CT real-time 4D gating imaging method based on projection data

Abstract

The invention discloses a CT real-time 4D gating imaging method based on projection data. The method comprises the following steps: 1) acquiring a CT projection of a target to be detected and dividing the CT projection into a plurality of groups according to an acquisition angle; carrying out pixel combination on each CT projection and then adopting an n-dimensional vector for representation; an original matrix is constructed according to the vectors corresponding to the same angle group, principal component information is extracted from the matrix based on a PCA method, and superposed signals containing respiration and heartbeat are obtained; 2) separating breathing and heartbeat signals from the superposed signals; performing respiratory gating processing on the CT projection according to the respiratory signal, and performing gating division on the CT projection in the respiratory movement period; carrying out heartbeat gating processing on the CT projection in the breathing stationary period according to the heartbeat signal; and 3) reconstructing the CT projection groups after breathing and heartbeat gating processing to obtain a 4D gating image of the to-be-detected target. The real-time performance of CT gating imaging is achieved, and meanwhile the gating information extraction precision is improved.

Description

Technical Field

[0001] The invention belongs to the field of medical imaging and X-ray imaging, and relates to a CT real-time 4D gated imaging method based on projection data. Background Art

[0002] Computed Tomography (CT) is an imaging method that uses an X-ray source to perform transmission scanning on an object, uses a detector to detect ray attenuation, and then uses a reconstruction algorithm to reconstruct a tomogram by a computer. CT imaging has important applications in clinical diagnosis, preclinical research, and industrial non-destructive testing. In the field of small animal imaging, CT imaging can provide high-contrast images of chest structures, such as lung tissue, bones, heart chambers and blood vessels with enhanced contrast. It is also suitable for most pathological tests related to chest organs, and has become an important imaging technology.

[0003] However, due to the periodic expansion and contraction of the breathing and heartbeat of small animals, the collected CT data is the superposition result of multiple periodic movements. This causes image blur when reconstructing the CT image, that is, motion artifacts. When there are lesions on the lungs, heart or surrounding tissues and organs, the periodic movement will cause the lesions to expand in size on the reconstructed image, the contrast of the lesions will decrease and they may even be submerged in the background, affecting the accuracy of diagnosis. In order to reduce the errors caused by the above movement, the collected CT data can be divided into multiple phases according to the movement cycle, and then dynamic imaging is performed, that is, 4D gating technology that takes into account breathing and heartbeat movement. The group reconstruction of CT data can be achieved through gating technology, which can effectively remove motion artifacts and can be used to achieve 4D dynamic CT reconstruction.

[0004] Data Driven Gating (DDG) is a technology that directly extracts respiratory and cardiac motion signals through changes in information in CT chest projection. In contrast, External Device Gating (EDG) requires additional respiratory and heartbeat monitoring equipment to be connected to the patient, which increases the discomfort of the subject. And due to its operational complexity and the difficulty in accurately aligning the time of motion signals with CT data, it has gradually been replaced by data driven gating.

[0005] The commonly used CT data-driven gating methods include the centroid method (COM), frequency analysis method, and principal component analysis (PCA). In the process of extracting motion information, the three methods all perform overall data-driven gating after the complete acquisition of CT projections. The centroid method (COM) requires manual delineation of the region of interest (ROI) or automatic determination of the ROI position through multiple acquisitions during implementation, which increases the acquisition time and radiation dose. The frequency analysis method determines the region of interest and respiratory frequency by analyzing the Fourier spectrum of each image or sinusoidal pixel, and then estimates the respiratory signal by the change of the total count in the region over time. This method has the characteristics of many parameters and many calculation stages, so the amount of calculation is relatively large. The overall PCA method directly performs calculations on the complete CT projection, which will reduce the accuracy of motion information extraction. When the projection is large and the number of projections is large, there will be a large amount of calculation, resulting in a long data processing time.

[0006] Problems with existing technologies:

[0007] (1) Multiple acquisition cycles are required, which increases the acquisition time and radiation dose;

[0008] (2) The accuracy of motion information extraction is not high;

[0009] (3) The additional data processing time caused by the large amount of calculation makes it impossible to achieve real-time performance. Summary of the invention

[0010] In view of the problems existing in the prior art, the purpose of the present invention is to provide a real-time CT 4D gated imaging method based on projection data, which solves the problem of inaccurate information caused by additional rotation information brought by projection angle changes when extracting overall PCA gated information, as well as the problem of additional time overhead caused by large calculation amount. The present invention achieves the goals of universal applicability, real-time performance and high accuracy for CT cardiothoracic imaging.

[0011] The process of the CT real-time 4D gated imaging method of the present invention is as follows: firstly, the CT projection data is preprocessed, and the preprocessing includes pixel merging and filtering of the CT projections at the same angle, and then the motion information is extracted by using the PCA-based cell-to-cell data-driven gating technology; secondly, a bidirectional Hamming window FIR filter is used to perform bandpass filtering on the extracted respiratory and heartbeat superposition signals; then, the separated respiratory signals are subjected to respiratory gating processing; then, the stationary period information obtained by respiratory gating is combined with the heartbeat signal to perform heartbeat gating processing; finally, the gated CT projections are grouped and reconstructed to obtain 4D gated imaging.

[0012] Before extracting motion information through CT projection, data preprocessing, pixel merging and filtering are performed to reduce noise and greatly reduce the amount of calculation.

[0013] The preprocessed CT projections are overlapped and grouped by inter-cell PCA data-driven gating to extract more accurate superposition signals of breathing and heartbeat.

[0014] A bidirectional Hamming window FIR filter is used to perform bandpass filtering on the superimposed signal of breathing and heartbeat, and the breathing and heartbeat signals are separated based on the frequency.

[0015] Cyclic step-by-step respiratory gating is used to accurately determine the respiratory stable period and the inhalation and exhalation boundary, and then the respiratory movement period is gated and grouped to obtain complete respiratory grouping. Each group contains CT projections of the chest in the same spatial state for dynamic reconstruction.

[0016] A respiratory and heartbeat dual gating method is used to perform heartbeat gating during the respiratory stable period, and the heartbeat cycle is confirmed and grouped based on periodic peak detection to obtain CT projection groupings of different cardiac states, including atrial systolic CT projection, ventricular systolic CT projection and whole heart diastolic CT projection.

[0017] The grouped projections obtained by this method can be used in CT imaging analytical imaging algorithms and iterative imaging algorithms and other reconstruction methods, and the method is not limited by the reconstruction method.

[0018] The technical solution of the present invention is:

[0019] A CT real-time 4D gated imaging method based on projection data, the steps of which include:

[0020] 1) Acquire CT projections of the target to be detected and divide the CT projections into multiple groups according to the acquisition angle; perform pixel merging on each CT projection in the same angle group, and represent each merged CT projection with an n-dimensional vector, where n is the number of pixels in the merged image; arrange the n-dimensional vectors corresponding to the same angle group according to the scanning time of each CT projection data in the same angle group to construct an original matrix, and then extract the principal component information from the original matrix based on the PCA method, and arrange the weight factors corresponding to the selected principal components according to the CT projection acquisition order to obtain a superposition signal containing breathing and heartbeat;

[0021] 2) using a bidirectional Hamming window FIR filter to perform bandpass filtering on the superimposed signal to separate the respiratory signal and the heartbeat signal; performing respiratory gating processing on the CT projection according to the respiratory signal to divide the CT projection of the respiratory movement period and the CT projection of the respiratory stable period, and performing gating division on the CT projection of the respiratory movement period to obtain CT projection groups in the same state of the chest during the respiratory process; performing heartbeat gating processing on the CT projection in the respiratory stable period according to the heartbeat signal to obtain CT projection groups corresponding to different heartbeat stages;

[0022] 3) Reconstruct the CT projection groups after respiratory gating processing and heartbeat gating processing to obtain 4D gated imaging of the target to be detected.

[0023] Furthermore, the method of performing gated grouping on the CT projections during the respiratory movement period to obtain the CT projection groups in the same state of the chest during the respiratory process is as follows:

[0024] 21) Find all peak positions peak through the AMPD peak-finding algorithm, and determine the boundary between inspiration and expiration in each respiratory cycle according to each peak position peak; wherein, the coordinate positions kpeak[i]+1 and peak[i+1]-1 corresponding to the left and right CT projections of the i-th peak position peak[i] are used as the boundaries highLeft and hightRight of the i-th interval;

[0025] 22) Taking the ith peak position peak[i] as the reference, move the coordinates corresponding to p CT projections in the direction of the next peak position peak[i+1] as the left boundary of the ith stable respiratory period stableLeft; taking the peak position peak[i+1] as the reference, move the coordinates corresponding to q CT projections in the direction of the peak position peak[i] as the right boundary of the ith stable respiratory period stableRight;

[0026] 23) Taking the area between two adjacent peak positions peak[i] and peak[i+1] as a cycle, taking the max(stableLeft, stableRight) in the cycle as the lower bound of the respiratory movement period in the cycle, and taking the min(highLeft, hightRight) in the cycle as the upper bound of the respiratory movement period in the cycle;

[0027] 24) The respiratory movement period in the cycle is divided into equal intervals according to a preset number of groups, so as to obtain CT projection groups in the same state of the chest during the respiratory process.

[0028] Furthermore, the CT projections in the respiratory stable period are subjected to heartbeat gating processing according to the heartbeat signal, and the method for obtaining the CT projection groupings in different heartbeat stages is as follows:

[0029] 31) Calculate the detection threshold k=(f2 / f1) by scanning the heart rate f1 of the target to be detected by CT and the acquisition frequency f2 of CT projection data; detect the CT projection in the i-th respiratory stable period stable[i], and obtain the position valley[0] of the first valley in the respiratory stable period stable[i] as the initial detection position;

[0030] 32) Detect backward from the current detection position;

[0031] 321) If the current valley position index exceeds the right boundary of the respiratory stable period stable[i], the last value of the valley valley[i] before the position index is changed to the opposite number, and jump to the next respiratory stable period stable[i+1] to start detection;

[0032] 322) If the current valley[i] is negative, the first valley value after valley[i] is detected to be negative will be stored in the data

[0033] Valley group, return to step 32);

[0034] 323) If the current valley[i] is a positive value, then record the distance between the detected valley position index and valley[i].

[0035] When |index-valley[i]-k| is the smallest, store the valley position index into the array valley and return to execute step 32);

[0036] 33) After detecting all CT projections in the stable respiratory period stable[i], the detection ends and the array valley is returned;

[0037] 34) Based on the array valley, the CT projections in the respiratory stable period stable[i] are divided into equal time intervals to obtain CT projection groups in different heartbeat stages.

[0038] Furthermore, an overlapping grouping method is used to divide the CT projections into multiple groups, wherein the CT projections of adjacent groups partially overlap.

[0039] Furthermore, the selected principal components are the first two principal components p1 and p2 among the obtained principal components.

[0040] Furthermore, according to the CT projection data after respiratory gating processing and heartbeat gating processing, analytical imaging, iterative imaging or filtered back projection method is used to reconstruct the 4D gated imaging of the target to be detected.

[0041] Furthermore, the CT projection data of each group are reconstructed respectively, and then the reconstruction results of each group are spliced ​​to generate a 4D dynamic image that changes with the motion.

[0042] The advantages of the present invention are as follows:

[0043] This method adopts the small-interval PCA method to reduce the computational time cost while improving the accuracy of motion information extraction and CT projection grouping, thereby achieving accurate real-time grouping with universal applicability and high precision.

[0044] (1) It can be executed simultaneously with CT acquisition and has real-time performance;

[0045] (2) Effectively improve the accuracy of CT data-driven gating;

[0046] (3) The method is universal. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 Flow chart for real-time 4D gated imaging.

[0048] Figure 2 is w corresponding to the first principal component ik ;

[0049] (a) only contains breathing motion; (b) contains breathing and heartbeat motion.

[0050] Figure 3 The projection image before and after pixel merging and w ik Comparison chart;

[0051] (a) is before the merger; (b) is after the merger.

[0052] Figure 4 is the w of the overall PCA and the small-area PCA ik Comparison chart;

[0053] (a) is the motion waveform of the corresponding interval selected from the overall PCA; (b) is the PCA of a small interval, and the red dot on the waveform marks the actual ventricular contraction period.

[0054] Figure 5 Bandpass filtering for analog data;

[0055] (a) is a simulation diagram; (b) is a heartbeat signal diagram, and (c) is a breathing signal diagram; among them, blue is the original signal and orange is the filtered signal.

[0056] Figure 6 A comparison chart of periodic step-by-step respiratory gating and equal data volume amplitude gating methods;

[0057] (a) is the equal data amount amplitude gating method; (b) is this method.

[0058] Figure 7 This is a 4D cardiac coronary image based on the periodic equal time interval gating method. DETAILED DESCRIPTION

[0059] The present invention is further described in detail below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention but not to limit the scope of the present invention.

[0060] The present invention's real-time 4D gated imaging method for CT based on projection data can be applied to real-time dynamic reconstruction of CT projections. Figure 1 As shown, each step will be described in detail below.

[0061] Step 1: Data-driven gating

[0062] The core of the real-time 4D gated imaging method of CT based on projection data is to extract data-driven information based on the PCA method, and use small angle interval grouping processing, which is real-time and more accurate than the traditional overall processing method. For example, 2000 CT projections are collected in a single circle, and every 200 CT projections are processed as a group according to the angle, and then the principal component information is extracted from each group of images as motion information based on the PCA method.

[0063] The PCA method is a commonly used data dimensionality reduction method. This method uses orthogonal transformation to convert observation data represented by linearly dependent variables into a few linearly independent variables, which are called principal components. These principal components are also called basis vectors to represent the largest contribution to the change in the data set. During the cardiothoracic CT scan, the CT projection data changes due to respiratory movement and heartbeat movement, so PCA can detect this change and extract motion information.

[0064] The present invention uses a single CT scan data as x i , x i is an n-dimensional vector representing the number of pixels in a single CT projection, i∈[1,m], representing m CT projections. n×m =(x1, x2, ..., x m ) represents the original matrix with pixels as samples and scanning time as features. Follow the steps below to solve its principal components.

[0065] 1. Normalize the original matrix;

[0066] 2. Get the covariance matrix ∑=(X T X) / (m-1)

[0067] 3. Solve for the eigenvector α of the covariance matrix k , and sort by the corresponding eigenvalue;

[0068] 4. Get the principal component p k =Xα k .

[0069] The original data x i It can be approximated by the principal component p k With x i The corresponding weight factor w ik To express it, the formula is as follows:

[0070]

[0071] is the mean of the i-th row in the matrix X. If all principal components are retained, that is, when K in formula (1) is equal to m, the equality holds. Since the principal components are orthogonal, formula (1) can be expressed as:

[0072]

[0073] Calculate w by formula (2) ik corresponds to the weight of the i-th CT projection under the basis of the k-th principal component, such as Figure 2 As shown, the weight factors w corresponding to the first two principal components p1 and p2 are i1 、w i2 It contains most of the motion information, and arranges the weight factors in the order of CT projection acquisition to obtain the breathing and heartbeat motion waveforms.

[0074] Pixel merging is a commonly used technique in image processing. It reduces image size, facilitates faster processing and analysis, and also reduces image noise. Figure 3 As shown, the present invention therefore considers performing data preprocessing pixel merging processing on the CT projection before performing data-driven gating.

[0075] After pixel merging, PCA-based data-driven gating extraction is performed. While completely retaining the motion information of breathing and heartbeat, it also significantly suppresses noise, which is more conducive to the subsequent processing of gating technology. Experimental verification shows that when 2062*252 data is pixel-merged, the motion information can be accurately retained after the pixel merging of up to 7*7. Since the data volume is reduced by 49 times, the operation time is reduced by 84.65% with the data volume of 400 CT projections.

[0076] The advantage of PCA with small angle interval grouping is that it can, to a certain extent, exclude the movement in CT projection caused by the change of angle during the rotation of CT scanning, which is not affected by breathing and heartbeat.

[0077] Depend on Figure 4It can be seen that the w obtained by PCA in small angle interval grouping is ik The accuracy is significantly improved while the calculation time is greatly shortened. However, due to the existence of boundaries between different grouping intervals, boundary information may be lost. The more groups there are, the more serious the loss is. Therefore, an overlapping grouping method (overlap) is adopted. The present invention allows each group to contain 10% of the data at the end of the previous group, solving the boundary problem with only a small calculation cost.

[0078] Step 2: Bandpass filtering

[0079] The chest CT projection is affected by both breathing and heartbeat movements, so the extracted motion waveform is a superposition signal of breathing and heartbeat movements. However, since the influence of breathing movement on the projection is much greater than that of heartbeat movement, the breathing amplitude in the extracted superposition signal is also much greater than the heartbeat amplitude, so the heartbeat signal needs to be separated before performing cardiac gating.

[0080] The Hamming window is a commonly used window function with good frequency response characteristics. It can achieve a smoother transition area in the frequency domain and can effectively suppress the sidelobe response of the filter.

[0081]

[0082] The present invention uses a bidirectional Hamming window FIR filter to perform bandpass filtering to separate the respiratory and heartbeat signals. The frequencies of the respiratory and heartbeat signals are determined by converting the motion signal into the frequency domain by Fourier transform. Figure 5 shown.

[0083] Step 3: Respiratory gating / cardiac gating

[0084] After the data-driven gating acquires the motion signal, the gated grouping technology needs to divide the CT data into different groups according to the corresponding motion signal, so that the CT data in the group corresponding to the same motion signal belongs to the same motion state, and the CT data set in the same group corresponds to the range of negligible motion differences in space, and then the data of each group are reconstructed separately to obtain a 4D dynamic image that changes with motion. In data grouping, the gating processing for breathing can mainly adopt an amplitude-based method; in the gating processing of the heart, the time-based method is mainly adopted. The gated grouping technology that performs better in respiratory motion is the amplitude-based division method. This is because respiratory motion is more likely to have an unstable state during the actual acquisition process, such as changes in the size of the respiratory amplitude and changes in the length of the respiratory cycle. These changes will cause the problem of data divided into the same group not matching in actual space when using the time-based division method.

[0085] In the amplitude-based gated grouping technology, the global equal-amplitude gating method only needs to determine the upper and lower boundaries, as well as the number of groups, and is relatively simple to implement. At the same time, it ensures that the CT data in each group is uniform in spatial position. However, due to the characteristics of respiratory movement, there will be a large difference in the amount of grouped data between the steady respiratory period and the movement period, which also leads to a large difference in the signal-to-noise ratio of the reconstructed image. The global equal-data-volume amplitude gating method is to first perform histogram statistics on the data, and then divide the data volume boundaries according to the data volume information reflected by the histogram and the number of groups. This method can ensure that the amount of CT data in each group is approximately the same, and the signal-to-noise ratio of the reconstructed image is basically uniform, but the data in each group is not uniform in spatial position, and motion blur may occur in the fastest moving interval.

[0086] The present invention adopts a new periodic respiratory gating grouping method - periodic step-by-step respiratory gating based on the physiological characteristics of respiratory movement. Respiratory movement can be divided into a movement period and a stable period. The movement period includes two processes: the expansion of the chest cavity during inspiration and the descent of the diaphragm, and the contraction of the chest cavity during exhalation and the rise of the diaphragm. The stable period is the period from the end of exhalation to the beginning of the next inhalation, when the chest cavity and the diaphragm are in a static state. Therefore, the method adopted by the present invention is to first determine the two groups of data of the stable period and the boundary area of ​​inspiration and exhalation (also a static state), and then group the data in the movement period. The specific algorithm is as follows:

[0087] 1. Determine all peak positions peak (corresponding to the boundary of inhalation and exhalation in each respiratory cycle) through the AMPD peak-finding algorithm, and use the amplitudes corresponding to the positions of peak[i]+1 and peak[i+1]-1 as the boundaries highLeft and hightRight of the i-th interval;

[0088] 2. Based on the position of the peak value peak[i], move p to the right peak value peak[i+1], and based on the position of the peak value peak[i+1], move q to the left, as the left and right boundaries stableLeft and stableRight of the i-th respiratory stable period (p, q are the horizontal coordinate positions of the points in the motion waveform, which also correspond to the number of CT projections; the size of p and q is related to the type of small animal, acquisition frame rate and anesthesia depth);

[0089] 3. Take the area between the two peaks as a cycle (a cycle includes a stable period and a movement period), take max(stableLeft, stableRight) as the lower bound of the respiratory movement period, and take min(highLeft, hightRight)

[0090] As the upper bound of the movement period;

[0091] 4. Divide the respiratory movement period into equal intervals according to the preset number of groups.

[0092] Compared with the global equal amplitude gating method and the global equal data volume amplitude gating method, this method comprehensively considers the uniform reconstruction signal-to-noise ratio (guaranteeing the motion period) and the physiological characteristics of respiratory movement, making the grouping of both the stable period and the motion period more reasonable. And because the respiratory signal obtained by data-driven gating is affected by rotation, there is a baseline drift problem. Compared with the overall amplitude gating method, the separate grouping of each cycle used in this method will not be affected by this, and accurate grouping can be obtained.

[0093] The time-based gating method is more commonly used in cardiac motion. A complete cardiac cycle consists of atrial contraction, ventricular contraction and global diastole. In a short period of time, both the cycle length and the heartbeat pattern have strong stability. However, due to the existence of respiratory movement, the heart morphology changes greatly during the movement of respiratory movement, which will cause serious motion artifacts during reconstruction. Therefore, the present invention adopts a respiratory-heart dual gating method. First, the steady period of respiratory movement is determined by respiratory gating technology. According to the heartbeat signal, the CT projection in the steady period of respiratory movement is subjected to heartbeat gating processing to obtain CT projection grouping corresponding to different heartbeat stages. Compared with the global equal time interval gating method, it can make full use of all data while being more in line with the physiological characteristics of the heartbeat. The specific algorithm is as follows:

[0094] The detection threshold k = (f2 / f1) is calculated by the heart rate (average) f1 of the CT scanned object and the acquisition frequency f2 of the CT data, and the detection is performed in the i-th respiratory stable period stable[i] to detect the position valley[0] of the first valley of the heartbeat movement in the respiratory stable period stable[i];

[0095] 1. Detect peaks or valleys backward from the current position;

[0096] 2. Take the valley as an example to judge:

[0097] (1) If the current valley position index exceeds the right boundary of the current respiratory stable period stable[i], then the current valley[i]

[0098] The last value is changed to the opposite number, and jumps to the next stable breathing period stable[i+1], and returns to execute step 1; index is a number used to record the current valley position (not detected), valley

[0099] The array stores the locations of valleys that have been detected without problems, and valley[i] is the i-th value in the valley array.

[0100] (2) If the current valley[i] is a negative value, the first valley value after valley[i] is detected to be negative is stored in the array valley, and the process returns to step 1.

[0101] (3) If the current valley[i] is a positive value, then record the distance between the detected valley position index and the previous valley[i]. When |index-valley[i]-k| is the smallest, store the valley position index into the array valley and return to execute step 1.

[0102] 3. After all stable period data are detected, the detection ends and the array valley is returned;

[0103] 4. Based on the array valley, the time intervals are divided into periods and equal periods. The range of each stable period starts from the detection of the last negative value to the detection of the next negative value.

[0104] Step 4: Image reconstruction

[0105] Using the respiration and heartbeat gating results in step 3, analytical imaging, iterative imaging or other imaging methods can be used. In this embodiment, a filtered back-projection (FBP) method is used for image reconstruction.

[0106] Example:

[0107] The present invention performs the real-time 4D gated imaging proposed by the present invention based on projection data collected by photon counting small animal spectral CT independently developed by the Institute of High Energy Physics of the Chinese Academy of Sciences.

[0108] Detector related parameters: detector equivalent rows (252), detector equivalent columns (2062), detector pixel size (0.1mm), detector effective field of view (80.91×80.91×7.59mm 3 ).

[0109] Experimental conditions: tube voltage (90 kV), tube current (0.2 mA), frame rate (40 Hz).

[0110] The results of reconstruction using the FBP method after 4D gating using the KM mouse CT projection data are as follows. Figure 7 The results of 4D cardiac gating imaging using the present invention show that the cardiac cycle is divided into 8 groups, and 12.5% ​​of the data is used for image reconstruction, which can clearly show the movement of the myocardium during ventricular contraction and relaxation.

[0111] When the present invention is used to separate the heartbeat and respiratory signals, a motion calibration method can also be used to first extract the respiratory signal, then calibrate the original data, and then extract the heartbeat signal. The calibration method is to treat the respiratory motion as an axial rigid body motion, calibrate the original CT projection data (i.e., move the pixels in the motion area), and then use the same method to extract the heartbeat signal from the calibrated new projection.

[0112] The search for the stable period of respiratory gating can also be determined by the statistical histogram method. By performing histogram statistics on the motion waveform according to the amplitude, the stable period is determined by percentage according to the respiratory type and the proportion of the stable period of different species.

[0113] The reconstruction algorithm may also adopt the SART algorithm, and add a total variation image denoising algorithm (Total Variation, TV) in the reconstruction process.

[0114] Although the specific embodiments of the present invention are disclosed for the purpose of illustration, the purpose is to help understand the content of the present invention and implement it accordingly, those skilled in the art will understand that various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the present invention should not be limited to the content disclosed in the best embodiment, and the scope of the present invention is subject to the scope defined in the claims.

Claims

1. A method for real-time 4D gated imaging of CT based on projection data, comprising the steps of: 1) Acquire CT projections of the target to be detected and divide the CT projections into multiple groups according to acquisition angles; Perform pixel merging on each CT projection in the same angle group, and represent each merged CT projection with an n-dimensional vector, where n is the number of pixels in the merged image; arrange the n-dimensional vectors corresponding to the same angle group according to the scanning time of each CT projection data in the same angle group to construct an original matrix, and then extract the principal component information from the original matrix based on the PCA method, and arrange the weight factors corresponding to the selected principal components according to the CT projection acquisition order to obtain a superposition signal containing breathing and heartbeat; 2) using a bidirectional Hamming window FIR filter to perform bandpass filtering on the superimposed signal to separate the respiratory signal and the heartbeat signal; performing respiratory gating processing on the CT projection according to the respiratory signal to divide the CT projection of the respiratory movement period and the CT projection of the respiratory stable period, and performing gating division on the CT projection of the respiratory movement period to obtain CT projection groups in the same state of the chest during the respiratory process; performing heartbeat gating processing on the CT projection in the respiratory stable period according to the heartbeat signal to obtain CT projection groups corresponding to different heartbeat stages; 3) Reconstruct the CT projection groups after respiratory gating processing and heartbeat gating processing to obtain 4D gated imaging of the target to be detected.

2. The method according to claim 1, characterized in that The method for performing gated grouping on the CT projections during the respiratory movement period to obtain the CT projection groups in the same state of the chest during the respiratory process is: 21) Find all peak positions peak through the AMPD peak-finding algorithm, and determine the boundary between inspiration and expiration in each respiratory cycle according to each peak position peak; wherein, the coordinate positions kpeak[i]+1 and peak[i+1]-1 corresponding to the left and right CT projections of the i-th peak position peak[i] are used as the boundaries highLeft and hightRight of the i-th interval; 22) Taking the ith peak position peak[i] as the reference, move the coordinates corresponding to p CT projections in the direction of the next peak position peak[i+1] as the left boundary of the ith stable respiratory period stableLeft; taking the peak position peak[i+1] as the reference, move the coordinates corresponding to q CT projections in the direction of the peak position peak[i] as the right boundary of the ith stable respiratory period stableRight; 23) Taking the area between two adjacent peak positions peak[i] and peak[i+1] as a cycle, taking the max(stableLeft, stableRight) in the cycle as the lower bound of the respiratory movement period in the cycle, and taking the min(highLeft, hightRight) in the cycle as the upper bound of the respiratory movement period in the cycle; 24) The respiratory movement period in the cycle is divided into equal intervals according to a preset number of groups, so as to obtain CT projection groups in the same state of the chest during the respiratory process.

3. The method according to claim 1, characterized in that The method of performing heartbeat gating processing on the CT projections in the respiratory stable period according to the heartbeat signal to obtain the CT projection grouping in different heartbeat stages is as follows: 31) Calculate the detection threshold k=(f2 / f1) by scanning the heart rate f1 of the target to be detected by CT and the acquisition frequency f2 of CT projection data; detect the CT projection in the i-th respiratory stable period stable[i], and obtain the position valley[0] of the first valley in the respiratory stable period stable[i] as the initial detection position; 32) Detect backward from the current detection position; 321) If the current valley position index exceeds the right boundary of the respiratory stable period stable[i], the last value of the valley valley[i] before the position index is changed to the opposite number, and jump to the next respiratory stable period stable[i+1] to start detection; 322) If the current valley[i] is a negative value, store the first valley value after valley[i] is detected to be a negative value into the array valley, and return to execute step 32); 323) If the current valley[i] is a positive value, then record the distance between the detected valley position index and valley[i]. When |index-valley[i]-k| is the smallest, store the valley position index into the array valley, and return to execute step 32); 33) After detecting all CT projections in the respiratory stable period stable[i], the detection ends and the array valley is returned; 34) Based on the array valley, the CT projections in the respiratory stable period stable[i] are divided into equal time intervals to obtain CT projection groups in different heartbeat stages.

4. The method according to claim 1, 2 or 3, characterized in that: The CT projections are divided into multiple groups using an overlapping grouping method, in which the CT projections of adjacent groups partially overlap.

5. The method according to claim 1, 2 or 3, characterized in that: The selected principal components are the first two principal components p1 and p2 among the obtained principal components.

6. The method according to claim 1, characterized in that According to the CT projection data after respiratory gating processing and heartbeat gating processing, analytical imaging, iterative imaging or filtered back projection method is used to reconstruct the 4D gated imaging of the target to be detected.

7. The method according to claim 1 or 6, characterized in that: The CT projection data of each group are reconstructed separately, and then the reconstruction results of each group are spliced ​​to generate a 4D dynamic image that changes with motion.

Images