KR20130124135A - Apparatus and method for generating positron emission tomography image - Google Patents

Abstract

A method and apparatus generate an image in positron emission tomography (PET). An image generating method comprises the steps of: detecting signals emitted from tracers introduced into a target; dividing the detected signals into sections at predetermined intervals, and generating unit signals for each of the sections by accumulating signals within the section with respect to each of the sections; classifying the unit signals into a plurality of groups based on the characteristics of each of the unit signals; and generating a medical image for the target from the unit signals classified into the groups. [Reference numerals] (20) Computer;(210) Unit signal generating unit;(220) Classifying unit;(230) Movement estimating unit;(240) Image generating unit;(AA) Detected signal;(BB,QQ) Image;(CC) Detected signal;(DD) Unit signal #1;(EE) Unit signal #2;(FF) Unit signal #3;(GG) Unit signal #N;(HH) Group #1;(II) Standard;(JJ) Conversion factor #1;(KK) Group #2;(LL) Motion information #2;(MM) Conversion factor #2;(NN) Group #M;(OO) Motion information #M;(PP) Conversion factor #M

Description

Apparatus and method for generating Positron Emission Tomography image}

A method and apparatus for generating images in positron emission tomography.

Medical imaging devices that acquire images of the inside of the human body to diagnose patients provide information necessary for diagnosis of diseases. The medical imaging methods currently being used or developed in hospitals are largely divided into methods of obtaining anatomical images and physiological images. First, MRI (Magnetic Resonance Imaging) or CT (Computed Tomography) are examples of imaging techniques that provide detailed anatomical images of the human body at high resolution. These two-dimensional images of a human body, or a plurality of two-dimensional images using three-dimensional images with high resolution to produce the exact position and shape of the organs in the body. Second, examples of physiological imaging techniques include positron emission tomography (PET), which captures the metabolic processes in the body and contributes to the diagnosis of metabolic abnormalities.

Positron emission tomography (CT) produces a special radioactive tracer that emits a positron in the form of a component that participates in human metabolism. The tracer is injected into the human body by intravenous injection or inhalation, and the positron emitted by this tracer is combined with electrons , Which are emitted in the opposite direction to each other, are detected by using an external device to observe the position of the tracer and to observe the distribution pattern and the change of the distribution pattern with time.

In general, the gamma rays emitted from the tracer injected into the object may reach the signal detection device in a much smaller amount than the amount actually emitted due to the influence of dispersion or attenuation. Therefore, a relatively long detection time of several minutes is required to secure a sufficient detection amount to generate an image. However, since human organs are accompanied by relatively short periods of movement due to breathing, heartbeat, etc., when the object is photographed for several minutes of image acquisition time, the movement of the object is reflected on the image and is blurred and recorded. The blurring or blurring of the image caused by the relative movement between the photographing apparatus and the subject is called dynamic noise. Such noise is a major cause of lowering the resolution of positron emission tomography.

In positron emission tomography, there is provided an image generating method and apparatus for accurately classifying detected data and generating still images of high resolution. The present invention also provides a computer-readable recording medium on which a program for causing the computer to execute the method is provided. The technical problem to be achieved by the present embodiment is not limited to the technical problems as described above, and further technical problems can be inferred from the following embodiments.

According to an aspect of the present invention, there is provided a method of generating an image, the method comprising: detecting signals emitted from a tracker injected into an object, dividing the detected signals into sections at predetermined intervals, Accumulating the signals in the section to generate a unit signal for each section, classifying the unit signals into a plurality of groups based on characteristics of each of the unit signals, and from the unit signals classified into the groups. The method may include generating a medical image of the object.

The image generating method according to another aspect of the present invention may further include generating a two-dimensional sinogram for each section by using each unit signal, and the classifying may include the sinograms. The sinograms may be classified into a plurality of groups based on respective characteristics.

According to another aspect of the present invention, the characteristic may be a characteristic of a gradient representing a two-dimensional gradient of the sinogram.

According to yet another aspect of the present disclosure, the classifying may include calculating a feature value representing the characteristic of each unit signal, and classifying the unit signals into a plurality of groups based on the calculated feature value.

According to another aspect of the present invention, the feature value may be calculated from a correlation value representing the similarity between the respective unit signals.

According to another aspect of the present invention, the classifying may include determining a maximum value and a minimum value among the feature values, and allocating a group to each interval by dividing the maximum value and the minimum value into a predetermined number of intervals. The unit signal may be further classified into a group allocated to a section including a feature value of each unit signal.

The classifying according to another aspect of the present invention may further include arranging the unit signals according to a result of comparing the feature values, and classifying the unit signals into a plurality of groups based on the listed order. Can be classified.

According to another aspect of the present invention, the classifying may be classified using a k-means clustering algorithm.

According to another aspect of the present invention, the generating may generate an image of the object from the unit signals by regulating the unit signals so that the position of the tracker represented by each of the groups is matched. .

According to still another aspect of the present invention, there is provided a method of generating an image, the method comprising: estimating motion information of the tracker from the position of the tracker indicated by one of the groups as a reference, to the position of the tracker indicated by each group; The generating may include generating the image of the object from the unit signals by aligning the unit signals based on the motion information.

According to another aspect of the present invention, the motion information may be estimated based on a result of comparing unit signals allocated to one group as the reference and unit signals allocated to each group.

According to another aspect of the present invention, the motion information is estimated based on a result of comparing a sinogram that accumulates unit signals allocated to one group as the reference and a sinogram that accumulates unit signals allocated to each group. Can be.

According to still another aspect of the present invention, there is provided a computer-readable recording medium having recorded thereon a program for executing the above-described image generating method on a computer.

In accordance with another aspect of the present invention, an image generating apparatus includes a signal detector configured to detect signals emitted from a tracker injected into an object, dividing the detected signals into sections of a predetermined interval, and for each of the sections, a signal within the section. A unit signal generator for accumulating the signals and generating unit signals for each of the sections, a classifier for classifying the unit signals into a plurality of groups based on characteristics of the unit signals, and a unit signal classified into the groups. The image generating unit may generate a medical image of the object.

The image generating apparatus according to another aspect of the present invention may further include a sinogram generator for generating a two-dimensional sinogram for each section by using the respective unit signals, and the classification unit may include the sinogram. The sinograms may be classified into a plurality of groups based on their respective characteristics.

According to another aspect of the present invention, the classification unit may calculate a feature value representing the characteristic of each unit signal, and classify the unit signals into a plurality of groups based on the calculated feature value.

According to another aspect of the present invention, the classification unit may determine the highest value and the lowest value of the feature values, and set a group for each period by dividing the maximum value and the minimum value into a predetermined number of sections.

According to another aspect of the present invention, the classifying unit may further include a unit signal arranging unit for arranging each unit signal according to a result of comparing the feature values, and the unit signals may be divided into a plurality of groups based on the listed order. Can be classified.

According to another aspect of the present invention, the classification unit may be classified using a k-means clustering algorithm.

According to another aspect of the present invention, the image generator may generate an image of the object from the unit signals by regulating the unit signals so that the position of the tracker represented by each of the groups is matched.

An image generating apparatus according to another aspect of the present invention is a motion estimation for estimating the motion information of the tracker from the position of the tracker represented by one of the groups as a reference among the groups from the position of the tracker represented by each group The image generating unit may generate an image of the object from the unit signals by aligning the unit signals based on the motion information.

According to another aspect of the present invention, the motion information may be estimated based on a result of comparing unit signals allocated to one group as the reference and unit signals allocated to each group.

According to another aspect of the present invention, the motion information is estimated based on a result of comparing a sinogram that accumulates unit signals allocated to one group as the reference and a sinogram that accumulates unit signals allocated to each group. Can be.

In generating PET medical images using the image alignment method, the detected data are classified without relying on an external device, and then the classified data are sorted to generate an image, thereby accurately classifying the data and maintaining a high resolution. An image can be generated.

1 shows an image generating apparatus according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating the configuration of the computer illustrated in FIG. 1 and a flow of data according to the operation of the computer.
3 shows an example of LOR data.
4 shows an example in which LOR data is represented by a sinogram.
5 shows an example of classifying a unit signal into a plurality of groups.
6 shows another example of classifying a unit signal into a plurality of groups according to the feature value of the unit signal.
FIG. 7 illustrates another example of classifying a unit signal into a plurality of groups according to feature values of the unit signal.
8 is a flowchart of an image correction method, according to an exemplary embodiment.

Hereinafter, exemplary embodiments will be described in detail with reference to the accompanying drawings. In order to clearly describe the features of the embodiments, detailed descriptions of matters well known to those skilled in the art will be omitted.

FIG. 1 illustrates an image generating apparatus according to an embodiment, and illustrates an overall system for generating an image of a patient's body section. Referring to FIG. 1, the image generating apparatus includes a signal detector 10, a computer 20, a display device 30, and a user input device 40.

The signal detector 10 detects a signal emitted from the tracker injected into the object. The subject may be an organism such as an animal or a human. For example, when the subject is a human, the operator creates a special radiotracer that emits a positron in the form of a component that participates in human metabolism, and injects the tracer into the human body by intravenous injection or inhalation. . The injected tracer emits a positron, and when the emitted positron combines with the electron, two gamma ray pairs of 511keV are emitted in opposite directions. The signal detection unit 10 detects such gamma rays and transmits data on the detected gamma rays to the computer 20 in the form of a line of response (LOR). Here, a positron is a kind of radiation and is emitted from radioisotopes such as C-11, N-13, O-15, and F-18. By injecting a substance constituting the living body, generated using the above elements, a special radiotracer can be generated. The most commonly used tracer is a glucose-like substance called F-18-FDG, which injects the tracer in the body where glucose metabolism is concentrated, such as cancer. In areas where the tracer is concentrated, positrons are concentrated.

LOR is a data format indicating the position of a straight line in space, and FIG. 3 shows an example of LOR data. Referring to FIG. 3, the reaction between the positron and the electron emitted from the tracer 32 located in the detection space in the scanner 31 emits two gamma rays in the 180 degree direction, and the two gamma rays are placed on one straight line. do. 3 shows a case where two such straight lines 33 and 34 are detected, that is, two reactions of a positron and an electron are detected. Referring to the straight line 33, when the waterline perpendicular to the straight line 33 is drawn from the origin of the scanner 31, the distance to the waterline is r1 and the angle to the waterline is θ1, so the LOR for the straight line 33 is (r1, θ1). Can be expressed as Similarly, referring to the straight line 34, when the waterline perpendicular to the straight line 34 is lowered based on the origin in the scanner 31, the distance to the waterline is r2 and the angle to the waterline is θ2, so the LOR for the straight line 34 is (r2, θ2). As such, when two or more pieces of LOR data are acquired, the position of the tracker may be determined as an intersection point of the LOR data. Therefore, the signal detector 10 transmits the LOR for the detected gamma rays to the computer 20, and the computer 20 finally determines the position of the tracker from such LOR to generate an image.

Referring back to FIG. 1, the computer 20 generates an image of an object by using data obtained from the signal detector 10. The display device 30 displays an image generated from the computer 20 on the display panel. The user may input information necessary for the operation of the computer 20, such as commanding to start and end the operation of the computer 20 by using the user input device 40. However, the computer 20 may obtain necessary information from a storage device other than the user input device 40.

FIG. 2 is a diagram illustrating the configuration of the computer illustrated in FIG. 1 and a flow of data according to the operation of the computer. Referring to FIG. 2, the computer 20 includes a unit signal generator 210, a classifier 220, a motion estimator 230, and an image generator 240. 2 shows the flow of data according to the operation of each unit.

The unit signal generator 210 of FIG. 2 obtains a signal detected by the signal detector 10, divides the obtained signal into sections of predetermined intervals, and accumulates signals within sections for each of the divided sections. Generate unit signals. For example, the signals detected by the signal detector 10 may be listed in the order of detected time, and then divided into predetermined time intervals. Accordingly, each unit signal may be generated by accumulating signals within a predetermined time interval. Can be. The generated unit signals may also be arranged in chronological order.

Referring to the right data flow of FIG. 2, on the right side of the unit signal generation unit 210, unit signals arranged in total N times from unit signal # 1 to unit signal #N are illustrated.

 When the LOR signal is detected by the signal detector 10, one LOR cannot determine the position of the tracker, and even if two or more are small, the reliability of the determined position of the tracker is low. Accordingly, the unit signal generator 210 may determine a predetermined interval forming a section so that a sufficient amount of data can be accumulated to determine the intersection point of the LOR data as the tracker's position. However, the wider the interval, the more difficult it is to locate the correct tracker because of the tracker's movement. Therefore, the predetermined interval should be determined in consideration of the degree of movement of the object, the period of the movement, the time interval at which the scanner detects the LOR, and the like. Such a section may be a short time, for example, 1 second or less, but is not limited thereto. In this case, since the unit signal generated by accumulating the signals during the predetermined interval may include noise, histogram clipping may be applied to minimize the influence of the noise. Histogram clipping is a method of determining the minimum and maximum values of a signal represented by a histogram and limiting the values of the signal to values between the minimum and maximum values. Accordingly, values below the minimum value are limited to the minimum value, and values above the maximum value are limited to the maximum value.

The unit signal generated as described above is composed of several LORs, and a sinogram graph may be used as a method of accumulating such LORs. In order to specifically describe a method of representing LOR data in a sinogram, the following description will be made with reference to FIG. 4.

4 shows an example in which LOR data is represented by a sinogram. The projection information may be expressed in a sinogram graph format in which the horizontal axis is r and the vertical axis is θ. The LOR data described above may also represent LOR on the sinogram since r and θ are variables. LOR data having a value of (r, θ) corresponds to a coordinate of (r, θ) in the sinogram graph. Graph 40 shows the LOR for several gamma rays emitted from the tracer 32 in the detection space of the scanner 31 in the sinogram. The position coordinate point of the tracer 32 in the detection space of the scanner 31 corresponds to one curve 41 in the graph of Fig. Thus, if multiple tracers are in different coordinates, the synogram for the signal detected from such tracers will appear as multiple curves.

Accordingly, the unit signal generator 210 may accumulate LORs included in each unit signal to generate the unit signal in the form of a sinogram. In detail, the unit signal generator 210 may generate a sinogram corresponding to each unit signal, and each sinogram may be similar or identical to each other according to the movement of the object including the tracker 32. It may be different from each other.

Referring again to FIG. 2, the classifier 220 classifies the unit signals into a plurality of groups based on characteristics of each of the unit signals generated by the unit signal generator 210. Referring to the data flow on the right side of FIG. 2, a total of M groups from the group # 1 to the group #M are illustrated on the right side of the classification unit 220. Here, M is a number smaller than N, which is the total number of unit signals, and in general, a value of 5 or more may be used, but is not limited thereto. The value of M may be set differently depending on the computational performance of the computer or the resolution of the final image required.

The position of the tracker shown in each unit signal may be different according to the movement of the object. In order to generate a single still image by adding all of these unit signals, it is necessary to align each unit signal so that the position of the tracker indicated by each unit signal is matched. For each unit signal, since the signal to noise ratio is low, Sorting can be difficult. Thus, in order to improve the signal-to-noise ratio, a plurality of unit signals may be combined into one. For example, a concept of extracting feature values of each unit signal and adding unit signals having similar feature values to each other may be introduced. As the unit signals accumulate, the signal-to-noise ratio improves, thus enabling more accurate alignment.

In addition, when the unit signals are accumulated, an operation time required for aligning the unit signals is shortened. There is a problem that it takes a long time to align all the unit signals individually. For example, when N unit signals are aligned, a total of N-1 operations are required based on one unit signal. On the other hand, if the unit signals corresponding to the same position are grouped into one group and the groups are aligned with each other, the total operation time can be shortened. For example, when sorting M groups, only M-1 operations are required based on one group. Here, M is a number smaller than N, which is the total number of unit signals, and the computation time for aligning each group is shortened as compared with aligning each unit signal.

5 illustrates an example in which the classifier 220 classifies a unit signal into a plurality of groups.

Graph 50 of FIG. 5 shows the change in position of the tracker indicated by each unit signal when the unit signals are arranged in the order of detection time. As shown in FIG. 5, the classifier 220 classifies the unit signals 51, 52, 53, 54, and 55 corresponding to the position 1 of the tracker in the detection space of the scanner 31 into a group 1, and the tracker. The unit signals 56, 57, 58, 59, and 60 corresponding to the position 2 may be classified into group 2, and the unit signals indicating the same position may be classified into the same group. The reason why the position of the tracer alternates between 1 and 2 is that the tracer injected into the organ can move together with the periodic movement of the organ. When classifying unit signals for M positions, the classifier 220 may generate a total of M groups.

The classifying unit 220 classifies the unit signals into a plurality of groups based on a result of analyzing the unit signals and grouping the unit signals indicating the same position into one group. For example, in the case where the unit signal is represented by a sinogram, the position of the tracker indicated by the unit signal corresponds to the curve of the sinogram, and the classifying unit 220 may determine the angle based on the characteristics of the sinogram of each unit signal. The unit signals may be classified by the position of the tracker according to the degree of similarity of the sinogram.

Such a classification method has an advantage of reducing the total number of generated groups. For example, the position of the tracker may be the same even if they correspond to different phases within the movement period of the object. Hereinafter, 'phase information' is used as a term meaning information of which phase a detection time of each data corresponds to within a movement period of an object.

Referring to FIG. 5, when the classification method including phase information is used, the unit signals 51, 53, and 55 correspond to the same phase, and the unit signals 52 and 54 correspond to the same phase. It will be classified into another group. However, if the classifier 220 classifies the unit signal based on the characteristics of the unit signal itself regardless of the phase information in the period as described above, 51, 52, 53, 54, 55 indicating the position of the same tracker are all one. Can be classified into groups. Accordingly, although the unit signals 51, 53, 55 and the unit signals 52, 54 are out of phase with each other, the total number of groups generated by the classifying unit 220 is reduced, and as a result, the motion estimating unit 230 is performed. And the number of operations of the image generator 240 may be reduced.

The classifier 220 may calculate a characteristic value of the sinogram in order to determine a degree to which each sinogram is similar based on the characteristics of the sinogram of each unit signal. For example, the classifier 220 may calculate a feature value of a sinogram of each unit signal, classify unit signals having similar feature values of the sinogram into a group, and classify the unit signals into a plurality of groups. Since the sinogram is represented by a 2D graph, various feature values for determining the similarity of the 2D graph may be used as the feature value of the sinogram.

As the feature value, the classifying unit 220 may use a correlation indicating the correlation of the sinogram of each unit signal. Accordingly, the classifying unit 220 may classify the unit signals into a plurality of groups by grouping unit signals having a small correlation value between the calculated unit signals, that is, highly related units, into the same group. As another example, the classifying unit 220 calculates a result of applying a specific two-dimensional filter to a sinogram of each unit signal as a feature value, groupes unit signals having similar feature values into the same group, and collects a plurality of unit signals. Can be divided into groups. An example of such a filter may be a Gabor filter or the like. In addition, various two-dimensional filters may be applied.

As another example, the classifying unit 220 generates a gradient, which is a two-dimensional differential image representing a two-dimensional gradient of a sinogram, calculates a feature value of the gradient, and unit signals having a similar feature value of the gradient. The unit signals may be grouped into the same group to classify the unit signals into a plurality of groups. In addition, various methods of calculating feature values of an image may be applied.

As a method of classifying unit signals having similar feature values calculated as described above, a unit signal may be classified into a plurality of groups by grouping unit signals having a difference in sinogram feature value equal to or less than a predetermined threshold value into the same group. Various data clustering algorithms may be used.

For example, a k-means clustering algorithm can be used that can quickly and effectively cluster given data into k groups. In detail, the classifying unit 220 may classify the unit signals into a plurality of groups by calculating the characteristic values of the sinogram of each unit signal and clustering the unit signals by a k-means clustering algorithm. .

In addition, various classification algorithms such as Gaussian mixture model analysis, Principal Components Analysis (PCA) classification method, and linear discriminant classification (LDA) classification method may be used.

Alternatively, the classifier 220 may use a gradient vector flow (GVF) SNAKE algorithm to classify the unit signals based on the characteristics of the sinogram of each unit signal. The SNAKE algorithm is an algorithm for detecting edges in an image, and the GVF SNAKE algorithm is a kind of SNAKE algorithm. An SNAKE algorithm is an algorithm for detecting an edge from a distribution of the slope vector of the edge in the image. The characteristic of the image may be extracted from the detected edge using the GVF SNAKE algorithm. When the GVF SNAKE algorithm is applied to a sinogram that is a two-dimensional graph, the classification unit 220 may classify the unit signal based on the characteristics of the sinogram of each unit signal generated using the GVF SNAKE algorithm. The GVF SNAKE algorithm has good processing performance even at low signal-to-noise ratio sinograms and is suitable for application to sinusoidal sinograms.

6 shows another example of classifying a unit signal into a plurality of groups according to the feature value of the unit signal. Referring to FIG. 6, the classifier 220 obtains a minimum value and a maximum value of a feature value, divides the minimum value and the maximum value into M sections, and outputs the same group of unit signals having a feature value corresponding to each section. Unit signals may be classified into a total of M groups.

FIG. 7 illustrates another example of classifying a unit signal into a plurality of groups according to feature values of the unit signal. Referring to FIG. 7, the classifying unit 220 lists each unit signal in the order of the magnitude of the feature values, bundles the N / M unit signals into one group in the listed order, and unites the unit signals into a total of M groups. Can be classified.

2, the motion estimator 230 of FIG. 2 estimates tracker motion information for each of the groups from the groups generated by the classifier 220. For example, the motion estimator 230 may track the motion of the tracker from the position of the tracker indicated by one group, which is a reference among the groups generated by the classification unit 220, to the position of the tracker indicated by each group. Estimate Such motion information may reflect information about a change in the position of the tracker accompanying the movement of the object. For example, if the position of the tracker indicated by each group is different according to the movement of the object, the motion information of each group is based on the position of the tracker indicated by group # 1, and indicates the distance and direction the tracker's position is moved. Can be used as meaning. Such motion estimation can be calculated only by analyzing unit signals classified into groups without the help of external devices (eg, motion sensors). Accordingly, the motion estimator 230 may estimate motion information according to a result of comparing unit signals allocated to one group as a reference and unit signals allocated to each group.

 The position of the tracker indicated by each group can be known from the sinogram, and accordingly, the motion estimator 230 accumulates the sinogram that accumulates unit signals assigned to one group as a reference and unit signals assigned to each group. According to the result of comparing the accumulated sinograms, the change of the position of the tracker may be estimated, and motion information may be estimated therefrom.

The motion estimator 230 may determine group # 1 of the M groups as a reference and estimate motion information from the group # 1 to each group. As such a method of estimating motion information of an object in an image, an optical flow for estimating a distribution of a motion vector of an object or a sensor in coordinates, or a snake algorithm for detecting an edge of the image (SNAKE) A technique such as an algorithm) can be used, and various methods for motion estimation in motion can be applied.

Referring to the right side of FIG. 2, the motion information # 2 means a change from the position of the tracker indicated by the group # 1, which is a reference, to the position of the tracker indicated by the group # 2, and although not shown in the drawing, The corresponding motion information # 3 means a change from the tracker's position indicated by the group # 1 to the tracker's position indicated by the group # 3. Similarly, the motion information #M means a change from the tracker's position indicated by the group # 1 to the tracker's position indicated by the group #M.

Hereinafter, an example of a method of estimating the motion information of the tracker in three-dimensional space will be described.

When the detection space in the scanner 31 is three-dimensional, the step of acquiring a signal on a two-dimensional plane (xy plane) is performed with respect to different z values on the z-axis (in the axial direction of the scanner when the scanner is cylindrical). The position of the tracker can be represented in three-dimensional space. To this end, the signal detector 10 may repeat the step of acquiring a signal on a two-dimensional plane (xy plane) with respect to different z values while moving in the z-axis direction, or the signal detector 10 may be different from each other. A signal on a two-dimensional plane (xy plane) may be simultaneously acquired for the z value.

According to this, the unit signal generator 210 accumulates the signals obtained as described above, and generates N unit signals, and the classifier 220 generates the same or similar trackers in the three-dimensional space represented by the unit signals. The unit signals may be grouped into the same group, and the N unit signals may be classified into the M groups. The motion estimator 230 may estimate motion information of the tracker for each of the groups from each of the groups.

For example, the motion estimator 230 is a three-dimensional position of the tracker represented by each of the remaining groups from the tracker's three-dimensional position represented by one of the groups generated by the classifier 220 as a reference It is possible to estimate the motion information of the tracker on the road in three dimensions.

As an example of a method of estimating motion information # 2, which means a change from the tracker's three-dimensional position indicated by group # 1 to the tracker's three-dimensional position indicated by group # 2, based on the group # 1, the motion estimation unit In operation 230, the motion corresponding to the z-axis direction of the tracker's three-dimensional motion may be estimated first, and then the motion corresponding to the xy plane (or z plane) direction may be estimated. The motion information of each group may be represented by a three-dimensional vector.

For example, if the position of the tracer is the same in the plane whose z-axis value is constant k in the reference group # 1 and the plane whose z-axis value in the group # 2 is constant k + a (a> 0) In estimating the motion information # 2, the motion estimator 230 may recognize that the tracker has moved in the positive direction on the z-axis. In this way, the motion estimation unit 230 determines the direction of the z-axis motion vector of the motion information # 2, and determines the magnitude of the z-axis motion vector from the size of a, thereby determining the z-axis of the motion information # 2. Directional motion information can be estimated. However, this is only one example of estimating motion information # 2 in the z-axis direction, but is not limited thereto.

When the Z-axis motion information of the motion information # 2 is estimated, the motion estimator 230 may estimate motion information on the z plane of the motion information # 2. To this end, the motion estimator 230 estimates the tracker's motion information on any one z plane as a representative and estimates the tracker's motion in the x-y direction. Alternatively, the motion estimator 230 may estimate motion information of the tracker on the plurality of z planes, and may estimate motion information of the tracker in x-y direction.

For example, the motion estimation unit 230 estimates the motion information of the tracker in the x-y direction with respect to each of the plurality of z planes in the detection space of the scanner 31. When the z plane corresponding to z = 1 is the first plane, when estimating the motion information # 2, the motion estimation unit 230 determines the group # 2 from the position of the tracker indicated by the first plane of the group # 1 as the reference. By estimating the change in the tracker's position indicated by the first plane, the motion information of the tracker on the first plane from the group # 1 to the group # 2 can be estimated.

Similarly, the motion information of the tracker on the second plane corresponding to z = 2 may be estimated. As the method is repeated for each z plane, the motion estimator 230 may estimate motion information of the tracker on the plurality of z planes for each z plane. Accordingly, the motion estimator 230 may estimate motion information of the motion information # 2 in the x-y direction from the motion information of the tracker on the plurality of z planes.

The motion estimation unit 230 may generate the tracker's motion information # 2 by combining the estimated tracker's x-y direction motion information with the tracker's estimated z-axis direction information according to the above-described method. As the method is performed for each group, the motion estimator 230 may generate motion information # 2 to motion information #M for M groups.

An example of a method of estimating the motion information on the two-dimensional plane by the motion estimator 230 is an optical flow for estimating a distribution of a motion vector of an object or a sensor in coordinates as described above, or an edge of an image. Techniques such as a Snake algorithm for detecting the signal may be used, and various methods for motion estimation in the image may be applied.

The motion estimator 230 may use the histogram distribution of the sinogram for each plane to estimate the motion information on the 2D plane. For example, the motion estimator 230 estimates the motion information on the xy plane of the motion information # 2 and the two-dimensional sinogram for the first plane of the group # 1 and the two-dimensional plane for the first plane of the group # 2. Each sinogram can be represented by a histogram distribution. Since the histogram distribution is an example of a method of representing the characteristics of the sinogram, the motion estimator 230 may be applied to estimating the motion information in the z-axis direction. In addition, the classifier 220 may also use the sinogram of the unit signal. By using the histogram distribution of, unit signals having similar histogram distribution may be classified. A method of representing a two-dimensional image or a two-dimensional graph as a histogram distribution can be understood by those skilled in the art.

As described above, the motion estimation unit 230 may first estimate the z-axis motion information of each group by using the histogram distribution of the sinogram, and then estimate the motion information on the xy plane.

The image generator 240 may generate a medical image of the object from the unit signals classified into a plurality of groups by the classifier 220. For example, the image generator 240 generates a medical image of the object from the unit signals by aligning the unit signals included in the groups based on the motion information of each group estimated by the motion estimator 230. can do.

For example, the image generator 240 generates a still image from each group based on the motion information estimated by the motion estimator 230 for each of the M groups generated by the classifier 220. do. To this end, the image generator 240 may generate a conversion factor including motion information of each group for each group, and the conversion factor may be used as a variable in the image generation process. Such conversion factors of each group are shown on the right side of the image generator 240 of FIG. 2.

The image generator 240 may generate an image by performing an iterative reconstruction by reflecting the motion information. For example, using a conversion factor including motion information of each group as a variable of the image generation algorithm, the position of the tracker in each group is matched to the position of the tracker in group # 1, which is the reference. As the unit signals are aligned, all unit signals may be finally converted to generate a still image.

Iterative reconstruction is an example of an algorithm that estimates the input signal when the transfer function is known and the output signal is known. That is, after the initial value of the input signal is set to a predetermined value, a transfer function is applied to the input signal, and the operation is repeatedly performed while changing the input signal until the output signal becomes a desired output signal. .

For example, in the case of positron emission tomography, an LOR signal obtained from the signal detector 10 may be an input signal, and an image generated therefrom may be an output signal. Accordingly, a system matrix for reconstructing an image from the LOR signal may be a conversion factor. The image generator 240 may have a conversion factor for each group in order to reconstruct one still image simultaneously from a plurality of group signals representing positions of different trackers. At this time, the image generating unit 240 reflects the motion information of each group to the conversion factors of each group, so that the position of the tracker represented by each group is aligned in the process of reconstructing the image, thereby finally generating a still image. can do.

For example, when the group # 1 is set as the reference group by the motion estimation unit 230, the image generator 240 converts the transform factor # 2 when reconstructing the unit signals classified into the group # 2 into an image. It can be used as a variable of an image generating algorithm, and when transforming and reconstructing unit signals classified into group #M into an image for an object, the transform factor #M can be used as a variable of the image generating algorithm. Since each transform factor includes motion information of each group, the image generator 240 may generate a still image having no noise from all unit signals included in the M groups.

8 is a flowchart of an image correction method, according to an exemplary embodiment. As illustrated in FIG. 8, the image correction apparatus detects a signal emitted from a tracer injected into an object (81), generates a unit signal from the detected signal (82), and classifies the unit signals into a plurality of groups (83). The motion information of each group is estimated (84) and the groups are aligned to generate an image (85).

In operation 81, the signal detecting apparatus signal detector 10 detects gamma rays emitted from the tracer injected into the object and transmits the gamma rays to the unit signal generator 210 of the computer 20 as LOR data. In operation 82, the unit signal generator 210 acquires a signal detected by the signal detection device signal detector 10, divides the acquired signal into intervals of a predetermined interval, and signals within the interval for each of the divided intervals. Is accumulated to generate a unit signal. In operation 83, the classifier 220 classifies the unit signals into a plurality of groups based on characteristics of each of the unit signals generated by the unit signal generator 210. In step 84, the motion estimator 230 estimates the tracker's motion information for each of the groups from the groups generated by the classifier 220 without the help of an external device (eg, a motion sensor). In operation 85, the image generator 240 generates a medical image of the object by arranging the groups based on the motion information of each group estimated by the motion estimator 230.

According to the embodiments as described above, in the method for generating an image by positron emission tomography for a moving object, by performing an accurate classification based on the characteristics of the unit signal itself, a higher resolution still image Can be generated. In order to classify unit signals, conventionally, a method of determining phase information by mapping each unit signal to a respiratory cycle or a heartbeat cycle using an external device and classifying them according to the phase information has been used. Phase information is easily determined by synchronizing the time information from which the unit signal is detected with the movement period, but the error is not because the respiratory cycle or the heartbeat cycle does not exactly match the movement of the subject or the tracer movement according to the movement of the subject. There was a disadvantage.

However, in the above-described embodiments, since the unit signals are classified without depending on the external device based on the characteristics of the unit signal indicating the position of the tracker, the disadvantage of using the external device as described above is overcome. Thus, more accurate image alignment is possible, resulting in a clear still image. In addition, in classifying the unit signal, the user may preset the number of groups or input other adjustment factors through the user input device, which may be used in a trade-off relationship between image quality and computational load on a computer. A user may generate an image of a desired quality.

On the other hand, the image generating method shown in Figure 8 can be written as a program that can be executed in a computer, it can be implemented in a general-purpose digital computer to operate the program using a computer-readable recording medium. The computer readable recording medium may include a storage medium such as a magnetic storage medium (eg, a ROM, a floppy disk, a hard disk, etc.), an optical reading medium (eg, a CD-ROM, a DVD, or the like).

10: signal detection unit 20: computer
30: display device 40: user input device
210: unit signal generation unit 220: classification unit
230: motion estimation unit 240: image generating unit

Claims (24)

Detecting signals emitted from the tracer injected into the subject;
Dividing the detected signals into sections of a predetermined interval, and accumulating signals in the sections for each of the sections to generate a unit signal for each section;
Classifying the unit signals into a plurality of groups based on characteristics of each of the unit signals; And
And generating a medical image of the object from unit signals classified into the groups.
The method of claim 1,
Generating a two-dimensional sinogram for each section by using the unit signals;
The classifying may include classifying the sinograms into a plurality of groups based on characteristics of each of the sinograms.
3. The method of claim 2,
Wherein the characteristic is a characteristic of a gradient representing a two-dimensional gradient of the sinogram.
The method of claim 1,
The classifying may include calculating a feature value representing a characteristic of each unit signal, and classifying the unit signals into a plurality of groups based on the calculated feature values.
5. The method of claim 4,
And the feature value is calculated from a correlation value representing a similarity between the unit signals.
5. The method of claim 4,
The classifying may further include determining a highest value and a lowest value among the feature values, and allocating a group to each interval by dividing the maximum value and the minimum value into a predetermined number of intervals.
And classifying the unit signals into groups allocated to the section including the feature values of the unit signals.
5. The method of claim 4,
The classifying may further include arranging the unit signals according to a result of comparing the feature values.
The medical image generating method of classifying the unit signals into a plurality of groups based on the listed order.
The method of claim 1,
The classification may be performed using a k-means clustering algorithm.
The method of claim 1,
The generating may include generating an image of the object from the unit signals by regulating the unit signals so that the position of the tracker represented by each of the groups coincides with each other.
The method of claim 9, wherein
Estimating movement information of the tracker from the position of the tracker represented by one of the groups as a reference, to the position of the tracker represented by each group;
The generating may include generating an image of the object from the unit signals by aligning the unit signals based on the motion information.
The method of claim 10,
And the motion information is estimated based on a result of comparing unit signals allocated to one group as the reference and unit signals allocated to each group.
The method of claim 10,
The motion information is estimated based on a result of comparing a sinogram that accumulates unit signals allocated to one group as the reference and a sinogram that accumulates unit signals allocated to each group.
A signal detector detecting signals emitted from a tracker injected into the object;
A unit signal generator for dividing the detected signals into sections of a predetermined interval and accumulating signals in the sections for each of the sections to generate a unit signal for each section;
A classification unit classifying the unit signals into a plurality of groups based on characteristics of each of the unit signals; And
And an image generator configured to generate a medical image of the object from unit signals classified into the groups.
The method of claim 13,
Further comprising a sinogram generator for generating a two-dimensional sinogram (sinogram) for each section by using the unit signal,
The classification unit classifies the sinograms into a plurality of groups based on characteristics of each of the sinograms.
15. The method of claim 14,
Wherein the characteristic is a characteristic of a gradient representing a two-dimensional gradient of the sinogram.
The method of claim 13,
The classification unit calculates a feature value representing a characteristic of each unit signal, and classifies the unit signals into a plurality of groups based on the calculated feature values.
17. The method of claim 16,
And the feature value is calculated from a correlation value representing a similarity between the unit signals.
17. The method of claim 16,
The classification unit may further include a group setting unit configured to determine a maximum value and a minimum value among the feature values, and set a group for each period by dividing the maximum value and the minimum value into a predetermined number of sections.
And classifying each unit signal into a group allocated to a section including a feature value of each unit signal.
17. The method of claim 16,
The classification unit further includes a unit signal arranging unit for arranging the unit signals according to a result of comparing the feature values.
The medical image generating apparatus classifies the unit signals into a plurality of groups based on the listed order.
The method of claim 13,
The classification unit classifies the image using a k-means clustering algorithm.
The method of claim 13,
And the image generating unit generates an image of the object from the unit signals by regulating the unit signals so that the position of the tracker represented by each of the groups coincides with each other.
22. The method of claim 21,
And a motion estimator for estimating the motion information of the tracker from the position of the tracker represented by one of the groups as a reference, to the position of the tracker represented by each group,
And the image generator generates an image of the object from the unit signals by aligning the unit signals based on the motion information.
23. The method of claim 22,
And the motion information is estimated based on a result of comparing unit signals assigned to one group as the reference and unit signals allocated to each group.
23. The method of claim 22,
And the motion information is estimated based on a result of comparing a sinogram that accumulates unit signals allocated to one group as the reference and a sinogram that accumulates unit signals allocated to each group.

Images