based on the affected part position, the position of the target region, and the image feature (step S406). FIG. 12 is a diagram schematically showing how the correspondence information 150 is constructed. The function f as the correspondence information 150 is based on the image feature of the target area, the position information of the target area (for example, the upper left coordinates and the lower right coordinates), and the affected area position, and the image feature of the target area, the target area and the affected area position. The relative positional relationship and the rule for the image feature are constructed based on the constituent elements. As described above, the rule for the image feature is a rule that outputs an index such as a correlation coefficient between arrays or a fitting rate, or a value obtained by discretizing these indices with a threshold value. Note that the rule for the image feature may be omitted from the component of the function f as long as the rule for the image feature is determined uniformly. Also, on the assumption that the affected part position is fixed to the center of gravity of the target area, information on the relative positional relationship between the position of the target area and the affected part position may be omitted from the component of the function f.

  According to the medical image processing apparatus 100A according to the second embodiment described above, based on the image feature of the target area OA in the learning image such as DRR and the position of the affected part of the subject specified in the learning image. Thus, by learning the correspondence information 150, the correspondence information 150 suitable for each subject (patient) can be used. This improves convenience and further improves the reliability of treatment with radiation using an image.

  The medical image processing apparatus 100 according to the first embodiment or the medical image processing apparatus 100A according to the second embodiment may use the function f having the respiratory phase as an input parameter as the correspondence information 150. (See equation (9)). Where b is the respiratory phase. Under this assumption, the medical image processing apparatus 100A according to the second embodiment may perform processing for learning the correspondence information 150 between the image feature and the affected part position for each respiratory phase.

f (→ x, b) = l _b (u, v) (9)

<< Treatment system >>
Hereinafter, an application example of the medical image processing apparatus 100A according to the second embodiment will be described. FIG. 13 is a diagram illustrating a configuration example of the treatment system 1 including the treatment device 10 and the medical image processing device 100A.

[Treatment device]
The treatment apparatus 10 includes, for example, a bed 11, radiation sources 12-1 and 12-2, radiation detectors 13-1 and 13-2, an irradiation gate (irradiation unit) 14, a control unit 15, and an input unit. 16 and a display unit 17. Hereinafter, the hyphen in the reference numeral and the number following it indicate which radiation source and radiation detector pair is a radiation for fluoroscopy or a fluoroscopic image.

  A subject P to be treated is fixed on the bed 11. The radiation source 12-1 irradiates the subject P with fluoroscopy radiation r-1. The radiation source 12-2 irradiates the subject P with radiation r-2 for fluoroscopy from an angle different from that of the radiation source 12-1. The radiations r-1 and r-2 for fluoroscopy are, for example, X-rays.

  The fluoroscopy radiation r-1 is detected by the radiation detector 13-1, and the fluoroscopy radiation r-2 is detected by the radiation detector 13-2. The radiation detectors 13-1 and 13-2 are, for example, a flat panel detector (FPD), an image intensifier, or a color image intensifier. The radiation detector 13-1 detects and converts the energy of the radiation r-1 into a digital image, and outputs it as a fluoroscopic image TI-1 to the medical image processing apparatus 100. The radiation detector 13-2 detects the energy of the radiation r-2, digitally converts it, and outputs it as a fluoroscopic image TI-2 to the medical image processing apparatus 100. Although two sets of radiation sources and radiation detectors are shown in FIG. 13, the treatment apparatus 10 may include three or more sets of radiation sources and radiation detectors.

  The irradiation gate 14 irradiates the subject P with the treatment beam B in the treatment stage. The therapeutic beam B includes, for example, X-rays, γ-rays, electron beams, proton beams, neutron beams, heavy particle beams, and the like. Although only one irradiation gate 14 is shown in FIG. 13, the treatment apparatus 10 may include a plurality of irradiation gates. Although FIG. 13 illustrates the case where the irradiation gate is in the vertical direction of the subject P, the treatment apparatus 10 may include the irradiation gate in the horizontal direction of the subject P.

  The control unit 15 is realized by, for example, a computer device placed in a treatment room where the treatment device 10 is installed. The control unit 15 controls the radiation sources 12-1 and 12-2 so as to emit the radiations r-1 and r-2 for fluoroscopy according to the treatment plan. The input unit 16 is an input device such as a dedicated key, dial, touch panel, general-purpose keyboard, or mouse. In addition, the control unit 15 controls the irradiation gate 14 so as to irradiate the treatment beam B based on the treatment plan in the treatment stage. The display unit 17 displays an image sent from the medical image processing apparatus 100.

[Medical image processing device]
Hereinafter, application examples of the medical image processing apparatus 100A will be described. The medical image processing apparatus 100A includes, for example, a registration unit 102, a DRR generation unit 104, an affected part position calculation unit 106, image acquisition units 110-1 and 110-2, and image feature derivation units 112-1 and 112-. 2, position estimation units 114-1 and 114-2, an output unit 118, correspondence information learning units 120-1 and 120-2, and an input unit 122. In FIG. 13, the correspondence information acquisition unit 116 is not shown because it is a function of the position estimation unit 114.

  The image acquisition units 110-1 and 110-2, the image feature derivation units 112-1 and 112-2, the position estimation units 114-1 and 114-2, and the correspondence information learning units 120-1 and 120-2 Alternatively, each functional block described in the second embodiment is a functional block expressed as corresponding to the perspective images TI-1 and TI-2 input in two systems, and the basic function is the first or This is as described in the second embodiment. Therefore, the description of the function of these functional blocks is omitted.

  The medical image processing apparatus 100 </ b> A holds correspondence information 150 and plan data 152. The correspondence information 150 and the plan data 152 are stored in a storage device such as a RAM, an HDD, or a flash memory.

  Hereinafter, each functional unit of the medical image processing apparatus 100A will be described with reference to the flow of treatment. FIG. 14 is a flowchart illustrating an example of a flow of treatment performed using the treatment system 1.

  First, a treatment plan is made before treatment is performed (for example, about one week before) (step S500). This will be described with reference to FIG. FIG. 15 is a flowchart illustrating an example of the flow of a treatment plan.

  In the treatment planning stage, first, 4DCT imaging of the subject P is performed (step S502). Next, the 4DCT image is stored in the storage device as the plan data 152 (step S504). The 4DCT image is obtained by arranging n CT images that are three-dimensional volume data in time series. The period obtained by multiplying the time intervals of the n images and time-series images is set to cover a period in which the respiratory phase changes by one cycle, for example. For example, n = 10. The image used in the treatment planning stage is not limited to the 4DCT image, and a moving image of other three-dimensional volume data may be used. For example, it may be a moving image of magnetic resonance imaging (MRI).

  Next, of the n CT images, for example, one CT image is displayed, and the input of the contour of the affected part by the user is accepted for the CT image (step S506). In this step, the medical image processing apparatus 100A displays a cross-sectional image in the CT image in a state where the subject P exhales on the display device. Using the input unit 122, the user inputs the outline of the tumor that is the affected area, the outline of the organ that the treatment beam B is not desired to be irradiated, and the like while changing the displayed cross section.

  Next, the medical image processing apparatus 100A generates contour information and stores it in the storage device as part of the plan data 152 (step S508). In this step, the medical image processing apparatus 100A sets the contour for each of (n−1) CT images other than the CT image to which the user has input the contour in step S506 by deformable registration. In FIG. 13, illustration of a block for performing deformable registration is omitted.

  Next, in the medical image processing apparatus 100A, a treatment plan is drawn up (step S510). Specifically, based on the contour information input and generated in steps S506 and S508, when the affected part is in which position, where and from which direction how much treatment beam B is irradiated, gated irradiation. Is planned based on therapies and treatment methods such as tracking irradiation. This plan is made by a treatment plan program stored in the medical image processing apparatus 100A. The planned plan is displayed on the display device and confirmed by the user. The treatment plan information is stored in, for example, a storage unit included in the control unit 15 of the treatment apparatus 10. In FIG. 13, illustration of a block for creating a plan is omitted.

  In the flowchart of FIG. 15, some of the various processes described as the processes executed by the medical image processing apparatus 100A may be executed by an external apparatus. For example, a process of displaying a cross-sectional image of a CT image, a process of accepting a user input related to the contour of an affected area, a process of executing deformable registration, a process of creating a treatment plan, and the like are performed by the medical image processing apparatus 100A. It may be performed by an external treatment planning device.

  When the treatment plan is made, the medical image processing apparatus 100A performs a learning process for learning the correspondence between the image feature in the target region of the subject P and the affected part position (step S600). The learning process is performed, for example, immediately before the rehearsal in step S700 on the same day as the treatment in step S500. If the treatment spans multiple days, learning may take place only on the first treatment day. Learning may also take place on a different day from treatment and rehearsal.

  In the learning process stage, the subject P is laid on the bed 11 and fixed with a shell or the like. Then, processing is performed according to the flow shown in FIG. FIG. 16 is a flowchart illustrating an example of the flow of learning processing.

  First, rough adjustment of the bed position is performed (step S602). At this stage, the user visually confirms the position and posture of the subject P, and moves the bed 11 to a position where the treatment beam B from the irradiation gate 14 is likely to hit. Thereby, the position of the bed 11 is adjusted roughly.

  Next, fluoroscopic images TI-1 and TI-2 used to finely adjust the bed position are photographed (step S604). The fluoroscopic images TI- 1 and TI- 2 are, for example, a set of the radiation source 12-1 and the radiation detector 13-1, and the radiation source 12-2 and the radiation detector 13 at the timing when the subject P exhales. -2 sets, respectively. Since the position of the bed 11 has been roughly adjusted in step S602, the vicinity of the affected part of the subject P is shown in the fluoroscopic images TI-1 and TI-2. The image acquisition unit 110-1 of the medical image processing apparatus 100A acquires the fluoroscopic image TI-1, and the image acquisition unit 110-2 acquires the fluoroscopic image TI-2, and outputs the acquired image to the registration unit 102 (step S606).

  Then, the registration unit 102 reads out the CT image of the expiratory phase from the 4DCT image from the plan data 152, compares it with the fluoroscopic images TI-1 and TI-2, and performs the subject on the bed 11 by 3D-2D registration. Information on the position and orientation of P is calculated (step S608). The expiration phase refers to a respiratory phase in a state where the subject P has exhaled. In 3D-2D registration, CT image data is virtually installed on the bed 11 and the angle of view of the DRR, which is a virtually generated perspective image, matches the perspective images TI-1 and TI-2. This is a process for calculating the position and orientation of CT image data. Thereby, the position and orientation information of the subject P on the bed 11 is calculated. Based on the information on the position and orientation, the position of the bed 11 is precisely adjusted automatically or by a human operation. The adjusted position and orientation are output to the DRR generation unit 104 and the image feature derivation units 112-1 and 112-2.

  Next, the DRR generation unit 104 generates time-series (n) DRRs from the 4DCT image based on the position and orientation information obtained in step S608 (step S610). In step S608, since the position of the bed 11 has been precisely adjusted, the DRR generated in this step has the same angle of view as the perspective images TI-1 and TI-2.

  Next, the affected part position calculation unit 106 calculates the affected part position in each DRR in time series based on the contour information of the affected part stored as the plan data 152 and the position and orientation information obtained in step S608. (Step S612). For example, the affected part position calculation unit 106 calculates the three-dimensional center of gravity position of the affected part from the contour information on the assumption that the mass of the affected part is uniform for each of the n CT images, and corresponds the calculated center of gravity position. The position projected on the DRR is taken as the affected area position.

  Note that the medical image processing apparatus 100A may directly accept an input of an affected part position instead of an outline of the affected part. In this case, the affected part position calculation unit 106 sets the affected part position for each of (n-1) CT images other than the CT image in which the user has input the affected part position by deformable registration, and corresponds the set affected part position. The affected part position is calculated by projecting on the DRR. FIG. 17 is a diagram schematically illustrating a state where an affected part position input with respect to one CT image is developed into another CT image.

  When the affected part position is calculated by the affected part position calculating unit 106, as described in the second embodiment, the correspondence information learning units 120-1 and 120-2 determine the target region from which the image feature is derived in the DRR. (Step S614), an image feature in the target region is derived (Step S616), and the correspondence information 150 is constructed (Step S618). A re-explanation of these processes is omitted.

[rehearsal]
When the learning process ends, rehearsal is performed (FIG. 14; step S700). In the rehearsal, it is confirmed whether or not the affected part of the subject P can be tracked with the moving image of the fluoroscopic image TI. The rehearsal is performed on the same day as the treatment in step S800, immediately before the treatment. If the treatment spans multiple days, the rehearsal may be performed only on the first treatment day. In addition, positioning of the subject P is performed immediately before the rehearsal, but in this embodiment, since the positioning of the subject P has already been completed, it is not necessary to perform positioning again. It is also possible to omit rehearsals.

  FIG. 18 is an example of a flowchart showing a flow of processing performed in rehearsal. In the rehearsal, first, the position estimation units 114-1 and 114-2 read the correspondence information 150 learned in the learning process (step S702).

  Next, moving images (for example, X-rays) of the fluoroscopic images TI-1 and TI-2 by the set of the radiation source 12-1 and the radiation detector 13-1 and the set of the radiation source 12-2 and the radiation detector 13-2. (Movie) starts to be shot (step S704). Each frame of the moving image is sequentially acquired as perspective images TI-1 and TI-2 by the image acquisition units 110-1 and 110-2, and sequentially output to the image feature deriving units 112-1 and 112-2.

  Next, the image feature deriving unit 112-1 and the position estimating unit 114-1, and the image feature deriving unit 112-2 and the position estimating unit 114-2 each estimate the affected part position (step S706). The processing in this step is the same as the processing described using, for example, FIG. 6 or FIG.

  Next, the position estimation units 114-1 and 114-2 superimpose the estimated diseased part positions on the fluoroscopic images TI-1 and TI-2 and display them on the display device (step S708). FIG. 19 is a diagram illustrating an example of a rehearsal stage display screen. As shown in the figure, on the display screen at the rehearsal stage, for example, the perspective images TI-1 and TI-2 are displayed side by side. In each perspective image TI, objects A-1, A-2, and the like indicating the position of the affected area are superimposed and displayed. The rehearsal stage display screen is repeatedly updated, for example, in the loop processing (steps S706 to S718) in the flowchart of FIG. 18, so that the perspective images TI-1 and TI-2 showing the position of the affected area are viewed from the user. The video appears to be displayed. In addition, the rehearsal stage display screen is provided with a pause switch SS which is a GUI switch for instructing to pause the moving image.

  The user visually recognizes this display screen, and on the perspective images TI-1 and TI-2 from two directions, the user A visually confirms the affected part position and the affected part position displayed as the estimation result. 1 and A-2 can be confirmed. Thereby, the reliability of treatment can be improved.

  Next, the medical image processing apparatus 100A determines whether or not the temporary stop switch SS has been operated (step S710). When the temporary stop switch SS is operated, the medical image processing apparatus 100A determines whether a correction instruction has been given (step S712). The correction instruction is performed, for example, when the user clicks the correct affected part position with the mouse. When the correction instruction is given, the position estimation unit 114-1 or 114-2 corrects the affected part position (step S714). The correction instruction is given to one or both of the fluoroscopic images TI-1 and TI-2. For example, the position estimation unit 114-1 or 114-2 obtains a difference between the displayed object A1 or A2 and the position instructed to be corrected as a correction amount (correction amount), and then estimates the affected part position. Then, the estimation result is determined with the correction amount taken into account. This correction amount is also carried over to the treatment stage. Further, the position estimation unit 114-1 or 114-2 may correct the correspondence information 150 itself based on the difference. As a result, the position of the displayed object A1 or A2 is also corrected.

  The correction instruction is accepted until the pause is released by operating the pause switch SS again (step S716). The medical image processing apparatus 100A continues to display the moving image until an end instruction is given by the user (step S718).

  By such control, the user can confirm the position of the affected part used by the treatment system 1 for controlling the irradiation gate 14 in the treatment stage. On the fluoroscopic image TI, it may be difficult to directly determine the position of the affected area. Therefore, in the medical image processing apparatus 100A of the present embodiment, first, the correspondence information 150 between the image feature of the target region and the affected part position is obtained on the DRR, and the affected part position is determined from the image feature of the target region on the fluoroscopic image TI. I try to reproduce it. That is, the medical image processing apparatus 100A according to the present embodiment estimates the affected area position based on the image area of the target area that can be recognized by computer processing even on the fluoroscopic image TI and the correspondence information 150. This makes it possible to confirm whether or not the affected part position is in line with the user's intention in the rehearsal stage or the treatment stage.

[Treatment stage]
In the rehearsal, when the affected part position is confirmed and corrected, treatment is started (FIG. 14; step S800). In the treatment stage, the treatment beam B is irradiated to the affected part of the subject P. Although the subject P is positioned immediately before the treatment, in this embodiment, since the subject P has already been positioned, it is not necessary to perform positioning again.

  FIG. 20 is a flowchart illustrating an example of a flow of processing performed in the treatment stage. First, as in the rehearsal stage, the position estimation units 114-1 and 114-2 read out the correspondence information 150 learned in the learning process (step S802). Next, imaging of moving images (eg, X-ray moving images) of the perspective images TI-1 and TI-2 is started (step S804).

  Next, the image feature deriving unit 112-1 and the position estimating unit 114-1, and the image feature deriving unit 112-2 and the position estimating unit 114-2 each estimate the affected part position (step S806). The processing in this step is the same as the processing described using, for example, FIG. 6 or FIG. Then, the output unit 118 outputs the affected part position estimated in step S806 to the treatment apparatus 10 (step S808).

  Next, the control unit 15 of the treatment apparatus 10 determines whether or not the affected part position estimated in Step S806 is within a preset setting range (Step S810). When the affected part position is within the preset setting range, the control unit 15 controls the irradiation gate 14 to irradiate the treatment beam B (step S812). When the affected part position is outside the preset setting range, the control unit 15 does not irradiate the irradiation gate 14 with the treatment beam B. Thereby, the affected part of the subject P is treated by the gated irradiation method.

  The processing in steps S806 to S812 is controlled so as to end when the cumulative amount of the treatment beam B irradiated to the affected area reaches a predetermined amount (step S814). In addition, although the flow of treatment by the gated irradiation method was illustrated, the tracking irradiation method is realized by tracking the irradiation by tracking the treatment beam B to the affected part position by tracking the diseased part position in the moving image. .

  Further, the treatment system 1 may display a display screen on the display unit 17 that can compare the affected part position and the set range in the treatment stage. FIG. 21 is a diagram illustrating an example of a treatment stage display screen. As shown in the drawing, on the display screen at the treatment stage, the setting range Ta-1 corresponding to the fluoroscopic image TI-1 can be compared with the object A-1 indicating the affected part position corresponding to the fluoroscopic image TI-1. Yes. In the treatment stage display screen, the setting range Ta-2 corresponding to the fluoroscopic image TI-2 and the object A-2 indicating the affected part position corresponding to the fluoroscopic image TI-2 can be compared.

  By such control, the medical image processing apparatus 100A of the present embodiment can improve the reliability of the estimation of the affected area, and thus the reliability of the treatment. Conventionally, control is performed such that a treatment beam is irradiated based on image information around the diaphragm in a fluoroscopic image taken at the treatment stage. However, this conventional method does not provide treatment support by tracking the position of the affected area itself, and thus may not be reliable enough. On the other hand, the medical image processing apparatus 100A of the present embodiment improves the reliability of treatment by estimating the affected part position from the fluoroscopic image TI based on the correspondence between the position of the characteristic part and the affected part position. Can do.

  In addition, the medical image processing apparatus 100A according to the present embodiment can save labor for treatment. Conventionally, an operation of taking images of various respiratory phases and selecting an image of a desired respiratory phase separately from the treatment plan is performed immediately before the treatment, which is a burden on the user. On the other hand, the medical image processing apparatus 100A according to the present embodiment learns the correspondence between the position of the characteristic part and the position of the affected part, thereby omitting the above trouble and saving labor for treatment. .

  In the above embodiment, the position of the affected part of the subject P is set using the contour information generated by deformable registration from the contour information of the affected part specified by the user. The user may directly specify the center position in the fluoroscopic images TI-1 and TI-2 of the affected part of the subject P. The user designates the center position of the affected part of the subject P on the plurality of fluoroscopic images TI having different respiratory phases by using a mouse or a cross key on the keyboard. In this case, the order of processing is the order of treatment plan, shooting of moving images of fluoroscopic images TI-1 and TI-2, learning of correspondence, rehearsal, and treatment. As a result, it becomes possible to learn from an image that is closer to treatment than at the time of treatment planning, so that it can be expected that errors in tracing a tumor during treatment are reduced.

  According to at least one embodiment described above, the image acquisition unit (110) that acquires the fluoroscopic image of the subject imaged by the imaging device, and part or all of the fluoroscopic image acquired by the image acquisition unit. Derived by the image feature deriving unit based on the image feature deriving unit (112) for deriving the image feature for the target area (OA) and the correspondence information (150) indicating the correspondence between the image feature and the position of the search part. By having the position estimation unit (114) that estimates the position of the search part of the subject in the fluoroscopic image from the image feature, the reliability can be improved.

  Although several embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are also included in the invention described in the claims and the equivalents thereof.

  That is, the present invention is not limited to the above-described embodiment as it is, and the constituent elements can be modified and embodied without departing from the spirit of the invention in the implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of components disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, constituent elements over different embodiments may be appropriately combined.

DESCRIPTION OF SYMBOLS 1 ... Treatment system, 10 ... Treatment apparatus, 100, 100A ... Medical image processing apparatus, 102 ... Registration part, 104 ... DRR production | generation part, 106 ... Affected part position calculation part, 110 ... Image acquisition part, 112 ... Image feature derivation part , 114 ... position estimation part, 116 ... correspondence information acquisition part, 118 ... output part, 120 ... correspondence information learning part, 122 ... input part, 150 ... correspondence information, 152 ... plan data

Claims (17)

An image acquisition unit for acquiring a fluoroscopic image of the subject imaged by the imaging device;
An image feature deriving unit for deriving image features for a target region that is a part or all of the fluoroscopic image acquired by the image acquiring unit;
A position estimation unit that estimates the position of the search site of the subject in the fluoroscopic image from the image feature derived by the image feature deriving unit, based on correspondence information indicating a correspondence relationship between the image feature and the position of the search site; ,
A medical image processing apparatus comprising: The correspondence information further includes information on a relative positional relationship between the position of the target region in the fluoroscopic image and the position of the search part,
The position estimation unit estimates a position of a search portion of the subject in the fluoroscopic image based on the image feature derived by the image feature deriving unit and a position of the target region in the fluoroscopic image;
The medical image processing apparatus according to claim 1. The correspondence information includes the position of the search region at one position on the fluoroscopic image by using the position of the target region in the fluoroscopic image and the image feature derived by the image feature deriving unit as input information. Contains derivation rules to derive certain probabilities,
The image feature deriving unit derives an image feature while changing the position of the target region in the fluoroscopic image, and the position estimating unit calculates the search site from the image feature derived for the plurality of target regions by the image feature deriving unit. Deriving the probability of being the position of the search part corresponding to a plurality of positions in the fluoroscopic image, respectively,
The medical image processing apparatus according to claim 2. The correspondence information includes, as input information, the position of the target region in the fluoroscopic image and the image feature derived by the image feature deriving unit, and the position of the search part at a plurality of positions on the fluoroscopic image. Contains derivation rules to derive certain probabilities,
The image feature deriving unit derives an image feature for one target region in the fluoroscopic image, and the position estimating unit calculates the image feature on the fluoroscopic image from the image feature derived for the one target region by the image feature deriving unit. A probability of being the position of the search part corresponding to a plurality of positions of
The medical image processing apparatus according to claim 1 or 2. The position estimation unit determines the position at which the probability of being the position of the search part corresponding to the plurality of positions on the fluoroscopic image is the maximum value as the position of the search part of the subject in the subject. presume,
The medical image processing apparatus according to claim 3 or 4. The position estimation unit estimates the derived gravity center of the probability distribution as the position of the search part corresponding to a plurality of positions on the fluoroscopic image as the position of the search part of the subject in the subject. ,
The medical image processing apparatus according to claim 3 or 4. The position estimation unit calculates the center of the derived local region having the maximum sum in the local region of the probability of being the position of the search part corresponding to a plurality of positions on the fluoroscopic image in the subject. Estimating the position of the search site of the subject,
The medical image processing apparatus according to claim 3 or 4. The correspondence information includes, as input information, the position of the target region in the fluoroscopic image and the image feature derived by the image feature deriving unit, so that the search site at one or more positions on the fluoroscopic image Including a plurality of derivation rules for deriving the probability of being the position of
The position estimation unit derives the probability of being the position of the search part corresponding to one or a plurality of positions on the fluoroscopic image by superimposing the probabilities derived by the plurality of derivation rules, Based on the derived probability, estimate the position of the search part of the subject in the fluoroscopic image,
The medical image processing apparatus according to claim 1 or 2. The position estimation unit superimposes the probabilities derived by the plurality of derivation rules by obtaining a sum of probabilities derived by the plurality of derivation rules,
The medical image processing apparatus according to claim 8. The position estimation unit superimposes the probabilities derived by the plurality of derivation rules by obtaining a product of the probabilities derived by the plurality of derivation rules,
The medical image processing apparatus according to claim 8. The position estimation unit superimposes the probabilities derived by the plurality of derivation rules by kernel density estimation,
The medical image processing apparatus according to claim 8. The image feature deriving unit derives information based on a luminance gradient of each pixel in the target region as the image feature;
The medical image processing apparatus according to any one of claims 1 to 11. A learning unit that learns the correspondence information based on the image characteristics of the target region in the learning image and the position of the search part specified in the learning image;
The medical image processing apparatus according to any one of claims 1 to 12. The learning unit further learns the correspondence information including a relative positional relationship between the target region and the position of the search part based on the position of the target region in the learning image.
The medical image processing apparatus according to claim 13. A specifying unit for specifying a position of a search portion of the subject in a learning image obtained by imaging the subject;
A learning unit for deriving an image feature for a target region that is a part or all of the learning image, and learning correspondence information indicating a correspondence relationship between the image feature and a position of a search part based on the derived image feature; ,
A medical image processing apparatus comprising: The learning unit further learns the correspondence information including a relative positional relationship between the target region and the position of the search part based on the position of the target region in the learning image.
The medical image processing apparatus according to claim 15. The medical image processing apparatus according to any one of claims 1 to 14,
An irradiation unit for irradiating the subject with a therapeutic beam;
A control unit that controls the irradiation unit to irradiate the treatment beam with respect to the position of the search portion of the subject estimated by the position estimation unit of the medical image processing apparatus;
A treatment system comprising:

Images