CN102814006B - Image contrast device, patient positioning device and image contrast method - Google Patents

Abstract

The present invention relates to an image contrast device, a patient positioning device, and an image contrast method. The image contrast device comprises a contrast processing unit. The contrast processing unit is used for contrasting a three-dimensional reference image and a three-dimensional current image, and calculating positional correction so that the position and posture of the patient in the current image site are consistent with the position and posture of the patient in the reference image. The contrast processing unit comprises: a primary contrast unit, which conducts primary contrast according to the reference image; and a secondary contrast unit, which conducts secondary contrast based on a predetermined template region for a predetermined retrieval target region, wherein the predetermined template region is generated according to one of the reference image or the current image and based on the primary contrast result, and the predetermined retrieval target area is generated according to one of the reference image or the current image different from the predetermined template region and based on the result of the primary contrast.

Description

Image comparison device, patient positioning device and image comparison method

technical field

The present invention relates to an image collating device and a patient positioning device. The image collating device utilizes CT image data, etc. , and the patient positioning device uses the image comparison device to position the patient at the radiation irradiation position where the radiation is irradiated.

Background technique

In recent years, among radiation therapy devices for cancer treatment, cancer therapy devices using particle beams such as protons and heavy ions (particularly referred to as particle beam therapy devices) have been developed and constructed. It is well known that, compared with conventional radiation therapy such as X-rays and γ-rays, particle beam therapy using particle beams can irradiate cancer patients intensively. Treat normal cells.

In particle beam therapy, it is important to precisely irradiate particle beams to affected areas such as cancer. Therefore, when particle beam therapy is performed, the patient is fixed with a fixture or the like so as not to be misaligned with respect to the treatment table in the treatment room (irradiation room). In order to accurately locate the affected part such as cancer in the radiation irradiation range, the patient is roughly fixed with a laser pointer or the like, and then the patient's affected part is precisely positioned using X-ray images or the like.

In Patent Document 1, a bed positioning device and its positioning method are proposed. In the bed positioning device and positioning method, the reference image of the X-ray fluoroscopic image and the current image captured by the X-ray receiver are compared. The same position of the same plurality of monuments is not specified in any of the images, and two-stage pattern matching is performed to generate positioning information for driving the treatment table. In one pattern matching, a second set area is set for the 2D current image, and the second set area is approximately the same size as the first set area, wherein the first set area includes the set area for the 2D reference image The isocenter (beam irradiation center) moves the second setting area sequentially within the area of the 2D current image, and at each position of the second setting area, the 2D benchmark in the first setting area The image is compared with the 2D current image in the second set area, and the second set area having the 2D current image most similar to the 2D reference image in the first set area is extracted. In the second pattern matching, the 2D current image in the second set area extracted in the first pattern matching is compared with the 2D reference image in the first set area, and the pattern matching is performed so that The two images are most consistent.

prior art literature

patent documents

Patent Document 1: Japanese Patent No. 3748433 (paragraphs 0007-0009, 0049, Fig. 8, Fig. 9)

The technical problem to be solved by the invention

Since the shape of the affected part is a three-dimensional shape, when positioning the affected part to the position of the affected part in the treatment plan, the positioning accuracy can be higher by using the three-dimensional image than by using the two-dimensional image. In general, when creating treatment plan data, X-ray CT (Computed Tomography: Computed Tomography) images are used to determine the three-dimensional shape of the affected part. In recent years, there has been a demand for an X-ray CT device to be installed in the treatment room, and the current X-ray CT image captured by the X-ray CT device during treatment and the X-ray CT image during treatment planning are required for positioning. X-ray fluoroscopy images do not reflect the affected part as soft tissue well, so the bones are basically used for position matching, and X-ray CT images are used for positioning because it is possible to position the affected parts reflected in X-ray CT images. match.

Therefore, in conventional two-stage pattern matching, a case where the reference image and the current image are expanded to a 3D image is considered. The 3D reference image and the 3D current image include a plurality of tomographic images (slice images) captured by an X-ray CT apparatus. The 3D current image is assumed to have a small number of images from the perspective of X-ray radiation, so it is necessary to compare the 3D reference image with dense image information and the 3D reference image with sparser image information than the 3D reference image. current image for comparison. In the existing two-stage pattern matching, there is a problem that although comparison can be made between 2D reference images having image information of the same density and the 2D current image, the comparison between 3D reference images with different image information densities When comparing with the 3D current image, the 2-stage pattern matching cannot be realized by simply increasing the image dimension of the prior art from 2D to 3D. That is, there is a problem that pattern matching cannot be performed simply once from the 3D reference image in the first set area to the 3D current image in the second set area as in the prior art. The extracted 3D current image in the second set area is simply compared with the 3D reference image in the first set area, so as to achieve the most consistent pattern matching between the two 3D images.

Contents of the invention

The object of the present invention is to realize high-precision two-stage pattern matching (2-stage pattern matching) even when the number of tomographic images in the 3D current image is smaller than that of the 3D reference image when positioning a patient for radiation therapy. phase comparison).

Technical solutions for technical problems

The image collation device according to the present invention includes: a 3D image input unit, which respectively reads a 3D reference image captured during radiation therapy treatment planning and a 3D current image captured during treatment; a collation processing unit that compares the 3D reference image and the 3D current image, and calculates a body posture correction amount so that the position and posture of the affected part in the 3D current image are consistent with the position and posture of the affected part in the 3D reference image; unanimous. The collation processing part has: a primary collation part; the primary collation part performs primary pattern matching on the 3D current image based on the 3D reference image; The search target area is subjected to secondary pattern matching, wherein the specified template area is generated based on one of the 3D reference image or the 3D current image based on the primary pattern matching result, and the specified search target area is generated based on the matching with the specified template area. The other of the 3D reference image or the 3D current image based on a different base is generated based on the primary pattern matching result.

Invention effect

The image matching device according to the present invention performs primary pattern matching on the 3D current image based on the 3D reference image, and then generates a predetermined template region and a predetermined search target region based on the primary pattern matching result, and executes the search target region and the predetermined search target region. 2-stage pattern matching in the template area enables high-precision 2-stage pattern matching even when the number of tomographic images in the 3D current image is smaller than that of the 3D reference image.

Description of drawings

FIG. 1 is a diagram showing configurations of an image collating device and a patient positioning device according to Embodiment 1 of the present invention.

Fig. 2 is a diagram showing the overall equipment structure related to the image collation device and the patient positioning device of the present invention.

3 is a diagram showing a three-dimensional reference image and a reference image template area according to Embodiment 1 of the present invention.

FIG. 4 is a diagram showing a three-dimensional current image according to Embodiment 1 of the present invention.

FIG. 5 is a diagram illustrating a primary pattern matching method according to Embodiment 1 of the present invention.

FIG. 6 is a diagram illustrating the relationship between a reference image template region and a slice image in the primary pattern matching method of FIG. 5 .

7 is a diagram showing primary extraction regions of slice images extracted by the primary pattern matching method according to Embodiment 1 of the present invention.

FIG. 8 is a diagram illustrating a secondary pattern matching method according to Embodiment 1 of the present invention.

FIG. 9 is a diagram illustrating the relationship between a reference image template region and a slice image in the secondary pattern matching method of FIG. 8 .

FIG. 10 is a diagram illustrating a primary pattern matching method according to Embodiment 2 of the present invention.

FIG. 11 is a diagram illustrating the relationship between a reference image template region and a slice image in the primary pattern matching method shown in FIG. 10 .

12 is a diagram showing a three-dimensional reference image after posture conversion according to Embodiment 2 of the present invention.

FIG. 13 is a diagram illustrating a secondary pattern matching method according to Embodiment 2 of the present invention.

FIG. 14 is a diagram showing configurations of an image collating device and a patient positioning device according to Embodiment 3 of the present invention.

Detailed ways

Embodiment 1

1 is a diagram showing configurations of an image collating device and a patient positioning device according to Embodiment 1 of the present invention, and FIG. 2 is a diagram showing overall equipment configurations related to the image collating device and patient positioning device of the present invention. In FIG. 2 , 1 is a CT simulator room for performing a treatment plan to be performed before radiation therapy. In this CT simulator room, there are a CT gantry 2 and a top plate 3 of a bed for CT image capturing, and the patient 4 lying on the top plate 3, CT image data for treatment planning is taken so as to include the affected part 5. On the other hand, 6 is a treatment room for radiotherapy. In this treatment room, there are CT gantry 7 and rotary treatment table 8, and there is a ceiling 9 on the top of the rotary treatment table 8, and a patient 10 is placed on the ceiling. 9, and CT image data for positioning are taken so as to include the affected part 11 during treatment.

Here, positioning refers to calculating the positions of the patient 10 and the affected part 11 during treatment using CT image data based on the treatment plan, calculating body posture correction amounts so as to match the treatment plan, and performing position matching so that the affected part 11 during treatment reaches the radiation. The therapeutic beam irradiates the center 12 . Position matching is achieved by moving the position of the top plate 9 by driving and controlling the rotary treatment table 8 in a state where the patient 10 is placed on the top plate 9 . The rotary table 8 is capable of driving and correcting six degrees of freedom of translation/rotation, and by rotating the top plate 9 of the rotary table 8 by 180 degrees, it can be moved from the CT imaging position (indicated by a solid line in FIG. A certain treatment position of the irradiated irradiation couch 13 (indicated by a dotted line in FIG. 2 ). In addition, although it is shown in FIG. 2 that the CT shooting position and the treatment position have a 180-degree opposing positional relationship, the arrangement is not limited to this, and the positional relationship between the two may be other angles such as 90 degrees.

The CT image data for treatment planning and the CT image data for positioning are transmitted to the positioning computer 14 . The CT image data for treatment planning becomes a 3D reference image, and the CT image data for positioning becomes a 3D current image. The image comparison device 29 and the patient positioning device 30 in the present invention are all related to the computer software existing in the positioning computer 14, and the image comparison device 29 calculates the above-mentioned body position correction amount (translation amount, rotation amount), and the patient positioning device 30 It includes the image collating device 29 and also has a function of calculating parameters for controlling each drive shaft of the rotary treatment table 8 (simply referred to as the treatment table 8 as the case may be) based on the body posture correction amount. The patient positioning device 30 controls the treatment table 8 based on the matching result (comparison result) obtained by the image collating device 29 to guide the affected part of the particle beam therapy so that it is positioned at the beam irradiation center 12 of the treatment device.

In conventional radiotherapy positioning, DRR (Digitally Reconstructed Radiography: Digitally Reconstructed Radiography) images generated from CT image data based on treatment plans or X-ray fluoroscopy images taken at the same time are compared with the time of treatment. The X-ray fluoroscopy images taken in the treatment room of the patient are used to calculate the position offset. In the X-ray fluoroscopy image, since the affected part which is soft tissue cannot be well reflected, position matching using bones is basically performed. The positioning using CT image data described in this embodiment has the following features: a CT gantry 7 is installed in the treatment room 6, and the positioning is performed by using the CT image data immediately before treatment and the CT image data for treatment planning. Matching, so the affected part can be directly described, and the location of the affected part can be matched.

Next, the calculation procedure of the above-mentioned body posture correction amount by the image collating device 29 and the patient positioning device 30 in this embodiment will be described. Fig. 1 shows the relationship between each data processing part that constitutes the image comparison device and the patient positioning device. Here, the image comparison device 29 is equipped with: a 3-dimensional image input part 21 that reads CT image data; a comparison processing part 22; a display unit 23 ; and a comparison result output unit 24 . The patient positioning device 30 is a device in which the treatment table control parameter calculation unit 26 is added to the image collating device 29 .

As described above, the 3D reference image is data captured for treatment planning when performing treatment planning, and is characterized in that affected area information (affected area shape, etc.) indicating an affected area to be treated with particle beams is manually input. The 3D current image is data captured for patient positioning during treatment, and is characterized in that the number of tomographic images (also referred to as slice images) is small in view of suppressing exposure to X-rays.

In the present invention, a two-stage pattern matching structure is adopted: a pattern matching is performed on the 3D current image based on the 3D reference image; then, based on the result of the primary pattern matching, a specified template area and a specified search object area are generated, Using this predetermined template area, pattern matching is performed twice in the same direction or in the opposite direction. In the two-step pattern matching, by making the matching parameters for the first pattern matching different from the matching parameters for the second pattern matching, high-speed and high-precision processing can be realized. For example, there is a method of performing pattern matching once at a low resolution with a wide area as an object, and using the found template area or search target area at a high resolution with a narrowed area as an object. Do 2 pattern matches.

The three-dimensional image input unit 21 will be described. The 3D image input unit 21 reads an image group composed of a plurality of tomographic images captured by an X-ray CT apparatus, and image data in the form of DICOM (Digital Imaging and Communications in Medicine) (slice image group). ) as 3D volume data. The CT image data for treatment planning is 3-dimensional volume data at the time of treatment planning, that is, a 3-dimensional reference image. The CT image data for positioning is 3D volume data at the time of treatment, that is, a 3D current image. In addition, CT image data is not limited to the DICOM format, and may be data in other formats.

The collation processing unit 22 compares the 3D reference image and the 3D current image (pattern matching), and calculates a body posture correction amount so that the position and posture of the affected part in the 3D current image are consistent with the position and posture of the affected part in the 3D reference image. consistent. The collation result display part 23 displays the result after collation by the collation processing part 22 on the display screen of the positioning computer 14 (the following body position correction amount, or the 3-dimensional current image and the 3-dimensional reference after being moved by the body position correction amount) Images displayed by superimposing images, etc.). The comparison result output unit 24 outputs the correction amount when the comparison processing unit 22 compares the 3D reference image and the 3D current image, that is, the body posture correction amount (translation amount, rotation amount) calculated by the comparison processing unit 22 . The treatment table control parameter calculation part 26 converts the output value of the comparison result output part 24 (3-axis translation [ΔX, ΔY, ΔZ], 3-axis rotation [ΔA, ΔB, ΔC], a total of 6 degrees of freedom) into the corresponding treatment table The parameters for controlling each axis of 8 are calculated parameters. The treatment table 8 drives the drive devices of the axes of the treatment table 8 based on the treatment table control parameters calculated by the treatment table control parameter calculation unit 26 . Thereby, the body position correction amount can be calculated so as to match the treatment plan, and position matching can be performed so that the affected part 11 at the time of treatment reaches the beam irradiation center 12 of the radiotherapy.

The comparison processing unit 22 includes: a position and posture conversion unit 25 ; a primary comparison unit 16 ; a secondary comparison unit 17 ; and a reference template region generation unit 18 . The position and posture conversion unit 25 changes the position and posture of the object data when the pattern matching is performed once or twice. The primary matching unit 16 performs primary pattern matching on the 3D current image based on the 3D reference image. The secondary comparison unit 17 performs secondary pattern matching on the predetermined search target region according to the predetermined template region, wherein the predetermined template region is generated based on the result of primary pattern matching based on one of the 3D reference image or the 3D current image, On the other hand, the predetermined search target region is generated based on the primary pattern matching result from the other of the 3D reference image or the 3D current image, which is different from the generation basis of the predetermined template region.

The collation processing unit 22 will be described in detail using FIGS. 3 to 9 . 3 is a diagram showing a three-dimensional reference image and a reference image template area according to Embodiment 1 of the present invention. FIG. 4 is a diagram showing a three-dimensional current image according to Embodiment 1 of the present invention. FIG. 5 is a diagram illustrating a primary pattern matching method according to Embodiment 1 of the present invention. FIG. 6 is a diagram illustrating the relationship between a reference image template region and a slice image in the primary pattern matching method of FIG. 5 . 7 is a diagram showing primary extraction regions of slice images extracted by the primary pattern matching method according to Embodiment 1 of the present invention. FIG. 8 is a diagram illustrating a secondary pattern matching method according to Embodiment 1 of the present invention. FIG. 9 is a diagram illustrating the relationship between a reference image template region and a slice image in the secondary pattern matching method of FIG. 8 .

The reference template region generation unit 18 of the comparison processing unit 22 generates a reference image template region 33 from a three-dimensional reference image 31 using the shape of the affected part (affected part information) input during treatment planning. The three-dimensional reference image 31 consists of a plurality of slice images 32 to form. In FIG. 3 , an example composed of five slice images 32 a , 32 b , 32 c , 32 d , and 32 e is shown for convenience. The shape of the affected part is input as ROI (Region of Interest) 35 and as a closed contour surrounding the affected part in each slice image. A region including the above-mentioned closed contour can be used as, for example, a circumscribing quadrilateral 34 , and a cuboid region including each circumscribing quadrangle 34 can be used as a template region. Let this template area be the reference image template area 33 . The primary matching unit 16 of the matching processing unit 22 performs pattern matching once to match the reference image template region 33 to the three-dimensional current image 36 . The three-dimensional current image 36 shown in FIG. 4 shows an example composed of three slice images 37a, 37b, and 37c. The current image area 38 shown in FIG. 5 is represented as a cuboid including three slice images 37a, 37b, and 37c. As shown in FIG. 5 , the reference image template area 33 ( 33 a , 33 b , 33 c ) is moved in a raster scanning manner in the current image area 38 to calculate a correlation value with the three-dimensional current image 36 . As the correlation value, all correlation values used in image matching (image collation) such as normalized cross-correlation values can be used.

The reference image template area 33a moves in the slice image 37a along the scanning path 39a in a raster scanning manner. Similarly, the reference image template region 33b moves in the slice image 37b in a raster scan along the scan path 39b, and the reference image template region 33c moves in the slice image 37c in a raster scan along the scan path 39c. In order to simplify the drawing, the scanning paths 39b, 39c are schematically shown.

When pattern matching is performed once, as shown in FIG. 6 , image comparison is performed for each slice image 53 constituting the reference image template region 33 with the slice image 37 constituting the current image region 38 . The slice image 53 is an image divided by the reference image template region 33 in the slice image 32 of the three-dimensional reference image 31 . The reference image template area 33 is composed of five slice images corresponding to the five slice images 32a, 32b, 32c, 32d, and 32e of the three-dimensional reference image. Therefore, when pattern matching is performed once, image comparison is performed on the slice images 37 a of the 3D current image 36 using the five slice images in the reference image template area 33 . For the slice images 37b and 37c of the 3D current image 36, image comparison is similarly performed.

The primary comparison unit 16 extracts the primary extraction region 43 from each slice image 37 of the three-dimensional current image 36 such that the region having the highest correlation value between the current image region 38 and the current image template region 33 is included. As shown in FIG. 7 , the primary extraction region 43 a is extracted from the slice image 37 a of the three-dimensional current image 36 . Similarly, primary extraction regions 43 b and 43 c are extracted from the slice images 37 b and 37 c of the three-dimensional current image 36 . The primary extraction current image area 42 which is the search target area for the secondary pattern matching is generated so as to include the primary extraction areas 43a, 43b, and 43c. In this way, the primary comparison unit 16 generates the primary extracted current image region 42 as the search target region for the secondary pattern matching.

Here, since the postures (3-axis rotation) of the 3D reference image 31 and the 3D current image 36 are inconsistent in the state before positioning, in a simple raster scan as shown in FIG. When the number of slices is small, it is not a problem to extract the primary extraction region 43 for secondary pattern matching, although high-precision matching that can detect even an angular offset cannot be performed. Therefore, in the primary pattern matching, the correlation value is calculated without detecting the angular offset, and in the subsequent secondary pattern matching, high-accuracy matching is performed so that even the angular offset can be detected.

The two-stage pattern matching will be described. In the secondary pattern matching, the position and orientation conversion unit 25 of the collation processing unit 22 generates a position and orientation converted template area 40 obtained by converting the position and orientation of the reference image template area 33 generated from the three-dimensional reference image 31 . In the secondary pattern matching, as shown in FIGS. 8 and 9 , when performing the matching, the amount of change in posture (rotation along three axes) of the reference image template region 33 is added as a parameter. The secondary comparison unit 17 performs continuous comparison between the position and posture conversion template region 40 after the position and posture conversion by the position and posture conversion unit 25 and the primary extraction current image region 42 of the 3D current image 36 with a small number of slice images. Angle offsets are also included for high precision matching. In this way, high-precision two-stage pattern matching including angular offset can be realized. By using a narrow range including the region obtained by the primary pattern matching as the search range of the secondary pattern matching, it is possible to use the low-resolution and wide range as the target to perform the primary pattern matching. The primary extraction current image area 42 including the found primary extraction area 43 performs pattern matching twice at high resolution, and the time required for pattern matching can be shortened.

The primary extraction current image area 42 shown in FIG. 8 is represented as a cuboid including three primary extraction areas 43a, 43b, and 43c. The position-posture conversion template region 40a, which is the reference image template region after the position-posture conversion, moves in the primary extraction region 43a of the slice image 37a in a raster-scanning manner along the scanning path 39a. Similarly, the position and posture transformation template region 40b, which is the transformed reference image template region, moves along the scanning path 39b in the primary extraction region 43b of the slice image 37b in a raster scanning manner, and serves as the transformed position and posture template region. The position-posture conversion template region 40c of the reference image template region moves along the scan path 39c in the primary extraction region 43c of the slice image 37c in a raster scan manner. In order to simplify the drawing, the scanning paths 39b, 39c are schematically shown.

When performing secondary pattern matching, as shown in FIG. 9 , the secondary comparison unit 17 is used to convert the cross section 41 of the template region 40 in position and posture and the primary extraction region 43 of the slice image 37 constituting the primary extraction current image region 42 . Compare images between them. In addition, image comparison may be performed between the slice image 55 that is divided into the slice image 37 of the three-dimensional current image 36 by extracting the current image region 42 once and the cross section 41 . A cross-section 41 of a position-posture conversion template region 40 is generated from a plurality of slice images 32 of a three-dimensional reference image 31 . For example, the data of the cross section 41 are extracted from the plurality of slice images 32 constituting the three-dimensional reference image 31 . Usually, the data density of the section 41 of the position and posture transformation template area 40 is different from the data density of the primary extraction area 43 of the 3D current image 36 , but it is only necessary to calculate the correlation value of each pixel of the section 41 . In addition, the section 41 of the position and posture conversion template area 40 may also include data that has been completed so that the data density of the section 41 is the same as the data density of the primary extraction area 43 of the 3D current image 36 .

Here, the two-stage pattern matching method of Embodiment 1 is summarized. First, the reference template region generation unit 18 of the comparison processing unit 22 generates a reference image template region 33 from the three-dimensional reference image 31 (reference image template region generation step). The primary matching unit 16 performs primary pattern matching on the three-dimensional current image 36 based on the reference image template region 33 (primary pattern matching step). The primary pattern matching performs image comparison for each slice image 53 constituting the reference image template region 33 with the slice image 37 constituting the current image region 38 . The primary comparison unit 16 calculates the correlation value between the current image area 38 and the reference image template area 33 each time the reference image template area 33 is scanned (correlation value calculation step), and extracts a single The region 43 is extracted so as to include the region with the highest correlation value between the current image region 38 and the reference image template region 33 (1 extraction region extraction step). The primary comparison unit 16 generates the primary extraction current image area 42 as the search target area for the secondary pattern matching so as to include the primary extraction area 43 (search target area 43) for each slice image 37 constituting the current image area 38. build step). The two-stage pattern matching method of Embodiment 1 includes: a reference image template region generation step; a first pattern matching step; and a second pattern matching step described below. One pattern matching step includes: a correlation value calculation step; one extraction region extraction step; and a search object generation step.

Next, the secondary comparison unit 17 of the comparison processing unit 22 extracts the current image from the three-dimensional current image 36 based on the position and posture transformation template region 40 obtained by converting the position and posture of the reference image template region 33 by the position and posture transformation unit 25. The area 42 performs pattern matching twice (two pattern matching steps). Secondary pattern matching generates a plurality of cross-sections 41 of the position and posture transformation template region 40 transformed into a predetermined position and posture (section generation step), and for each cross-section 41, the slice image 37 constituting the primary extraction current image region 42 is generated. Image comparison is performed between the primary extraction region 43 or the slice image 55 and the section 41 . Every time the position and posture conversion template region 40 is scanned, the secondary comparison unit 17 calculates the correlation value between the primary extracted current image region 42 and the plurality of sections 41 of the position and posture conversion template region 40 (correlation value calculation step). In addition, the position and posture conversion unit 25 performs conversion to a position and posture different from the previous position and posture (position and posture conversion step), and the secondary comparison unit 17 generates a plurality of cross-sections 41 ( Section generation step), when scanning the position and posture transformation template region 40 each time, calculate and extract the correlation value of multiple sections 41 between the current image region 42 and the position and posture transformation template region 40 (correlation value calculation step). The secondary comparison unit 17 of the comparison processing unit 22 selects the position and posture relationship (position and posture information) between the 3-dimensional reference image and the 3-dimensional current image, which is the highest correlation value among the calculated correlation values, as the optimal solution (the optimal solution selected steps). Pattern matching is thereby achieved such that the 3D reference image and the 3D current image - the two 3D images are most consistent. The two pattern matching steps include: the section generation step; the correlation value calculation step; the position and posture transformation step; and the optimal solution selection step.

After the pattern matching is completed, the collation processing unit 22 converts the position and posture of the template region 40 according to the position and posture whose correlation value is the highest among the calculated correlation values, and calculates when the 3D reference image 31 and the 3D current image 36 are compared. The body position correction amount (translation amount, rotation amount) (body position correction amount calculation step). The collation result display unit 23 displays, on the monitor screen of the computer 14 , the body posture correction amount, an image in which the 3D current image moved by the body posture correction amount is superimposed on the 3D reference image, and the like. The comparison result output unit 24 outputs the body posture correction amount (translation amount, rotation amount) when the comparison processing unit 22 compares the 3D reference image 31 and the 3D current image 36 (body posture correction amount output step). The treatment table control parameter calculation part 26 converts the output value of the comparison result output part 24 (3-axis translation [ΔX, ΔY, ΔZ], 3-axis rotation [ΔA, ΔB, ΔC], a total of 6 degrees of freedom) into the corresponding treatment table 8, the parameters controlled by each axis are the calculated parameters (the calculation step of the control parameters of the treatment table). The treatment table 8 drives the driving device of each axis of the treatment table 8 based on the treatment table control parameters calculated by the treatment table control parameter calculation unit 26 (treatment table driving step).

The image collating device 29 according to Embodiment 1 performs one-time pattern matching from the three-dimensional reference image 31 to the three-dimensional current image 36, and then, based on the result of the one-time pattern matching, generates a predetermined image from the three-dimensional reference image 31. The position and orientation of the template area for the secondary pattern matching is transformed into the template area 40, and the primary extraction current image area 42, which is a predetermined search target area for the secondary pattern matching, is generated from the 3D current image 36 so as to include the primary extraction. Therefore, even when the number of tomographic images (the number of slice images) of the 3D current image 36 is smaller than that of the 3D reference image 31, high-precision two-stage pattern matching can be realized.

Since the image collating device 29 according to the first embodiment can realize highly accurate two-step pattern matching even when the number of tomographic images (the number of slice images) of the 3D current image 36 is smaller than that of the 3D reference image 31, Therefore, the number of tomographic images of the three-dimensional current image 36 from the X-ray CT apparatus at the time of position matching can be reduced, and the radiation exposure of the patient due to the X-ray CT apparatus at the time of position matching can be reduced.

The image collating device 29 according to Embodiment 1 generates a primary extracted current image region 42 based on the result of primary pattern matching from the 3D reference image 31 to the 3D current image 36 , and makes the region smaller than the current image region 38 . The narrow primary extraction current image area 42 is used as the search object, thereby using the primary extraction current image area 42 including the primary extraction area 43, high-resolution secondary pattern matching can be performed, and the time required for pattern matching can be shortened. The primary extraction region 43 is found by performing primary pattern matching with low resolution and targeting a wide range.

The patient positioning device 30 according to Embodiment 1 can match the position and posture at the time of treatment planning based on the body posture correction amount calculated by the image collating device 29 . Since the position and posture at the time of treatment planning can be matched, position matching can be performed so that the affected part 11 at the time of treatment reaches the beam irradiation center 12 of radiation therapy.

The patient positioning device 30 according to Embodiment 1 can use the position and posture conversion unit 25 to generate a position and posture conversion template region 40 suitable for matching from the reference image template region 33 obtained from the 3D reference image 31 . Up to the 3D current image 36 whose number of tomographic images (the number of slice images) is smaller than that of the 3D reference image 31, high-precision two-stage pattern matching including angular offset can be realized.

The image collating device 29 according to Embodiment 1 includes a three-dimensional image input unit 21 that reads the three-dimensional reference image 31 captured during the treatment planning of radiation therapy and the three-dimensional reference image captured during the treatment. 3-dimensional current image 36; and a comparison processing unit 22, which compares the 3-dimensional reference image 31 and the 3-dimensional current image 36, and calculates a body position correction amount so that the position and posture of the affected part in the 3-dimensional current image 36 are consistent with The position and posture of the affected part in the 3D reference image 31 are consistent, and the comparison processing part 22 has: a primary comparison part 16, and the primary comparison part 16 performs pattern matching on the 3D current image 36 based on the 3D reference image 31; And the secondary comparison unit 17, the secondary comparison unit 17 performs secondary pattern matching on the predetermined search object region 42 based on the predetermined template region (position and posture transformation template region 40), wherein the predetermined template region is based on the three-dimensional reference image 31 or one of the 3D current images 36 and is generated based on the primary pattern matching result, and the predetermined search object region 42 is based on the 3D reference image 31 or The other one of the 3D current image 36 is generated based on the primary pattern matching result. Therefore, even when the number of tomographic images in the 3D current image 36 is smaller than that of the 3D reference image 31, high-precision 2 stage pattern matching.

The patient positioning device 30 according to Embodiment 1 includes: an image collating device 29 ; and a treatment table control parameter calculation unit 26 for controlling the treatment table based on the body posture correction amount calculated by the image collating device 29 8, and the image comparison device 29 includes: a 3-dimensional image input unit 21, the 3-dimensional image input unit 21 respectively reads the 3-dimensional reference image 31 captured during the treatment plan of radiation therapy and the 3-dimensional reference image 31 captured during the treatment 3-dimensional current image 36; and a comparison processing unit 22, which compares the 3-dimensional reference image 31 and the 3-dimensional current image 36, and calculates a body position correction amount so that the position and posture of the affected part in the 3-dimensional current image 36 It matches the position and orientation of the affected part in the three-dimensional reference image 31 . The collation processing unit 22 includes: a primary collation unit 16 that performs primary pattern matching on the 3D current image 36 based on the 3D reference image 31 ; and a secondary collation unit 17 that performs pattern matching on the basis of the The predetermined template area (position and posture transformation template area 40) performs secondary pattern matching on the predetermined search object area 42, wherein the predetermined template area is based on one of the 3D reference image 31 or the 3D current image 36 and is based on the primary pattern The predetermined search object region 42 is based on the primary pattern based on the other of the 3D reference image 31 or the 3D current image 36 which is different from the generation basis of the predetermined template region (position and posture transformation region 40). Therefore, even when the number of tomographic images in the 3D current image 36 is smaller than that of the 3D reference image 31, high-precision positioning can be performed.

Embodiment 1 relates to an image comparison method. The image comparison method compares the 3D reference image 31 captured during radiation therapy treatment planning with the 3D current image 36 captured during treatment; the image comparison method includes: 1 time A pattern matching step, the first pattern matching step performs one pattern matching on the 3D current image 36 according to the 3D reference image 31; The template area 40) performs pattern matching twice on the specified search object area 42, wherein the specified template area is generated based on the result of primary pattern matching based on one of the 3D reference image 31 or the 3D current image 36, and the specified The search target region 42 is generated based on the primary pattern matching result from the other of the 3D reference image 31 or the 3D current image 36 that is different from the generation basis of the predetermined template region (position and posture conversion region 40 ). Even when the number of tomographic images in the three-dimensional current image 36 is smaller than that in the three-dimensional reference image 31, high-precision two-stage pattern matching can be realized.

Embodiment 2

In the two-stage pattern matching of Embodiment 2, one pattern matching is performed from the three-dimensional reference image 31 to the three-dimensional current image 36, and then, based on the result of the one-pass pattern matching, a predetermined pattern is generated from the three-dimensional current image 36. The current image template region 44 of the template region used for the secondary pattern matching takes the pose transformation reference image region 47 after transforming the position and posture of the 3D reference image 31 as the retrieval object, and transforms the pose transformation reference image according to the current template region 44 Area 47 performs pattern matching twice. The 2nd pattern matching is the reverse of the 1st pattern matching.

10 is a diagram illustrating a primary pattern matching method according to Embodiment 2 of the present invention, and FIG. 11 is a diagram illustrating a relationship between a reference image template region and a slice image in the primary pattern matching method of FIG. 10 . In Embodiment 2, by pattern matching once, the comparison unit 16 performs searching including three axes of rotation to obtain the posture change amount.

The current image area 38 shown in FIG. 10 is represented as a cuboid including three slice images 37a, 37b, and 37c. The position and posture transformation template regions 40 a , 40 b , and 40 c serving as the reference image template regions in Embodiment 2 are regions subjected to position and posture transformation by the position and posture conversion unit 25 . However, the initial position and posture are in the default state, for example, the parameter of the rotation 3 axis is 0. The position-posture conversion template region 40a, which is the reference image template region after the position-posture conversion, moves in the slice image 37a in a raster-scanning manner along the scan path 39a. Similarly, the position and posture transformation template region 37b, which is the position and posture transformed reference image template region 40b, moves along the scanning path 39b in the slice image 37b in a raster scanning manner, and the position and posture transformation template region 40c is obtained. Along the scanning path 39c, it moves in the slice image 37c in a raster scanning manner. In order to simplify the drawing, the scanning paths 39b, 39c are schematically shown.

While transforming the position and posture, the correlation calculation between the slice images 37 a , 37 b , and 37 c of the 3D current image 36 and the position and posture conversion template area 40 is performed. For example, the correlation calculation is performed by changing each of the three rotation axes at a predetermined change amount or rate of change, and moving to the next scanning position to perform the correlation calculation. As shown in FIG. 11 , the primary comparison unit 16 performs image comparison between the cross section 41 of the position and posture transformation template region 40 and the slice image 37 constituting the current image region 38 . The cross section 41 of the position and posture conversion template region 40 is a plane in which the position and posture conversion template region 40 is cut along a plane parallel to the slice image 32 of the 3D reference image 31 as the initial position and posture, and is obtained from the 3D reference image 31. A plurality of slice images 32 are generated (section generating step). For example, the method described in Embodiment 1 can be used. That is, the data of the cross-section 41 can be extracted from the plurality of slice images 32 constituting the three-dimensional reference image 31 . In addition, the section 41 of the position and posture transformation template region 40 may also include data that has been completed so that the data density of the section 41 is the same as that of the 3D current image 36 .

Next, the primary comparison unit 16 generates the current image template area 44 used for the secondary matching. The primary comparison unit 16 obtains, for example, the cross-section 41 of the position-posture conversion template region 40 with the highest correlation value based on search results including the three axes of rotation in each of the slice images 37a, 37b, and 37c. The amount of posture change in the position and posture transformation template area 40 and the extraction area of the slice image 37 corresponding to the section 41 . The primary comparison unit 16 generates the current image template region 44 from the obtained extracted regions of each slice image so as to include the extracted region of the 3D current image with the highest correlation value. The current image template area 44 is a 2-dimensional image.

Next, as shown in FIG. 12 , the position and posture conversion unit 25 of the comparison processing unit 22 changes the overall posture of the three-dimensional reference image 31 by the above-mentioned posture change amount obtained when the current image template region 44 is generated, and generates The 3D pose transformation reference image 45 after the pose transformation is the pose transformation reference image area 47 . 12 is a diagram showing a three-dimensional reference image after posture conversion according to Embodiment 2 of the present invention. The slice images 46 a , 46 b , 46 c , 46 d , and 46 e are slice images obtained by changing the postures of the slice images 32 a , 32 b , 32 c , 32 d , and 32 e by the amount of posture change described above, respectively.

Next, as shown in FIG. 13 , the secondary comparison unit 17 matches the current image template region 44 to the pose transformation reference image region 47 which is the three-dimensional pose transformation reference image 45 after the pose transformation in a raster scanning manner along the scanning path 49, Therefore, only translational offset can be detected at high speed. FIG. 13 is a diagram illustrating a secondary pattern matching method according to Embodiment 2 of the present invention. The posture transformation reference image area 47 after the posture transformation is represented as a cuboid including five slice images 46 a , 46 b , 46 c , 46 d , and 46 e. The collation execution surface 48 is an image surface corresponding to a posture having the highest correlation value with the posture corresponding to the slice image 37 of the three-dimensional current image 36 by primary pattern matching, that is, a posture corresponding to the following A plane having an equivalent posture corresponding to the slice image 37 of the three-dimensional current image 36 in the posture conversion reference image area 47 . The secondary collation unit 17 generates a predetermined collation execution surface 48 from the pose transformation reference image region 47 and generates it from a plurality of slice images 46 of the three-dimensional pose transformation reference image 45 (comparison execution surface generation step). For example, the method described in Embodiment 1 can be used. That is, the data of the collation execution surface 48 can be extracted from a plurality of slice images constituting the three-dimensional pose transformation reference image 45 . In addition, the comparison execution surface 48 includes completion so that the data density of the comparison execution surface 48 is the same as the data density of the current image template area 44 .

The two-stage pattern matching method of Embodiment 2 is summarized. First, the collation processing unit 22 uses the position and posture conversion unit 25 to generate a position and posture transformation template region 40 from the three-dimensional reference image 31 (position and posture transformation template region generation step). The primary collation unit 16 of the collation processing unit 22 performs pattern matching on the three-dimensional current image 36 with the position and orientation transformation template region 40 once (primary pattern matching step). Each time the position and posture of the position and posture conversion template region 40 is changed (every time the position and posture conversion step is executed), the cross-section of the position and posture conversion template region 40 is generated for each slice image 37 constituting the current image region 38 by one pattern matching 41 (section generation step), image comparison is performed between the section 41 of the position and posture conversion template area 40 and the slice image 37 constituting the current image area 38 .

Each time the position and posture of the position and posture conversion template area 40 is changed, the primary comparison unit 16 calculates a correlation value between the current image area 38 and the position and posture conversion template area 40 (correlation value calculation step). In addition, each time the position and posture transformation template region 40 is scanned, the primary comparison unit 16 calculates the correlation value between the current image region 38 and the position and posture transformation template region 40, and uses pattern matching once to generate the current image template region 44, so that It includes the extraction region of the position and posture transformation template region 40 with the highest correlation value between the current image region 38 and the position and posture transformation template region 40 (current image template region generating step).

Next, the comparison processing unit 22 uses the position and posture conversion unit 25 to change the posture of the entire three-dimensional reference image 31 by the above-mentioned posture change amount obtained when the current image template region 44 is generated, and generates a three-dimensional posture after the posture transformation. The transformation reference image 45, that is, the pose transformation reference image area 47 is generated (posture transformation reference image area generating step). The secondary matching unit 17 performs secondary pattern matching on the current image template area 44 and the pose transformation reference image area 47 (secondary pattern matching step). In the secondary pattern matching, the comparison execution surface 48 is generated through the comparison execution surface generation step, and image comparison is performed on the comparison execution surface 48 generated in the comparison execution surface generation step and the current image template area 44 . When performing this image comparison, the current image template area 44 is translated without being rotated, and the correlation value between the comparison execution surface 48 and the current image template area 44 is calculated (correlation value calculation step).

In the secondary pattern matching, the secondary comparison unit 17 of the comparison processing unit 22 compares the position and posture relationship (position and posture information) between the 3D posture transformation reference image 45 and the current image template area 44 with the highest correlation value among the calculated correlation values. ) is selected as the best solution (best solution selection step). Thus, pattern matching is realized by two-stage matching so that the two types of 3D images, the 3D reference image 31 and the 3D current image 36 , are the most consistent. The two-stage pattern matching method of Embodiment 2 includes: a position and posture transformation template region generation step; a pattern matching step once; a posture transformation reference image region generation step; and a pattern matching step twice. A pattern matching step includes: a section generation step; a correlation value calculation step; a position and posture transformation step; and a current image template region generation step. The two pattern matching steps include: the step of generating the comparison execution surface; the step of calculating the correlation value; and the step of selecting the best solution.

After the pattern matching is completed, the collation processing unit 22 calculates the position and posture of the high correlation value region in the 3D posture transformation reference image 45 whose correlation value is the highest among the calculated correlation values. Body position correction amount (translation amount, rotation amount) when the current image 36 is compared (body position correction amount calculation step). The collation result display unit 23 displays, on the monitor screen of the computer 14 , the body posture correction amount, or an image obtained by superimposing the 3D current image moved by the body posture correction amount on the 3D reference image. The comparison result output unit 24 outputs the body posture correction amount (translation amount, rotation amount) when the comparison processing unit 22 compares the 3D reference image 31 and the 3D current image 36 (body posture correction amount output step). The treatment table control parameter calculation part 26 converts the output value of the comparison result output part 24 (3-axis translation [ΔX, ΔY, ΔZ], 3-axis rotation [ΔA, ΔB, ΔC], a total of 6 degrees of freedom) into the corresponding treatment table 8, the parameters controlled by each axis are the calculated parameters (the calculation step of the control parameters of the treatment table). The treatment table 8 drives the driving device of each axis of the treatment table 8 based on the treatment table control parameters calculated by the treatment table control parameter calculation unit 26 (treatment table driving step).

The image collating device 29 according to Embodiment 2 performs one-time pattern matching as image collating including three-axis rotation on the three-dimensional current image 36 based on the position and posture transformation template region 40 of the three-dimensional reference image 31, and then, based on one As a result of secondary pattern matching, the current image template region 44 serving as a template region for secondary pattern matching is generated from the 3D current image 36. Therefore, even if the number of tomographic images (the number of slice images) of the 3D current image 36 is greater than that of the 3D current image 36 Even when the number of quasi-images 31 is small, highly accurate two-step pattern matching can be realized.

The image collating device 29 according to Embodiment 2 generates a 3D pose transformation reference image 45 as a pose transformed 3D reference image from the 3D reference image 31 , that is, generates a pose transformation reference image region 47 , so that 2 dimensional current image template region 44, and utilizes translational movement without rotational movement to perform direct pattern matching on the pose transformation reference image region 47. In the two-pass pattern matching, since only the correlation value is calculated for each translational movement, compared with the case of calculating the correlation value for every rotational movement and translational movement, the speed-up of the two-pass pattern matching is realized.

Embodiment 3

The difference between Embodiment 3 and Embodiments 1 and 2 is that the human body database (atlas model: atlas model) is used to generate the reference image template area 33 for primary pattern matching in Embodiment 1, or the position and posture in Embodiment 2. The transformation template region 40 is based on the reference image template region 33 . FIG. 14 is a diagram showing configurations of an image collating device and a patient positioning device according to Embodiment 3 of the present invention. The image collating device 29 according to Embodiment 3 differs from the image collating devices 29 according to Embodiments 1 and 2 in that it includes: a human body database input unit 50 ; and an average template area generating unit 51 . The patient positioning device 30 according to Embodiment 3 includes an image comparison device 29 and a treatment table control parameter calculation unit 26 .

The human body database input unit 50 acquires a human body database (atlas model) from a storage device such as a database device. The average template region generation unit 51 cuts out the average template region 54 from the organ parts of the human body database corresponding to the affected parts 5 and 11 of the patients 4 and 10 . The reference template region generation unit 18 of the comparison processing unit 22 automatically generates the reference image template region 33 by pattern-matching the average template region 54 to the three-dimensional reference image 31 (reference image template region generation step).

Using the reference image template region 33 described above, the two-stage pattern matching of Embodiment 1 or the two-stage pattern matching of Embodiment 2 is performed. In this way, two-step pattern matching can be realized without preparing information indicating the affected part (the shape of the affected part, etc.) on the 3D reference image in advance.

In addition, it is considered that the average template region generation unit 51 cuts out a two-dimensional average template region from the organ parts of the human body database corresponding to the affected parts 5 and 11 of the patients 4 and 10 . In the case of the two-dimensional average template region 54 , a plurality of two-dimensional average template regions are cut out, and the plurality of two-dimensional average template regions are collected and output to the collation processing unit 22 . The reference template region generation unit 18 of the comparison processing unit 22 automatically generates the reference image template region 33 by pattern-matching the plurality of two-dimensional average template regions to the three-dimensional reference image 31 .

Label description

16...First-time comparison part; 17...Second-time comparison part; 18...Reference template area generation part; 21...3-dimensional image input part; 22...Contrast processing part; 25...Position and posture transformation part; 29...image comparison device; 30...patient positioning device; 31...3-dimensional reference image; 33...reference image template area; 36...3-dimensional current image; 40, 40a, 40b, 40c...position and posture transformation template area; 41 ...section; 42...1 time extraction of the current image area; 44...current image template area; 45...3-dimensional pose transformation reference image; 48...contrast execution surface; 50...human body database input unit; 51...average template area generation unit.

Claims (16)

1. An image comparison device, characterized in that, comprising: a 3D image input unit that reads a 3D reference image captured during treatment planning of radiation therapy and a 3D current image captured during treatment; and a comparison processing unit that compares the 3D reference image with the 3D current image, and calculates a body posture correction amount so that the position and posture of the affected part in the 3D current image are consistent with the 3D reference image. The position and posture of the affected part in the image are consistent, The control treatment part has: a primary matching unit that performs pattern matching once on the 3D current image from the 3D reference image; and a secondary comparison unit, the secondary comparison unit performs pattern matching on the predetermined search object region twice according to the predetermined template region, wherein the predetermined template region is based on the 3D reference image or the 3D current image One is generated based on the result of the primary pattern matching, and the predetermined search object region is based on the 3D reference image or the 3D current image that is different from the generation basis of the predetermined template region. The other is generated based on the result of the first pattern matching. 2. The image comparison device according to claim 1, wherein: The comparison processing unit includes a reference template region generation unit for generating a reference image template region of a three-dimensional region from the three-dimensional reference image based on the affected part information prepared in the three-dimensional reference image. 3. The image comparison device according to claim 1, characterized in that, comprising: a human body database input unit that acquires a human body database from the database device; and an average template area generation unit that generates an average template area from an organ portion corresponding to an affected portion of a patient in the human body database, The comparison processing unit includes a reference template region generating unit that performs pattern matching on the 3D reference image based on the average template region, and generates Image generation 3D regions of reference image template regions. 4. The image comparison device according to claim 2 or 3, characterized in that, During the primary pattern matching, the primary matching unit performs pattern matching on the 3D current image based on the reference image template area. 5. The image comparison device according to claim 4, characterized in that, The primary matching unit generates a primary extraction current image region as the search target region from the three-dimensional current image so as to include a region having the highest correlation value with the reference image template region. 6. The image comparison device according to claim 5, characterized in that, The collation processing unit includes a position and posture conversion unit that converts the position and posture of the 3D image, The position and posture conversion unit generates a position and posture conversion template region obtained by converting the position and posture of the reference image template region into a predetermined position and posture, During the secondary pattern matching, the secondary matching unit performs pattern matching on the primary extracted current image area based on the position and posture template area that is the predetermined template area. 7. The image comparison device according to claim 6, characterized in that, The secondary matching unit generates a cross-section of the position-posture transformation template region, and performs pattern matching between the primary extracted current image region and the cross-section. 8. The image comparison device according to claim 2 or 3, characterized in that, The collation processing unit includes a position and posture conversion unit that converts the position and posture of the 3D image, The position and posture conversion unit generates a position and posture conversion template region obtained by converting the position and posture of the reference image template region into a predetermined position and posture, During the primary pattern matching, the primary comparison unit performs pattern matching on the 3D current image according to the position and posture conversion template area. 9. The image comparison device according to claim 8, characterized in that, The primary matching unit generates a cross-section of the position and posture transformation template region, performs pattern matching between the 3D current image and the cross-section, and performs the following operation: A high-correlation profile of the highest profile; calculating the pose variation in the position-posture transformation template region; and extracting an extraction region corresponding to the high-correlation profile in the 3D current image. 10. The image comparison device according to claim 9, characterized in that, The primary comparison unit generates a current image template area as the predetermined template area so as to include the extraction area extracted at the time of the primary pattern matching, The position and posture transformation unit generates a 3-dimensional posture transformation reference image region of the extraction region in which the position and posture transformation of the 3-dimensional reference image is converted to correspond to the current image template region. Formed by the size of the posture change, During the secondary pattern matching, the secondary comparison unit performs pattern matching on the 3D pose transformation reference image region as the search target region based on the current image template region. 11. The image comparison device according to claim 10, characterized in that, The secondary matching unit generates a matching execution surface that is a cross section of the position and posture conversion template area, and performs pattern matching between the current image template area and the matching execution surface. 12. A patient positioning device, comprising: The image comparison device according to any one of claims 1 to 3, 5 to 7, 9 to 11; and A treatment table control parameter calculation unit that calculates parameters for controlling each axis of the treatment table based on the body posture correction amount calculated by the image collating device. 13. An image comparison method, the image comparison method compares a 3-dimensional reference image taken during radiation therapy treatment planning with a 3-dimensional current image taken during treatment, the image comparison method is characterized in that it includes: 1 pattern matching step, the 1 pattern matching step performs 1 pattern matching on the 3D current image according to the 3D reference image; and a second pattern matching step, the second pattern matching step performs pattern matching on the prescribed search target area twice according to the prescribed template area, wherein the prescribed template area is based on the 3D reference image or the 3D current image One of them is generated based on the result of the primary pattern matching, and the predetermined search object region is based on the 3D reference image or the 3D current image that is different from the generation basis of the predetermined template region The other one is generated based on the result of the first pattern matching. 14. The image comparison method according to claim 13, characterized in that, comprising a reference image template region generating step, which generates a reference image template region of a 3-dimensional region from said 3-dimensional reference image, The one-time pattern matching step performs one-time pattern matching on the 3D current image according to the reference image template area. 15. A patient positioning device, comprising: The image comparison device as claimed in claim 4; and A treatment table control parameter calculation unit that calculates parameters for controlling each axis of the treatment table based on the body posture correction amount calculated by the image collating device. 16. A patient positioning device, comprising: The image comparison device according to claim 8; and A treatment table control parameter calculation unit that calculates parameters for controlling each axis of the treatment table based on the body posture correction amount calculated by the image collating device.