Image processing apparatus and image processing method

Abstract

The present invention provides an image processing device capable of improving the robustness of patient positioning. [Solution] The image estimation unit 24 generates time-series data of an estimated 2D perspective image, which is an estimated 2D perspective image based on the time-series data of the 3D volume image. The curve comparison unit 25 calculates a comparison result by comparing a plurality of measurement curves showing the time change of the pixel value of each pixel in the 2D perspective image with a plurality of estimation curves showing the time change of the pixel value of each pixel in the estimated 2D perspective image, based on the time-series data of the 2D perspective image and the time-series data of the estimated 2D perspective image. The optimization unit 26 generates time-series data of an optimized 3D volume image, which is an optimized 3D volume image, which is an optimized 3D volume image, with each of the 3D volume images adjusted according to the comparison result.

Description

Technical Field

[0001] The present disclosure relates to an image processing apparatus and an image processing method.

Background Art

[0002] In order to achieve high-precision radiation therapy, it is important to accurately position the patient. The patient positioning is usually performed by matching an X-ray image that captures the periphery of the bone region obtained at a specific moment on the treatment day when radiation therapy is performed, and a DRR (Digitally Reconstructed Radiograph) generated from a CT (Computed Tomography) image obtained at the treatment planning stage for creating the treatment plan of radiation therapy. Since the DRR image is an image generated from the CT image, although it is different from the actually detected X-ray image, there is a clear contrast between the bone region and other regions, so the X-ray image and the DRR image can be matched (see Patent Document 1).

[0003] However, the above technique assumes that the relative positions of the tumor and the bone do not change throughout the entire treatment process of radiation therapy. However, since the DRR image is generated from the CT image obtained at the treatment planning stage about several days before the treatment day, there is a possibility that the relative positions of the tumor and the bone are different between the X-ray image and the DRR image. In addition, the relative positions of the tumor and the bone change over time according to the patient's respiratory movement and the like. Therefore, the above technique has a problem that the irradiation area where the radiation is actually irradiated deviates from the ideal irradiation area.

[0004] In contrast, Patent Document 2 discloses a technique that calculates the temporal change in the patient's bone position from the treatment planning stage based on two-dimensional fluoroscopic images acquired during the treatment planning stage and two-dimensional fluoroscopic images acquired during treatment, and irradiates the patient with radiation only when the change in water equivalent thickness falls below an acceptable value based on that change. In this case, radiation can be irradiated only when the relative positional displacement between the tumor and the bone is small, thus reducing the displacement of the irradiation area. [Prior art documents] [Patent Documents]

[0005] [Patent Document 1] Japanese Patent Publication No. 2010-246733 [Patent Document 2] Japanese Patent Publication No. 2016-144573 [Overview of the Initiative] [Problems that the invention aims to solve]

[0006] However, in the technology described in Patent Document 2, the temporal changes of each part captured in the X-ray image are strongly affected by noise, and areas with clear contrast, such as bones, become blurred, making it difficult to accurately detect the amount of displacement of bones. In other words, the technology described in Patent Document 2 has a problem with robustness.

[0007] The purpose of this disclosure is to provide an image processing apparatus and an image processing method capable of improving the robustness of patient positioning. [Means for solving the problem]

[0008] An image processing apparatus according to one aspect of the present disclosure includes: a volume acquisition unit that acquires time-series data of a three-dimensional volume image showing the anatomical structure of a subject; an image acquisition unit that acquires time-series data of a two-dimensional perspective image showing the anatomical structure; an estimation unit that generates time-series data of an estimated two-dimensional perspective image estimated from the time-series data of the three-dimensional volume image; a comparison unit that calculates a comparison result by comparing a plurality of measurement curves showing the time change of the pixel value of each pixel in the two-dimensional perspective image with a plurality of estimation curves showing the time change of the pixel value of each pixel in the estimated two-dimensional perspective image, based on the time-series data of the two-dimensional perspective image and the time-series data of the estimated two-dimensional perspective image; and a generation unit that generates time-series data of an optimized three-dimensional volume image adjusted from the three-dimensional volume image based on the comparison result. [Effects of the Invention]

[0009] According to the present invention, it becomes possible to improve the robustness of patient positioning. [Brief explanation of the drawing]

[0010] [Figure 1] This figure shows the configuration of an image processing system according to one embodiment of the present disclosure. [Figure 2] This is a front view showing an example of a 3D volume imaging device. [Figure 3] This is a side view showing an example of a 3D volume imaging device. [Figure 4] This is a front view showing an example of an image acquisition device. [Figure 5] This is a side view showing an example of an image acquisition device. [Figure 6] This is a flowchart illustrating an example of the processing steps of an image processing system. [Figure 7] This is a diagram illustrating an example of optimization processing. [Figure 8] This is a diagram illustrating other examples of optimization processes. [Modes for carrying out the invention]

[0011] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0012] Figure 1 is a diagram showing the configuration of an image processing system according to one embodiment of the present disclosure. The image processing system 1 shown in Figure 1 is a system for performing image processing on images used in radiotherapy, which treats tumors by irradiating the tumor, which is the affected area of ​​a patient, with radiation, and comprises a 3D volume imaging device 11, an image imaging device 12, and an image processing device 13.

[0013] The 3D volume imaging device 11 is a first imaging device that images a patient in order to acquire time-series data of 3D volume images showing the anatomical structures of the patient, who is the subject of the imaging. The 3D volume imaging device 11 is, for example, a CT scanner, a cone-beam computed tomography (CBCT) scanner, a fan-beam computed tomography (CBCT) scanner, or an MRI (Magnetic Resonance Imaging) scanner. The 3D volume imaging device 11 images the patient in the treatment planning stage, which is the first stage in creating a treatment plan for radiation therapy for the patient. The 3D volume images are treatment planning images used to create the treatment plan. The anatomical structures of the patient include, for example, tumors, bones, and organs, and their position and other characteristics change due to movements such as respiratory movements.

[0014] The image acquisition device 12 is a second imaging device that images the patient in order to acquire time-series data of two-dimensional fluoroscopic images showing the patient's anatomical structure. In this embodiment, the image acquisition device 12 images the patient while changing the imaging angle. The image acquisition device 12 is, for example, a cone-beam CT scanner and images the patient in the pre-treatment stage, which is a second stage in which pre-processing such as positioning the patient during radiation therapy based on the treatment plan is performed. The pre-treatment stage is usually performed several days after the treatment planning stage, immediately before the start of radiation therapy on the treatment day. The two-dimensional fluoroscopic image is a positioning image used for positioning the patient in pre-processing.

[0015] FIG. 2 is a front view showing an example of the three-dimensional volume imaging device 11, and FIG. 3 is a side view showing an example of the three-dimensional volume imaging device 11.

[0016] The three-dimensional volume imaging device 11 shown in FIGS. 2 and 3 is a fan beam CT device that images a patient 51 placed on a treatment table 50a, and includes an X-ray source 111a and an X-ray detector 112a.

[0017] The treatment table 50a moves in a predetermined direction (direction X indicated by the arrow in FIG. 3). The X-ray source 111a is installed so as to be rotatable around the patient 51 with the X-axis as the rotation axis, and a plurality of X-ray detectors 112a are installed around the patient 51, or are installed so as to be rotatable around the patient 51 in the same manner as the X-ray source 111a.

[0018] The X-ray source 111a irradiates the patient 51 placed on the treatment table 50a with X-rays in a fan beam form as radiation for imaging. The X-ray detector 112a detects the X-rays irradiated from the X-ray source 111a through the patient 51.

[0019] In the present embodiment, the affected part to be treated (the irradiation target irradiated with a treatment beam that is radiation for treatment) is a tumor 31 in the patient 51. The region irradiated with X-rays by the X-ray source 111a is a three-dimensional region of interest 33 including the tumor 31 and a rib tissue 32 that is a reference tissue, which is a tissue that moves differently from the tumor 31. Thereby, the three-dimensional volume imaging device 11 can acquire a three-dimensional volume image showing the tumor 31 and the rib tissue 32 as the anatomical structure of the patient who is the subject. Note that the reference tissue is not limited to the rib tissue, and may be other bone tissues or tissues other than bone tissues as long as it is a tissue that moves differently from the tumor 31.

[0020] Further, the three-dimensional volume imaging device 11 includes a surrogate motion acquisition device 15 that acquires the motion of a surrogate (proxy target) 151 disposed on the body surface of the patient 51. The surrogate motion acquisition device 15 has a motion tracking camera 152 that tracks the motion of the surrogate 151.

[0021] Figure 4 is a front view showing an example of the image acquisition device 12, and Figure 5 is a side view showing an example of the image acquisition device 12.

[0022] The image acquisition device 12 shown in Figures 4 and 5 is a cone-beam CT scanner that images a patient 51 placed on a treatment table 50b, and includes an X-ray source 111b and an X-ray detector 112b.

[0023] The X-ray source 111b emits X-rays in a cone beam format towards the patient 51 placed on the treatment table 50b as radiographic radiation. The X-ray detector 112b detects the X-rays emitted from the X-ray source 111b via the patient 51. The area irradiated with X-rays by the X-ray source 111b is a three-dimensional region of interest 33 that includes the tumor 31 and the rib tissue 32, similar to the area irradiated with X-rays by the X-ray source 111a described above.

[0024] The X-ray source 111b is mounted so as to be rotatable around the patient 51, and the X-ray detectors 112b are mounted multiple times around the patient 51, or are mounted so as to be rotatable around the patient 51, similar to the X-ray source 111b.

[0025] Returning to the explanation of Figure 1, the image processing device 13 is comprised of a computer system, for example, a processor (computer) and memory (neither of which are shown). In this case, each component and function of the image processing device 13 described below is realized, for example, by the processor reading a program and executing the program it has read. The program can be recorded on a computer-readable recording medium, such as external memory. The image processing device 13 may also be connected to an input device that receives various information from a user of the image processing device 13, an output device that outputs various information to the user, and a network interface that sends and receives various information via a communication network such as the Internet.

[0026] The image processing device 13 has the following functional configurations: a volume acquisition unit 21, an image acquisition unit 22, a proxy operation acquisition unit 23, an image estimation unit 24, a curve comparison unit 25, an optimization unit 26, and a positioning unit 27.

[0027] The volume acquisition unit 21 acquires time-series data of a three-dimensional volume image depicting the anatomical structure of the patient 51 based on the imaging data from the three-dimensional volume imaging device 11. In this embodiment, the time-series data of the three-dimensional volume image is acquired over one or more respiratory cycles.

[0028] The image acquisition unit 22 acquires time-series data of two-dimensional fluoroscopic images depicting the anatomical structure of the patient 51 based on the imaging data from the image acquisition device 12. In this embodiment, the time-series data of the two-dimensional fluoroscopic images is acquired over one or more respiratory cycles.

[0029] The proxy operation acquisition unit 23 acquires proxy operation data indicating the proxy operation of the surrogate 151, based on the imaging data from the surrogate operation acquisition device 15 mounted on the 3D volume imaging device 11.

[0030] The image estimation unit 24 is an estimation unit that generates time-series data of an estimated 2D fluoroscopic image, which is an estimated 2D fluoroscopic image depicting the anatomical structure of patient 51, based on the time-series data of the 3D volume image acquired by the volume acquisition unit 21 and the proxy action data acquired by the proxy action acquisition unit 23.

[0031] The curve comparison unit 25 calculates measurement curves and estimated curves that show the time change of the pixel value of each pixel in the two-dimensional perspective image and the estimated two-dimensional perspective image, respectively, based on the time-series data of the two-dimensional perspective image acquired by the image acquisition unit 22 and the time-series data of the estimated two-dimensional perspective image generated by the image estimation unit 24, and compares the measurement curve and the estimated curve corresponding to each pixel.

[0032] The optimization unit 26 is a generation unit that generates optimized 3D volume time series data by adjusting each 3D volume image included in the time series data of the 3D volume image based on the comparison results from the curve comparison unit 25.

[0033] The positioning unit 27 adjusts the position of the treatment table 50b based on the time-series data of the optimized 3D volume generated by the optimization unit 26, thereby positioning the patient 51 placed on the treatment table 50b.

[0034] In this embodiment, the image processing device 13 also had the function of a positioning device (positioning unit 27) for positioning the patient 51, but a positioning device may be provided separately from the image processing device 13.

[0035] Figure 6 is a flowchart illustrating an example of the processing of the image processing system 1. The processing of the image processing system 1 includes a treatment planning stage 301 in which a treatment plan for radiotherapy is created, and a pretreatment stage 302 performed immediately before radiotherapy on the treatment day.

[0036] In the treatment planning stage 301, first, the volume acquisition unit 21 of the image processing device 13 acquires time-series data of a 3D volume image showing the temporal changes in the anatomical structure of the patient 51, based on the imaging data from the 3D volume imaging device 11 (step S101). The volume acquisition unit 21 may acquire the time-series data of the 3D volume image from imaging data acquired at multiple time points, or it may acquire it from a still image 3D volume image based on imaging data acquired at a specific time point using a machine learning model or the like.

[0037] Next, the surrogate motion acquisition unit 23 acquires surrogate motion data, which represents the surrogate motion of the surrogate 151 installed on the patient 51, based on the imaging data from the surrogate motion acquisition device 15 mounted on the 3D volume imaging device 11. Then, the surrogate motion acquisition unit 23 generates a motion model that estimates the 3D volume image from the surrogate motion data, based on the time-series data of the 3D volume image and the surrogate motion data (step S102).

[0038] The motion model is generated using statistical methods such as principal component analysis, which generalizes the motion of the patient's tissues into eigenvectors, and in this embodiment, it is represented by the following equation 1.

number

[0039] The image estimation unit 24 generates time-series data of estimated 2D fluoroscopic images, which are estimated 2D fluoroscopic images depicting the anatomical structure of patient 51, based on the time-series of 3D volume images acquired by the volume acquisition unit 21 (step S103).

[0040] In this embodiment, the image estimation unit 24 first inputs surrogate motion data into a motion model generated based on the time-series data of the 3D volume image to obtain the predicted time-series data of the 3D volume image as the time-series data of the predicted 3D volume image. Subsequently, the image estimation unit 24 converts each of the predicted 3D volume images included in the time-series data into a DRR image, which is a projected 2D image obtained by projecting it onto a virtual plane corresponding to the imaging angle of the 2D fluoroscopic image, thereby generating the time-series data of the DRR image as the time-series data of the predicted 2D fluoroscopic image. Note that the time-series data of the predicted 3D volume image is smoothed out compared to the time-series data of the original 3D volume image because variations caused by changes in anatomical structures due to respiratory movements and the like are smoothed out, thus enabling the generation of time-series data of the predicted 2D fluoroscopic image with higher accuracy.

[0041] The above method for generating time-series data of estimated 2D perspective images is merely an example and is not limited to this. For example, the image estimation unit 24 may convert the time-series of the 3D volume image into a DRR image projected onto a virtual plane corresponding to the imaging angle of the 2D perspective image, without using a motion model. In this case, the processing in step S102 may be skipped.

[0042] Next, in the pretreatment stage 302, the image acquisition unit 22 first acquires time-series data of a two-dimensional fluoroscopic image based on the imaging data from the image acquisition device 12 (step S201).

[0043] Next, the curve comparison unit 25 calculates a measurement curve and an estimated curve showing the time change of the pixel value of each pixel in the two-dimensional perspective image and the estimated two-dimensional perspective image, respectively, based on the time-series data of the two-dimensional perspective image acquired by the image acquisition unit 22 and the time-series data of the estimated two-dimensional perspective image generated by the image estimation unit 24. For each pixel, the curve comparison unit 25 calculates a comparison result by comparing the measurement curve and the estimated curve corresponding to that pixel (step S202).

[0044] In this embodiment, the curve comparison unit 25 calculates a measurement curve and an estimated curve for each pixel of the two-dimensional region of interest in the two-dimensional perspective image and the estimated two-dimensional perspective image, respectively. The two-dimensional region of interest is the region that reflects the three-dimensional region of interest 33 described above in the two-dimensional perspective image and the estimated two-dimensional perspective image.

[0045] The optimization unit 26 generates time-series data of the optimized 3D volume by adjusting the 3D volume image based on the comparison results calculated by the curve comparison unit 25 (step S203).

[0046] The positioning unit 27 adjusts the position of the treatment table 50b based on the time-series data of the optimized 3D volume generated by the optimization unit 26, thereby positioning the patient 51 placed on the treatment table 50b (step S204).

[0047] Figure 7 illustrates an example of the optimization process that generates time-series data of the optimized 3D volume in steps S202 and S203, showing the measurement curve 40a and the estimation curve 40b.

[0048] The X-ray detector 112b of the image acquisition device 12 has multiple photodetectors (not shown) that detect the intensity of X-rays irradiated from the X-ray source 111a and passed through the patient 51. The two-dimensional fluoroscopic image has multiple pixels, each having a pixel value corresponding to the intensity of the X-rays detected by each photodetector of the X-ray detector 112b. Therefore, the pixel value of each pixel differs depending on the anatomical structure of the patient 51 through which the X-rays incident on the photodetector corresponding to that pixel have passed. Furthermore, the position of anatomical structures within the patient 51 changes over time due to respiratory movements, etc. Therefore, the tissue of the patient 51 through which the X-rays incident on the photodetector have passed also changes over time, and as a result, the pixel value of each pixel in the two-dimensional fluoroscopic image also changes over time. Similarly, the pixel value of each pixel in the estimated two-dimensional fluoroscopic image also changes over time.

[0049] The measurement curve 40a shows the time change in the pixel value of pixel 30 in the two-dimensional fluoroscopic image, and the estimation curve 40b shows the time change in the pixel value of pixel 30 in the estimated two-dimensional fluoroscopic image. Pixel 30 is a pixel within the two-dimensional region of interest 33b included in the two-dimensional fluoroscopic image and the estimated two-dimensional fluoroscopic image. As described above, the two-dimensional region of interest 33b includes the tumor 31 and the reference tissue, rib tissue 32, which is tissue that moves differently from the tumor 31.

[0050] In the example shown in Figure 7, as time progresses from t0 to t1 to t2 to t3 to t4, the tumor 31 moves to the lower left and the rib tissue 32 moves to the upper left. Furthermore, the tumor 31 has a negative contribution to the intensity of the X-rays detected by the photodetector (i.e., the pixel value), while the rib tissue 32 has a positive contribution to the intensity of the X-rays detected by the photodetector.

[0051] At time t0, neither the tumor 31 nor the rib tissue 32 is present in pixel 30. Subsequently, as the rib tissue 32 passes through pixel 30, the pixel value (X-ray intensity) of pixel 30 increases, peaking at time t1 just before the tumor 31 enters pixel 30. Then, as the tumor 31 passes through pixel 30, the pixel value of pixel 30 decreases, and at time t2, as the rib tissue 32 begins to leave pixel 30, the pixel value of pixel 30 decreases further. Furthermore, at time t3, as the rib tissue 32 has completely left pixel 30 and the tumor 31 also begins to leave pixel 30, the pixel value of pixel 30 increases. Finally, at time t4, when neither the tumor 31 nor the rib tissue 32 is left in pixel 30, the pixel value of pixel 30 returns to its original state (the state at time t0).

[0052] The curve comparison unit 25 compares the measured curve 40a and the estimated curve 40b for each pixel in the two-dimensional region of interest 33b and calculates the similarity between them as a comparison result. Based on the comparison results from the curve comparison unit 25, the optimization unit 26 adjusts the time series data of the three-dimensional volume image so as to maximize the similarity between the measured curve 40a and the estimated curve 40b, and generates the time series data of the optimized three-dimensional volume image.

[0053] For example, the curve comparison unit 25 adjusts the time series data of the predicted 3D volume image predicted by the motion model by adjusting the parameters of the motion model, thereby generating time series data of the optimized 3D volume image. Since the predicted 3D volume image is generated from the motion model which is based on the original 3D volume image, the predicted 3D volume image is an image obtained by adjusting the original 3D volume image.

[0054] The optimization method used by the optimization unit 26 to generate time-series data of the optimized 3D volume image is not limited to the example described above. For example, the optimization method may also involve adjusting the positions of the tumor 31 and rib tissue 32 within the 3D volume image.

[0055] Figure 8 illustrates another example of the optimization process, showing the measurement curve 40a and the tumor intensity component 40c and rib intensity component 40d of the estimation curve 40b. The tumor intensity component 40c represents the contribution of the tumor 31 to the pixel value, and the rib intensity component 40d represents the contribution of the rib tissue 32 to the pixel value. The tumor intensity component 40c and the rib intensity component 40d can be generated individually by the image estimation unit 24.

[0056] The optimization unit 26 adjusts the positions of the tumor 31 and rib tissue 32 by changing the shapes of the tumor strength component 40c and the rib strength component 40d, respectively, and generates time-series data of an optimized 3D volume image so as to maximize the similarity between the measurement curve 40a and the estimated curve 40b.

[0057] As described above, according to this embodiment, the image estimation unit 24 generates time-series data of an estimated 2D fluoroscopic image, which is an estimated 2D fluoroscopic image based on the time-series data of the 3D volume image. The curve comparison unit 25 calculates a comparison result by comparing a plurality of measurement curves showing the time change of the pixel value of each pixel in the 2D fluoroscopic image with a plurality of estimation curves showing the time change of the pixel value of each pixel in the estimated 2D fluoroscopic image, based on the time-series data of the 2D fluoroscopic image and the time-series data of the estimated 2D fluoroscopic image. The optimization unit 26 generates time-series data of an optimized 3D volume image, which is an optimized 3D volume image, adjusted according to the comparison result. Therefore, even if areas with clear contrast, such as bone, are blurred due to the effect of noise, it is possible to generate appropriate time-series data of a 3D volume image, thereby improving the robustness of patient positioning.

[0058] Furthermore, in this embodiment, the image estimation unit 24 generates a motion model that obtains a predicted 3D volume image by predicting the 3D volume image from the surrogate motion, based on the time-series data of the 3D volume image and the surrogate motion data. Using the motion model, it obtains the predicted 3D volume image and generates time-series data of the estimated 2D perspective image based on the predicted 3D volume image. In this case, it is possible to generate time-series data of the estimated 2D perspective image with high accuracy.

[0059] Furthermore, in this embodiment, the image estimation unit 24 adjusts the parameters of the motion model to generate time-series data of the adjusted 3D volume image as time-series data of the optimized 3D volume image. In this case, it becomes possible to generate time-series data of the estimated 2D perspective image with high accuracy.

[0060] Furthermore, in this embodiment, the measurement curve and the estimation curve, respectively, represent the time evolution of the pixel values ​​of each pixel included in the pixel region that captures multiple structures moving in different directions, in the two-dimensional perspective image and the estimated two-dimensional perspective image, respectively. In this case, it is possible to generate more appropriate time-series data for optimized three-dimensional volume images.

[0061] Furthermore, in this embodiment, the optimization unit 26 generates time-series data of the optimized 3D volume image so as to maximize the similarity between the measurement curve and the estimation curve for each pixel. In this case, more appropriate time-series data of the optimized 3D volume image can be generated.

[0062] The embodiments of the Disclosure described above are illustrative for illustrative purposes and are not intended to limit the scope of the Disclosure to those embodiments only. Those skilled in the art can implement the Disclosure in various other forms without departing from the scope of the Disclosure. [Explanation of Symbols]

[0063] 1: Image processing system 11: 3D volume imaging device 12: Image imaging device 13: Image processing device 15: Surrogate motion acquisition device 21: Volume acquisition unit 22: Image acquisition unit 23: Surrogate motion acquisition unit 24: Image estimation unit 25: Curve comparison unit 26: Optimization unit 27: Positioning unit

Claims

1. A volume acquisition unit that acquires time-series data of a three-dimensional volume image showing the anatomical structure of the subject, An image acquisition unit that acquires time-series data of a two-dimensional fluoroscopic image showing the aforementioned anatomical structure, An estimation unit generates time-series data of an estimated two-dimensional perspective image, which is obtained by estimating the two-dimensional perspective image based on the time-series data of the three-dimensional volume image. A comparison unit calculates a comparison result by comparing a plurality of measurement curves showing the time change of the pixel value of each pixel in the two-dimensional perspective image with a plurality of estimation curves showing the time change of the pixel value of each pixel in the estimated two-dimensional perspective image, based on the time series data of the two-dimensional perspective image and the time series data of the estimated two-dimensional perspective image. An image processing apparatus having a generation unit that generates time-series data of an optimized three-dimensional volume image obtained by adjusting the three-dimensional volume image based on the comparison results.

2. The three-dimensional volume image shows the anatomical structure in the first stage of creating a treatment plan for radiotherapy for the subject. The image processing apparatus according to claim 1, wherein the two-dimensional fluoroscopic image shows the anatomical structure in a second step of positioning the subject during radiotherapy, which is performed after the first step.

3. The system further includes a proxy action acquisition unit that acquires proxy action data indicating the operation of a surrogate placed on the body surface of the subject in the first step, The image processing apparatus according to claim 2, wherein the estimation unit generates a motion model that obtains a predicted three-dimensional volume image by predicting the three-dimensional volume image from the proxy motion data, based on the time-series data of the three-dimensional volume image and the proxy motion data, obtains the predicted three-dimensional volume image using the motion model, and generates time-series data of the estimated two-dimensional perspective image based on the predicted three-dimensional volume image.

4. The image processing apparatus according to claim 3, wherein the generation unit generates time-series data of the adjusted 3D volume image as time-series data of the optimized 3D volume image by adjusting the parameters of the motion model based on the comparison results.

5. The image processing apparatus according to claim 1, wherein the measurement curve and the estimation curve each show the time change of the pixel value of each pixel included in a pixel region that captures a plurality of structures performing different movements in the two-dimensional perspective image and the estimated two-dimensional perspective image, respectively.

6. The image processing apparatus according to claim 4, wherein the generation unit generates time-series data of the optimized three-dimensional volume image by adjusting the position of each structure depicted in the three-dimensional volume image based on the comparison results.

7. The image processing apparatus according to claim 1, wherein the generation unit generates time-series data of the optimized three-dimensional volume image such that the similarity between the measurement curve and the estimation curve is maximized.

8. The image processing apparatus according to claim 5, wherein the structure comprises a tumor to be subjected to radiotherapy on the subject and bone tissue surrounding the tumor.

9. An image processing method using an image processing device, We acquire time-series data of three-dimensional volume images showing the anatomical structure of the subject. Time-series data of two-dimensional fluoroscopic images showing the aforementioned anatomical structures is obtained. Based on the time-series data of the three-dimensional volume image, time-series data of the estimated two-dimensional perspective image is generated by estimating the two-dimensional perspective image. Based on the time-series data of the two-dimensional perspective image and the time-series data of the estimated two-dimensional perspective image, a comparison result is calculated by comparing a plurality of measurement curves showing the time change of the pixel value of each pixel in the two-dimensional perspective image with a plurality of estimation curves showing the time change of the pixel value of each pixel in the estimated two-dimensional perspective image. An image processing method that generates time-series data of an optimized three-dimensional volume image obtained by adjusting the three-dimensional volume image based on the comparison results.

Images