Image configuration device, image configuration method, image configuration program, and storage medium

Abstract

The present invention makes it possible to obtain a highly accurate PET image in PET inspection without fixing a subject and while reducing a calculation cost. This image configuration device 1 acquires measurement data measured by a PET device 2 in time series, divides the measurement data into a plurality of frame data, generates an inverse-projection image from the frame data, generates a mask image in a frame by setting RoI by using segmentation from the inverse-projection image, derives the center of mass from the mask image, derives a main component vector of an inertia matrix obtained from the center of mass, configures a homogeneous transformation matrix composed of the center of mass and a rotation matrix, calculates the amount of change between the homogeneous transformation matrix in an initial frame and the homogeneous transformation matrix in an nth frame, and corrects a rotation matrix serving as a main component vector column component in accordance with continuity of motion so that the amount of change is minimized.

Description

Image configuration device, image configuration method, image configuration program, and recording medium

[0001] This invention relates, for example, to an image composition apparatus, an image composition method, an image composition program, and a recording medium for non-invasively imaging the state of an object being imaged.

[0002] Traditionally, PET scans, using positron emission tomography (PET), have been performed as a medical imaging technique to identify lesions, examine the affected area, and determine diagnoses and treatment plans without harming the patient's (subject's) body. The images obtained from these PET scans (PET images) are then used as medical images for diagnosis.

[0003] PET scans require imaging time ranging from several minutes to several tens of minutes. Movement of the subject during this imaging time can cause image blurring, degrading the accuracy of PET images. In conventional PET scans, body movement is suppressed by fixing the subject's imaging area to the examination bed or other device. However, fixing the head, for example, in a PET scan targeting the head, places a burden on the subject. Therefore, instead of fixing the subject, a method has been proposed that allows for body movement and performs image correction according to the body movement.

[0004] For example, a data-driven framing method has been proposed that estimates the amount of body movement at each time point by analyzing measurement data obtained from PET scans and extracts frame images with low body movement. However, since the data-driven framing method generates frame images using only the imaging data from the static interval, frame images from intervals with high body movement frequency (dynamic intervals) are discarded, resulting in low data statistics and a problem of the PET image containing a lot of noise. Furthermore, obtaining high-precision PET images requires using a large number of frames, but increasing the number of frames leads to increased computational costs, which is a problem.

[0005] Enette Mae Revilla, Jean-Dominique Gallezot, Mika Naganawa, Takuya Toyonaga, Kathryn Fontaine, Tim Mulnix, John A Onofrey, Richard E Carson, Yihuan Lu. Adaptive data-driven motion detection and optimized correction for brain PET. NeuroImage. 2022. 252. 119031-119031.

[0006] The purpose of this invention is to provide an image composition device, an image composition method, an image composition program, and a recording medium that can reduce computational costs in PET examinations and obtain high-precision PET images without fixing the subject.

[0007] This invention provides a measurement data acquisition means for acquiring measurement data measured by a PET device in a time series,

[0008] The image configuration device, image configuration method, image configuration program, and recording medium are characterized by comprising: frame division means for dividing the measurement data into a plurality of frame data; back projection image generation means for generating a back projection image from the frame data; mask image generation means for generating a mask image in a frame by setting the RoI by segmentation from the back projection image; center of mass derivation means for deriving the center of mass from the mask image; principal component vector derivation means for deriving the principal component vectors of the inertia matrix obtained from the center of mass; homogeneous transformation matrix configuration means for configuring a homogeneous transformation matrix consisting of the center of mass and a rotation matrix with the principal component vector column components; calculation means for calculating the amount of change between the homogeneous transformation matrix in the initial frame and the homogeneous transformation matrix in the nth frame; and principal component vector correction processing means for correcting the rotation matrix with the principal component vector column components according to the continuity of the motion of the object to be measured so that the amount of change is minimized.

[0009] This invention makes it possible to reduce computational costs in PET scans and obtain highly accurate PET images without fixing the subject in place.

[0010] A block diagram showing an example of the configuration of the image processing device of the present invention. A flowchart showing the operation of the motion correction process. A graph showing the rotation vector estimation raw data and the correction result. A diagram showing the results of the motion correction experiment. A graph showing the resolution evaluation result.

[0011] The image constructing device 1 of the present invention is communicatively connected to an imaging device (PET device) 2 for positron emission tomography (PET) examinations, and has the function of constructing (reconstructing) a PET image based on measurement data output from the PET device 2.

[0012] The target of PET imaging (image acquisition target) can be an animal (for example, a human or a monkey) or a container made of a rigid body. Furthermore, when the imaging target is an animal, the target area for PET imaging can be a rigid part of the body (such as the head or shin), typically the human head. The following embodiment describes the case of acquiring a PET image of a human head.

[0013] PET is a technique in which a radiopharmaceutical labeled with a positron-emitting nuclide is administered to the imaging target, and the radiation emitted from inside the target (inside the body) is measured by a PET scanner. The acquired data is reconstructed into a tomographic image (PET image) with various corrections. PET images are used as medical images, and by analyzing PET images, they are used for evaluating physiological and pathological functions and for diagnostic imaging. In other words, examinations using a PET scanner (PET scan) allow for the identification of lesions and detailed examination of the relevant area without harming the patient's (subject's) body.

[0014] In PET, a tracer containing a radionuclide that emits positrons is introduced into the subject's body by injection or inhalation. The tracer accumulates in specific locations within the subject's body depending on its properties. For example, when a sugar-based tracer is used, it selectively accumulates in areas with high metabolic activity, such as cancer cells. At this time, positrons are emitted from the radionuclide contained in the tracer, and when the emitted positrons combine with surrounding electrons and annihilate each other, two gamma rays (so-called annihilation gamma rays) are emitted at approximately 180 degrees from each other. These two gamma rays are then detected by radiation detectors placed around the subject, and a PET image can be acquired (imaged) by reconstructing the detected data (measurement data). In this embodiment, measurement data is transmitted from the PET device 2 to the image constructor 1, where the measurement data is reconstructed.

[0015] Figure 1 is a block diagram showing the configuration of the image configuration device 1 of the present invention. As shown in Figure 1, the image configuration device 1 comprises a control unit 11, an input unit 12, a display unit 13, a communication unit 14, and an auxiliary storage unit 15. Each of the input unit 12, the display unit 13, the communication unit 14, and the auxiliary storage unit 15 is connected to the control unit 11.

[0016] The control unit 11 includes an arithmetic unit 16 and a main memory unit 17, and performs various calculations and control operations in the image composing device 1. The arithmetic unit 16 is an arithmetic processing unit having a CPU or MPU, etc. The main memory unit 17 has RAM (DRAM) and ROM, etc. The main memory unit 17 appropriately loads program data and data necessary for program execution stored in the auxiliary storage unit 15. The main memory unit 17 also stores the startup program of the image composing device 1 and default values ​​for various information.

[0017] The input unit 12 includes an input component that receives operation input from a user of the image generating device 1 (for example, a medical professional), and an input detection circuit interposed between the input component and the control unit 11. The input component is, for example, a touch panel (touch input means) and / or hardware keys. The input detection circuit outputs an operation signal or operation data to the control unit 11 in accordance with the operation (operation input) of each input component.

[0018] The display unit 13 has a display such as an LCD (liquid crystal display) or an organic EL display, and a display control circuit interposed between the display and the arithmetic unit 16. On the display, in addition to an appropriate operation screen, various medical images such as PET images are displayed.

[0019] When the input unit 12 has a touch panel, the touch panel is provided so as to overlap the display surface of the display of the display unit 13, and the touch panel can cooperate with the display of the display unit 13 to constitute a display with a touch panel (touch panel display). Further, the touch panel display may be configured to display a GUI (graphical user interface) having software keys (operation keys or operation buttons).

[0020] The communication unit 14 has a communication circuit for connecting to a communication line. The communication circuit is a wired communication circuit or a wireless communication circuit, and is communicably connected to other electronic devices (for example, the PET device 2) via the communication line according to an instruction from the control unit 11.

[0021] The auxiliary storage unit 15 is composed of a non-volatile memory such as an HDD, SSD, flash memory, EEPROM, etc., and stores programs and various data for the arithmetic unit 16 to control the operation of the image forming apparatus 1.

[0022] Various data used in the image forming apparatus 1 are stored (registered) in the auxiliary storage unit 15 of the image forming apparatus 1. Further, in the auxiliary storage unit 15, an image forming processing program 18 for automatically executing image forming processing in response to a user's operation input or when measurement data is input, and image forming processing data 19 necessary for executing the image forming processing are stored.

[0023] The image forming processing program 18 and the image forming processing data 19 are read from the auxiliary storage unit 15 as necessary and stored (expanded) in the main storage unit 17 (RAM). The operation of the image forming apparatus 1 is realized by the arithmetic unit 16 executing the image forming processing program 18 expanded in the main storage unit 17 (RAM).

[0024] The image composition processing program 18 has a plurality of programs for executing at least the processing of each step of the image composition processing described later. For example, the image composition processing program 18 includes a measurement data acquisition program for acquiring measurement data from the PET device 2, a frame division program for setting frame delimiters (frame division points), a correction data generation program for generating correction data for correcting measurement data so as to cancel the body movement of the subject, and a reconstruction program for reconstructing a PET image from the measurement data based on the correction data, etc.

[0025] The image composition processing data 19 has measurement data acquired from the PET device 2, data of frame division points (frame division point data) configured based on the measurement data, correction data for correcting the measurement data according to body movement, and data of a PET image (PET image data) reconstructed from the measurement data based on the correction data, etc.

[0026] Note that the configuration of the image composition device 1 shown in FIG. 1 is merely an example and does not necessarily have to be limited to this. For example, the image composition device 1 may be configured by a plurality of computer groups.

[0027] Hereinafter, an operation example of the image composition device 1 will be described. When the image composition device 1 starts the image composition processing, it acquires the measurement data transmitted from the PET device 2 (measured by the PET device 2) in time series, performs frame division processing by a predetermined method, performs body movement correction processing for processing each frame data to correct the measurement data, performs reconstruction processing with body movement correction, and generates a PET image with body movement correction. Note that a known technique can be applied to the reconstruction processing.

[0028] Furthermore, various methods can be employed in the frame division process, such as the fixed frame method and the adaptive frame method. The fixed frame method is a method in which measurement data is divided into fixed frames of several minutes to several tens of seconds, imaged, and then the images are aligned. The adaptive frame method is a method devised by the present inventors, which analyzes the trend of motion from the CoD (Center of Distribution) trace obtained by averaging the center coordinates of the LoR (Line of Response) contained in each 0.1-second frame, and realizes frame division that matches the timing of motion occurrence (Japanese Patent Application No. 2023-108744).

[0029] In short, the present invention's motion correction process automatically sets the Region of Interest (RoI) through segmentation (extraction of organ or tumor regions), constructs a mathematical algorithm to separate the translation component, and corrects the posture vector corresponding to the subject's motion. Specifically, it directly estimates body motion by mathematically analyzing frame data as described below.

[0030] Figure 2 is a flowchart showing the operation of the motion correction processing performed by the image composition device 1. As shown in Figure 2, when the image composition device 1 starts the motion correction processing, it generates a back projection image from each frame data (step S1), performs sensitivity correction (step S2), automatically sets the Region of Interest (RoI) by segmentation (extraction of organ or tumor regions) and generates a mask image (an image masked outside the region of interest) (step S3), derives the center of mass from the mask image (step S4), derives principal component vectors according to the center of mass (step S5), performs principal component vector correction processing (step S6), constructs a homogeneous transformation matrix (step S7), and then finishes the image composition processing. Note that the order of each step in the motion correction processing is not limited.

[0031] In step S1 (inverse projection image generation step), an inverse projection image taking into account ToF (Time of Flight) information is generated for each frame data. The inverse projection image is an image in which the locus of LoR taking into account ToF information is plotted in the image space. In step S2 (sensitivity correction step), the inverse projection image is divided by the overall sensitivity image pre-calculated from the geometric arrangement of the radiation detector and the normalization imaging data, and sensitivity correction is performed. The image after sensitivity correction generated in step S2 is the frame image F in frame n ｎ is referred to as.

[0032] Subsequently, in step S3 (RoI setting step), the frame image F ｎ is subjected to binarization processing with a threshold value derived by discriminant analysis, and then an RoI (Region of Interest) in which noise is removed by morphological operation is configured. For example, when the imaging object is a human head, in step S3, an RoI (head RoI) related to the head is configured. The image generated by step S3 is the mask image M in frame n ｎ and is composed of binary values of "inside RoI area = 1" and "outside RoI area = 0".

[0033] In step S4 (center of mass derivation step), the center of mass G ｎ is derived by the following [Equation 1]. In [Equation 1], the values at the i-th voxel position v ｎ =(x ｎ , y ｉ , z ｉ , y ｉ , z ｉ ) of the frame image F ｉ and the mask image M ｉ are respectively denoted as w

[0034] Also, when the coordinates based on the center of gravity are expressed by [Equation 2], the inertia tensor (inertia matrix) I ｎ is represented by the following [Equation 3], [Equation 4], and [Equation 5].

[0035]

[0036]

[0037]

[0038]

[0039] In step S5 (principal component vector derivation step), the inertia matrix I ｎ The principal component vector obtained by performing principal component analysis on r is ｎ１ ,r ｎ２ ,r ｎ３ Let these be the column elements of the rotation matrix R. ｎ The following is defined. The principal component vectors represent the three orthogonal axes with large head distributions, hence the rotation matrix R ｎ This represents the head's orientation in n frames. Center of mass G ｎ and rotation matrix R ｎ A 4x4 homogeneous transformation matrix consisting of these elements is represented by [Equation 6].

[0040]

[0041] The matrix represented by [Equation 6] represents the head position and orientation in n frames. The homogeneous transformation matrix in the initial frame is T ０ In this case, T in frame n ０ The change from [equation 7] is given by [equation 8].

[0042]

[0043]

[0044] The resulting change in [Equation 7] corresponds to the motion (body movement) in frame n necessary for body movement correction. Therefore, motion estimation is possible with the processing up to step S5, but the sign of the principal component vector cannot be mathematically stable and axis inversion occurs, so the rotation matrix R ｎ This will not be stable as is. Furthermore, the column components of the principal component vectors obtained in descending order of contribution rate are r ｎ１ ,r ｎ２ ,r ｎ３ The rotation matrix R ｎAlthough it is composed of these, in analyses targeting the human head, there is no significant difference in the contribution rates of the three vectors, and the order of the first, second, and third principal component vectors is frequently changed. Also, when body movement involving rotation of 45° or more occurs, the rotation matrix R ｎ It is not possible to uniquely determine this. To address these problems, in step S6, the present invention performs a principal component vector correction process that focuses on the continuity of motion, and in step S7, the rotation matrix R corrected in step S6 is determined. ｎ Using [Equation 6] and [Equation 8], the final homogeneous transformation matrix is ​​constructed.

[0045] Rotation matrix R ｎ For this, the principal component vector (r ｎ１ ,r ｎ２ ,r ｎ３ There are a total of 6 permutations of ). Furthermore, since there are positive and negative patterns for each of the three components for each permutation, there are a total of 48 patterns for one set of principal component vectors. In this invention, a 4x4 homogeneous transformation matrix [Equation 6] is created for all patterns, and the calculation [Equation 9] is performed.

[0046]

[0047] Based on the continuity of the translation component obtained in [Equation 9] [Equation 10], the optimal rotation matrix R is determined. ｎ We search for a reference T. Specifically, we search for a reference T. ０ Rotation matrix R in ０ We can arbitrarily determine this. No matter what pattern we choose, [Number 11] becomes the identity matrix and [Number 12].

[0048]

[0049]

[0050]

[0051] Next, the rotation matrix R in frame n ｎ From this, construct [Number 6], T ０ The change from is derived from [Equation 9]. At this time, the rotation matrix R that minimizes err in [Equation 13] for the obtained translation component [Equation 10] is determined. ｎThis is determined (set) as the optimal solution. Therefore, continuity between frames can be maximized. However, T ０ and T ｎ The axes of the right-handed and left-handed systems must coincide; those that do not are excluded from the calculation of the discriminant formula below. Therefore, by aligning the axes of the right-handed and left-handed systems, continuity between frames can be ensured. By continuously performing these processes, the rotation matrix R that maximizes continuity between frames is obtained. ｎ The rotation matrix R is found, ｎ This can resolve instability.

[0052] <Experimental Results>

[0053] For evaluation purposes, 50 frame data sets were created by rotating a single frame image by 60° in the Z-axis direction (body axis direction). Specifically, 50 3D images were created from a single 3D image. The 50 frames were divided into five intervals of 10 frames each. In the first interval (frames 1-10), there was no change from the initial state. In the second interval (frames 11-20), each frame was rotated by 6° in the Z-direction relative to the previous frame, for a total rotation of 60°. In the third interval (frames 21-30), the rotated state was maintained. In the fourth interval (frames 31-40), each frame was rotated by 6° in the Z-direction relative to the previous frame, returning to the initial state. In the final interval (frames 41-50), the initial state was maintained. Principal component analysis was performed on this data, and Figure 3 shows the amount of rotation around each axis obtained from the raw data of the rotation matrix, as well as the amount of rotation after correction using the proposed method. The instability of the principal component vectors described above can be confirmed from the raw data (Figure 3(a)). On the other hand, the corrected data to which the present invention is applied (Figure 3(b)) records a maximum rotation amount of 60 degrees only in the z-axis direction, and is almost identical to the true value (Figure 3(c)), thus confirming the effectiveness of the proposed method.

[0054] The actual motion, as shown in Figure 3(c) as the true value, involves rotating at a constant speed in the Z-axis direction (shown by the solid line), stopping when it reaches 60°, and then, after a certain period of time, decreasing until it returns to 0°. However, in the raw data (Figure 3(a)), the movement in the Z-axis direction (shown by the solid line) reaches 80°, and then decreases to around 20°, exhibiting a bizarre motion compared to the true value (Figure 3(c)). The same applies to the X-axis (shown by the dotted line) and the Y-axis (shown by the dashed line). This deviation in rotational components is called instability. This stems from the instability of the order and direction of the axes mentioned above, and the correction of the present invention can correct it to a value almost equivalent to the true value (Figure 3(c)), as shown in Figure 3(b).

[0055] This invention enables highly accurate motion correction of PET data acquired using general clinical protocols without requiring complex pre-processing, sensor installation, or lengthy post-processing. Furthermore, by implementing the method of this invention, it becomes possible to perform PET examinations with reduced burden on elderly patients or patients with severe illnesses who have difficulty with head fixation.

[0056] To demonstrate the effectiveness of the present invention, an experiment was conducted using test data obtained with a head-specific PET scanner. The subjects were eight healthy adult males aged 22 to 45 years with no history of brain injury or mental illness. After a 6-hour fast, they were administered 285 ± 23 MBq of FDG (8F-fluorodeoxyglucose). Forty-five minutes after administration, a 10-minute PET scan was performed using a head-specific PET scanner. For the first measurement, the jaw and forehead were restrained with bands for comparison, and imaging was performed while the subjects were motionless. Subsequently, a 15-minute measurement was performed with the head restraints removed. During this time, the subjects were instructed to move their entire head and gaze toward markers placed on the wall in front of them. The timing of gaze movement from marker to marker was 1 minute for the first 5 minutes, 30 seconds for the next 5 minutes, and 20 seconds for the last 5 minutes. To allow for variability in the timing of gaze movement, the elapsed time was not explicitly stated, and the count was based on each participant's brain count. The experimental results are shown in Figure 4. Figure 4 shows, from left to right, the reconstructed image from the first static imaging as a reference, the image reconstructed from the second imaging without motion correction (the head is moving but uncorrected), the reconstructed image after correction using direct motion measurement data from Kinect (after correction using motion measured by an external sensor), the reconstructed image after correction using the adaptive frame method, and the reconstructed image after correction according to the present invention. The image without motion correction shows significant blurring throughout the entire image, but it can be seen that the blurring is greatly reduced in each corrected reconstructed image. In particular, the image obtained using the proposed method is superior to that obtained with static imaging. This is called the Wobbling effect, and it is because the LoR sampling density was increased with high precision by motion correction.

[0057] Furthermore, to perform a quantitative evaluation, the peak-to-valley ratio, which is the ratio of peaks to valleys in the line profile, was determined from images of eight subjects for a pair of minute projection regions on the left and right sides of the midbrain called the inferior colliculus (corresponding to the central part of each cross-sectional view in Figure 4), and evaluated as an indicator of resolution (Figure 5). As shown in Figure 5, it was demonstrated that the peak-to-valley ratio value of the present invention is higher than the motion correction results of the other two methods. In addition, while the adaptive frame method required about 5 to 10 hours for post-processing, it was demonstrated that the method of the present invention could be completed in an extremely short time of about 2 to 3 minutes. Therefore, the present invention enables low computational cost motion estimation for frame data, thereby reducing computational costs. In other words, according to the present invention, it is possible to reduce computational costs in PET examinations and obtain highly accurate PET images without fixing the subject. Furthermore, it is possible to eliminate the need for repetitive image alignment according to the present invention, thereby shortening computation time. Specifically, while conventional methods required tens of thousands of calculation times to obtain detailed images, this invention makes it possible to obtain PET images with higher accuracy than before in a calculation time of 100 to 200 seconds.

[0058] In this invention, the image configuration device corresponds to the image configuration device 1, and similarly, the PET device corresponds to the PET device 2. The measurement data acquisition means corresponds to the measurement data acquisition program and the control unit 11 that operates thereacco. The frame division means corresponds to the frame division program and the control unit 11 that operates thereacco. The back projection image generation means corresponds to the control unit 11 that executes step S1. The mask image generation means corresponds to the control unit 11 that executes step S3. The center of mass derivation means corresponds to the control unit 11 that executes step S4. The principal component vector derivation means corresponds to the control unit 11 that executes step S5. The homogeneous transformation matrix configuration means corresponds to the control unit 11 that executes step S6. The calculation means corresponds to the control unit 11 that executes step S6. The principal component vector correction processing means corresponds to the control unit 11 that executes step S6. However, this invention is not limited to this embodiment and can take various other forms. Furthermore, the specific configurations and other details listed in the above embodiments are just examples and can be appropriately modified according to the actual product.

[0059] For example, the present invention can also be provided in the form of a storage medium, such as a computer-readable recording medium on which an image configuration program for operating an image output device is recorded. Examples of recording media include disk media such as CDs (Compact Discs), DVDs (Digital Versatile Discs), and flexible disks. Semiconductor media such as USB memory and SD (Secure Digital) memory cards are also applicable as recording media. In addition, embedded (internal) media incorporated into image output devices, such as ROMs, HDDs, or SSDs, rather than portable media, are also applicable as recording media.

[0060] This invention can be used in the industry for image-forming devices that non-invasively image the activity state within a subject being imaged.

[0061] 1...Image composition device 2...PET device 11...Control unit 15...Auxiliary storage unit 18...Image composition processing program

Claims

1. An image constructing apparatus comprising: measurement data acquisition means for acquiring measurement data measured by a PET device in a time series; frame division means for dividing the measurement data into a plurality of frame data; back projection image generation means for generating a back projection image from the frame data; mask image generation means for generating a mask image in a frame by setting the RoI by segmentation from the back projection image; center of mass derivation means for deriving the center of mass from the mask image; principal component vector derivation means for deriving the principal component vectors of the inertia matrix obtained from the center of mass; homogeneous transformation matrix constructing means for constructing a homogeneous transformation matrix consisting of the center of mass and a rotation matrix that is the column component of the principal component vector; calculation means for calculating the amount of change between the homogeneous transformation matrix in the initial frame and the homogeneous transformation matrix in the nth frame; and principal component vector correction processing means for correcting the rotation matrix that is the column component of the principal component vector according to the continuity of the motion of the object to be imaged so that the amount of change is minimized.

2. The image constructing apparatus according to claim 1, wherein the homogeneous transformation matrix constructing means constructs homogeneous transformation matrices for all patterns arising from a set of principal component vectors, and the principal component vector correction processing means sets the rotation matrix of the pattern in which the translation component is minimized as the amount of change as the optimal solution.

3. The image constructing apparatus according to claim 1 or 2, wherein the calculation means calculates the amount of change only when the axial directions of the right-handed system and the left-handed system coincide.

4. An image construction method comprising: acquiring measurement data measured by a PET device in a time series; dividing the measurement data into multiple frame data; generating a back-projection image from the frame data; generating a mask image for each frame by setting the RoI through segmentation from the back-projection image; deriving the center of mass from the mask image; deriving the principal component vectors of the inertia matrix obtained from the center of mass; constructing a homogeneous transformation matrix consisting of the center of mass and a rotation matrix with the principal component vectors as column components; calculating the change between the homogeneous transformation matrix in the initial frame and the homogeneous transformation matrix in the nth frame; and correcting the rotation matrix with the principal component vectors as column components according to the continuity of the motion of the object being imaged so that the change is minimized.

5. An image configuration program that causes a computer to function as: measurement data acquisition means for acquiring measurement data measured by a PET device in a time series; frame division means for dividing the measurement data into a plurality of frame data; back projection image generation means for generating a back projection image from the frame data; mask image generation means for generating a mask image in a frame by setting the RoI by segmentation from the back projection image; center of mass derivation means for deriving the center of mass from the mask image; principal component vector derivation means for deriving the principal component vectors of the inertia matrix obtained from the center of mass; homogeneous transformation matrix configuration means for constructing a homogeneous transformation matrix consisting of the center of mass and a rotation matrix that is the column component of the principal component vector; calculation means for calculating the amount of change between the homogeneous transformation matrix in the initial frame and the homogeneous transformation matrix in the nth frame; and principal component vector correction processing means for correcting the rotation matrix that is the column component of the principal component vector according to the continuity of the motion of the object being imaged so that the amount of change is minimized.

6. A computer-readable recording medium that records a program for causing a computer to function as: measurement data acquisition means for acquiring measurement data measured by a PET device in a time series; frame division means for dividing the measurement data into a plurality of frame data; back projection image generation means for generating a back projection image from the frame data; mask image generation means for generating a mask image in a frame by setting the RoI by segmentation from the back projection image; center of mass derivation means for deriving the center of mass from the mask image; principal component vector derivation means for deriving the principal component vectors of the inertia matrix obtained from the center of mass; homogeneous transformation matrix construction means for constructing a homogeneous transformation matrix consisting of the center of mass and a rotation matrix that is the column component of the principal component vector; calculation means for calculating the amount of change between the homogeneous transformation matrix in the initial frame and the homogeneous transformation matrix in the nth frame; and principal component vector correction processing means for correcting the rotation matrix that is the column component of the principal component vector according to the continuity of the motion of the object being imaged so that the amount of change is minimized.

Images