CT imaging control method and CT imaging control system

Abstract

The invention provides a CT imaging control method and a CT imaging control system, and relates to the technical field of medical imaging. By introducing the double X-ray sources and the two flat panel detectors and combining with the 3D optical camera to obtain breathing signals of a patient, CT imaging of the patient in a free breathing state is achieved. In the scanning process, based on real-time breathing signals, the two X-ray sources are triggered in sequence at different rotation angles for exposure, and the double-angle projection data obtained by the corresponding flat panel detectors are read. And reconstructing an organ mask image sequence from the double-angle projection data by using a pre-trained deep neural network, further analyzing the sequence to quantify the amplitude and phase of respiratory movement, distributing respiratory phase values for the projection data with different rotation angles, and classifying the respiratory phase values as gating data frames. And performing image reconstruction on each gating data frame to obtain four-dimensional CT images of different breathing phases. The problem of motion artifacts of the CT image in the free breathing state is solved, so that the image quality and the diagnosis accuracy are improved.

Description

Technical Field

[0001] This application relates to the field of medical imaging technology, and more specifically, to a CT imaging control method and a CT imaging control system. Background Technology

[0002] In the field of nuclear medicine imaging, multimodal imaging equipment such as SPECT / CT and PET / CT have become important clinical diagnostic tools. Among them, CT scans not only provide anatomical information, but more importantly, they perform attenuation correction on SPECT or PET data to improve the quantitative accuracy of functional images and achieve precise fusion of anatomical and functional images.

[0003] However, this technology faces an inherent challenge caused by respiratory motion: SPECT and PET signal acquisition times can be several minutes or even more than ten minutes, during which patients typically breathe freely and shallowly. In contrast, routine diagnostic CT scans require completion in the instant the patient holds their breath to obtain clear images free of motion artifacts. This fundamental difference in acquisition modality leads to a clinical dilemma: If a CT scan uses breath-hold mode, it captures a brief, isolated phase of the respiratory cycle. PET / SPECT images, however, are an average superposition of signals throughout the entire respiratory cycle. These two methods represent different physiological states, directly leading to a severe spatial mismatch between anatomical and functional images, especially in areas significantly affected by respiratory movements such as the diaphragm, lungs, and liver. This severely impacts diagnostic accuracy and the precision of radiotherapy target delineation.

[0004] If CT scans also use the same free breathing mode as PET / SPECT, although the breathing state is the same, the CT images themselves will produce motion artifacts such as blurring and stripes due to respiratory movements. This not only reduces the diagnostic value of CT images, but more seriously, attenuation correction and image fusion based on these artifact images will introduce new errors, leading to inaccurate quantitative analysis of functional images and overall distortion of the fused images.

[0005] Therefore, it is necessary to study a CT imaging control method that can simultaneously obtain high-precision CT images without motion artifacts under free breathing conditions, in order to solve the problems of spatial mismatch and artifact introduction between CT and PET / SPECT images under respiratory motion in the existing technology.

[0006] There is currently no effective technical solution to the above problems. Summary of the Invention

[0007] The purpose of this application is to provide a CT imaging control method and a CT imaging control system, which aims to solve the problems of spatial mismatch and artifact introduction between CT and PET / SPECT images under respiratory motion in the prior art.

[0008] In a first aspect, this application provides a CT imaging control method based on a CT imaging control system. The CT imaging control system includes two X-ray sources and two flat panel detectors, with the scanning centerlines of the two X-ray sources forming a preset angle with each other. Each X-ray source corresponds to one flat panel detector. The CT imaging control method includes the following steps: A1. Acquire the patient's respiratory signals using a 3D optical camera; A2. According to the reconstruction mode, the CT imaging control system is controlled to perform rotational scanning. During the scanning process, based on the respiratory signal, two X-ray sources are triggered to expose at different rotation angles, and the projection data obtained by the corresponding flat panel detector is read to obtain dual-angle projection data. The reconstruction modes include energy spectrum reconstruction mode and standard reconstruction mode. A3. Based on dual-angle projection data acquired at different rotation angles, an organ mask image sequence is reconstructed using a pre-trained deep neural network; A4. Analyze the organ mask image sequence to obtain the amplitude and phase of the quantified respiratory motion, and assign respiratory phase values ​​to the projection data at different rotation angles. At the same time, group the projection data corresponding to the same respiratory phase value to obtain the corresponding gated data frame. A5. Perform image reconstruction on each gated data frame to obtain four-dimensional CT images of different respiratory phases.

[0009] Through this technical solution, this application can acquire dual-angle projection data through dual X-ray sources and CT imaging control system under the condition of free breathing of patients, and reconstruct four-dimensional CT images of different breathing phases by combining deep learning and respiratory gating technology. This effectively solves the problems of motion artifacts and spatial mismatch with PET / SPECT images in traditional CT imaging under free breathing conditions, thereby improving the diagnostic accuracy of multimodal imaging and the precision of treatment planning.

[0010] Optionally, in the energy spectrum reconstruction mode, step A2 includes: The CT imaging control system is controlled to perform a rotational scan. During the scan, the two X-ray sources operate at the first tube voltage and the second tube voltage, respectively. Based on the breathing signal, the CT imaging control system is controlled to perform a 360-degree rotational scan in less than 15 seconds. The two X-ray sources are triggered to expose at different rotation angles, and the corresponding flat panel detectors are read to obtain the first dual-angle projection data with two different energies.

[0011] Through this technical solution, this application can efficiently acquire dual-angle projection data of different energies by rapidly rotating scanning and exposing X-ray sources with different tube voltages in the energy spectrum reconstruction mode, providing a foundation for subsequent energy spectrum image reconstruction, thereby obtaining richer information on material composition while ensuring imaging speed.

[0012] Optionally, in the standard reconstruction mode, step A2 includes: Two X-ray sources operate with the same third tube voltage. Based on the respiratory signal, the CT imaging control system is controlled to perform a 360-degree rotation scan in less than 15 seconds. The two X-ray sources are triggered to expose at different rotation angles, and the corresponding flat panel detectors are read to obtain two second dual-angle projection data with the same energy.

[0013] Through this technical solution, this application can quickly acquire high-quality conventional CT projection data in standard reconstruction mode by using rapid scanning and X-ray source exposure with the same tube voltage, providing an efficient and reliable data source for subsequent anatomical image reconstruction.

[0014] Optionally, in step A3, the input of the pre-trained deep neural network is dual-angle projection data, and the output is an organ mask image, where the pixel values ​​in the organ mask image are predefined organ index values.

[0015] Through this technical solution, this application utilizes deep neural networks to directly reconstruct organ mask images from dual-angle projection data, achieving accurate identification and quantification of organ motion. This provides accurate motion information for subsequent respiratory gating and four-dimensional image reconstruction, significantly improving the automation and intelligence level of image reconstruction.

[0016] Optionally, in the energy spectrum reconstruction mode, each of the first dual-angle projection data includes low-energy projection data and high-energy projection data. The low-energy projection data is obtained by one of the X-ray sources operating at a first tube voltage, and the high-energy projection data is obtained by another X-ray source operating at a second tube voltage. Step A4 includes: By statistically analyzing the boundary changes of lung or diaphragm organs in organ mask image sequences, the amplitude of respiratory motion is determined. A complete respiratory cycle is divided into a preset number of phases, and respiratory phase values ​​are assigned to low-energy projection data and high-energy projection data corresponding to different rotation angles. Low-energy projection data with the same phase value are grouped together, and high-energy projection data with the same phase value are grouped together to obtain the corresponding low-energy gated data frames and high-energy gated data frames.

[0017] Through this technical solution, in the energy spectrum reconstruction mode, this application can accurately quantify respiratory motion based on organ mask image sequences, and assign and gate respiratory phases to projection data of different energies respectively, thereby providing refined raw data classified by respiratory phase for subsequent four-dimensional energy spectrum material image reconstruction, effectively avoiding the interference of respiratory motion on energy spectrum analysis.

[0018] Optionally, in energy spectrum reconstruction mode, step A5 includes: All low-energy gated data frames for each phase are parsed and reconstructed to obtain the low-energy reconstructed image of the corresponding phase; All high-energy gated data frames for each phase are parsed and reconstructed to obtain the high-energy reconstructed image of the corresponding phase. AI algorithms were applied to all low-energy and high-energy reconstructed images for noise reduction and artifact removal. Then, a pixel-by-pixel linear merging method was used to obtain four-dimensional energy spectrum material images of different substances.

[0019] Optionally, in the standard reconstruction mode, step A4 includes: By statistically analyzing the boundary changes of lung or diaphragm organs in organ mask image sequences, the amplitude of respiratory motion is determined. A complete respiratory cycle is divided into a preset number of phases, and respiratory phase values ​​are assigned to all second dual-angle projection data corresponding to different rotation angles. All second dual-angle projection data corresponding to the same phase value are grouped together to obtain a standard gated data frame.

[0020] Optionally, in the standard reconstruction mode, step A5 includes: Each standard gated data frame is analyzed and reconstructed using a filtered back projection method or an iterative reconstruction algorithm. The reconstructed image is then processed with an AI algorithm for noise reduction and artifact removal to obtain a four-dimensional energy spectrum anatomical image.

[0021] Optionally, in step A2, the time difference between the exposure of the two X-ray sources is not less than the afterglow time of the corresponding flat panel detector.

[0022] Secondly, this application also provides a CT imaging control system, which includes a controller, a gantry, two X-ray sources, two flat panel detectors and multiple 3D optical cameras. The two X-ray sources and two flat panel detectors are mounted on the gantry and the scanning center lines of the two X-ray sources are at a preset angle to each other. Each X-ray source corresponds to one flat panel detector. The X-ray source is used to emit X-rays and the flat panel detector is used to receive the projection data of the corresponding X-rays. Multiple 3D optical cameras are positioned in multiple directions above the patient to acquire information about the entire upper surface of the patient's body. The controller is mounted on the gantry and is used to control the rotation of the gantry, the operating status of the two X-ray sources and multiple 3D optical cameras, receive all projection data from the two flat panel detectors and data from all the 3D optical cameras, and execute the steps in the CT imaging control method described in any of the preceding descriptions.

[0023] As described above, the CT imaging control method and control system provided in this application achieve CT imaging under the patient's free breathing state by introducing dual X-ray sources and two flat panel detectors, combined with a 3D optical camera to acquire the patient's respiratory signals. During the scanning process, based on the real-time respiratory signal, the two X-ray sources are triggered sequentially at different rotation angles, and the dual-angle projection data acquired by the corresponding flat panel detectors are read. Subsequently, a pre-trained deep neural network is used to reconstruct the organ mask image sequence from these dual-angle projection data, and then the sequence is analyzed to quantify the amplitude and phase of respiratory motion, assigning respiratory phase values ​​to the projection data at different rotation angles, and classifying them into gated data frames. Finally, image reconstruction is performed on each gated data frame to obtain four-dimensional CT images with different respiratory phases. This solves the motion artifact problem in CT images under free breathing state, thereby improving image quality and diagnostic accuracy.

[0024] Other features and advantages of this application will be set forth in the following description and will be apparent in part from the description or may be learned by practicing embodiments of this application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Attached Figure Description

[0025] Figure 1 This is a flowchart of a CT imaging control method provided in an embodiment of this application.

[0026] Figure 2 An organ mask image of the lungs provided for an embodiment of this application.

[0027] Figure 3 A sequence of organ mask images of the lungs during a respiratory cycle, provided for an embodiment of this application.

[0028] Figure 4 The breathing curve over time is provided for the embodiments of this application.

[0029] Figure 5 This is a schematic diagram of the structure of the CT imaging control system provided in an embodiment of this application.

[0030] Labeling explanation: 110, rack; 120, X-ray source; 130, flat panel detector. Detailed Implementation

[0031] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0032] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0033] In a first aspect, this application provides a CT imaging control method based on a CT imaging control system. The CT imaging control system includes two X-ray sources and two flat panel detectors, with the scanning centerlines of the two X-ray sources forming a preset angle with each other. Each X-ray source corresponds to one flat panel detector. The CT imaging control method includes the following steps: A1. Acquire the patient's respiratory signals using a 3D optical camera; A2. According to the reconstruction mode, the CT imaging control system is controlled to perform rotational scanning. During the scanning process, based on the respiratory signal, two X-ray sources are triggered to expose at different rotation angles, and the projection data obtained by the corresponding flat panel detector is read to obtain dual-angle projection data. The reconstruction modes include energy spectrum reconstruction mode and standard reconstruction mode. A3. Based on dual-angle projection data acquired at different rotation angles, an organ mask image sequence is reconstructed using a pre-trained deep neural network; A4. Analyze the organ mask image sequence to obtain the amplitude and phase of the quantified respiratory motion, and assign respiratory phase values ​​to the projection data at different rotation angles. At the same time, group the projection data corresponding to the same respiratory phase value to obtain the corresponding gated data frame. A5. Perform image reconstruction on each gated data frame to obtain four-dimensional CT images of different respiratory phases.

[0034] This application introduces dual X-ray sources, two flat-panel detectors, and multiple 3D optical cameras, combined with respiratory signals to synchronously trigger exposure, and utilizes deep neural networks for organ mask reconstruction and respiratory motion analysis, ultimately achieving four-dimensional CT image reconstruction under respiratory gating. This enables the acquisition of high-precision, motion artifact-free CT images under free breathing conditions, thereby solving the problem of severe spatial mismatch between anatomical and functional images, improving diagnostic accuracy and the precision of radiotherapy target delineation.

[0035] The core of the CT imaging control method proposed in this application lies in solving the motion artifact problem in CT images under free breathing conditions, thereby improving image quality and diagnostic accuracy. This method is based on a CT imaging control system, which typically includes a gantry, two X-ray sources, two flat panel detectors, and multiple 3D optical cameras. The two X-ray sources and two flat panel detectors are mounted on the gantry, with the scanning centerlines of the two X-ray sources forming a preset angle with each other. Each X-ray source corresponds to one flat panel detector; the X-ray source emits X-rays, and the flat panel detector receives the projection data of the corresponding X-ray. Multiple 3D optical cameras are positioned along multiple directions of the gantry to acquire information about the entire upper surface of the patient. A controller, mounted on the gantry, controls the rotation of the gantry, the operating status of the two X-ray sources and the multiple 3D optical cameras, and receives all projection data from the two flat panel detectors and data from all the 3D optical cameras.

[0036] In CT imaging control methods, such as Figure 1 As shown, the first step is to acquire the patient's respiratory signal. This acquisition can be achieved in various ways. For example, a 3D optical camera can be used to acquire point cloud images of the patient's body surface, and respiratory movements can be measured by analyzing changes in the height of surface undulations. Specifically, a 3D optical camera can capture real-time three-dimensional morphological changes in the patient's chest and abdomen, and by calculating the vertical displacement of a specific region (such as the central chest or abdomen), a respiratory curve varying over time can be obtained, such as... Figure 3 As shown in the diagram, the troughs in the respiratory curve typically correspond to the end of expiration (when the lungs are at their lowest and the chest and abdomen are most collapsed), while the peaks correspond to the end of inspiration (when the lungs are at their highest and the chest and abdomen are most expanded). By identifying consecutive troughs and peaks, the start and end times of each respiratory cycle can be determined, and thus the duration of the respiratory cycle can be calculated. For example, a threshold can be set; when the difference between several consecutive respiratory cycles and the average cycle is mostly within this threshold, the patient's breathing is considered stable.

[0037] After acquiring the respiratory signal, the CT imaging control system will control the CT imaging system to perform rotational scanning according to the preset reconstruction mode. During the scan, based on the real-time acquired respiratory signal, two X-ray sources are triggered sequentially at different rotation angles. For example, the CT imaging control system continuously monitors the patient's respiratory signal and acquires the rotation angle of the gantry in real time. When the CT imaging control system rotates to the first preset target angle A (e.g., 0 degrees), it waits for the respiratory signal to enter a preset stable respiratory phase (e.g., end-expiration). Once the respiratory signal meets the gating condition, the first X-ray source is triggered for exposure. After the first X-ray source exposure ends, and after waiting for a preset time difference of not less than the afterglow time of the flat panel detector (>1 ms), the second X-ray source is immediately triggered for exposure. Subsequently, the gantry continues to rotate. During the rotation, the patient continues to breathe. When the CT imaging control system rotates to the second preset target angle B (e.g., 90 degrees), it again waits for the respiratory signal to enter the same (or another preset) stable respiratory phase. Once the respiratory signal meets the gating condition again, the first X-ray source is triggered for exposure. After the first X-ray source exposure ends, and after waiting for a preset time difference of no less than the afterglow time of the flat panel detector (>1 ms), the second X-ray source is immediately triggered for exposure, and the projection data acquired by the corresponding flat panel detector is read, thus obtaining dual-angle projection data. The reconstruction modes here can include spectral reconstruction mode and standard reconstruction mode. In spectral reconstruction mode, the two X-ray sources can operate with different tube voltages, for example, one X-ray source operates with a first tube voltage, and the other X-ray source operates with a second tube voltage, thereby acquiring projection data with different energies. In standard reconstruction mode, the two X-ray sources can operate with the same third tube voltage, acquiring projection data with the same energy. During the scan, to reduce the influence of respiratory motion, the CT imaging control system can be controlled to rotate the scan time by less than 15 seconds in a 360-degree rotation. Furthermore, to avoid crosstalk between X-ray source exposures, the time difference between triggering the two X-ray source exposures should be no less than the afterglow time of the corresponding flat panel detector.

[0038] Next, based on dual-angle projection data acquired at different rotation angles, a pre-trained deep neural network is used to reconstruct the organ mask image sequence. The input to this deep neural network is the dual-angle projection data, and the output is the organ mask image. Pixel values ​​in the organ mask image can be predefined as organ index values; for example, the pixel value for the lung region is 1 (e.g., ...). Figure 2 (The yellow portion in the image shows the pixel value; the heart region has a pixel value of 2, the liver region has a pixel value of 3, etc.) This deep neural network can be trained using a large amount of CT projection data and corresponding organ mask images, enabling it to accurately identify and segment different organs from the projection data.

[0039] Subsequently, the organ mask image sequence needs to be analyzed to quantify the amplitude and phase of respiratory movements. The amplitude of respiratory movements can be determined by statistically analyzing the boundary changes of specific organs (such as the lungs or diaphragm) in the organ mask image sequence. For example, the positional changes of the lungs or diaphragm at different time points can be tracked, and the maximum amplitude (peak inspiratory phase) and minimum amplitude (peak expiratory phase) of the respiratory signal can be determined based on the amplitude of respiratory movements. This amplitude range is then divided into several equal intervals; for example, the amplitude range of the respiratory signal can be divided into 10 intervals, each interval corresponding to a respiratory phase, such as... Figure 4 As shown. Then, appropriate phase intervals are assigned to the projection data at different rotation angles. Based on the acquisition timestamp corresponding to each projection data, and according to the established respiratory motion curve, it is determined which pre-divided phase interval the respiratory state of the projection data belongs to. For example, if it is in the 0%-10% interval of the respiratory cycle, its phase value is marked as 1; if it is in the 10%-20% interval, it is marked as 2, and so on, until the tenth interval (90%-100%) is marked as 10. Finally, the projection data corresponding to the same respiratory phase value are grouped together (that is, all projection data with an assigned phase value of 1 are grouped together) to obtain the corresponding gated data frame.

[0040] Finally, image reconstruction is performed on each gated data frame to obtain four-dimensional CT images at different respiratory phases. For each gated data frame, either filtered backprojection or iterative reconstruction algorithms can be used for image reconstruction. Filtered backprojection is a classic CT image reconstruction algorithm that reconstructs images by filtering and backprojecting the projection data. Iterative reconstruction algorithms, on the other hand, gradually approximate the true image through an iterative optimization process, typically achieving better image quality, especially under low-dose or sparse sampling conditions. By reconstructing gated data frames at different respiratory phases, a series of CT images at different respiratory phases can be obtained. These images collectively constitute a four-dimensional CT image, clearly demonstrating the movement of organs during the respiratory cycle.

[0041] Therefore, the method of this application can obtain high-precision CT images without motion artifacts under conditions of free breathing by the patient. The CT imaging control system shortens the scanning time and reduces the generation of motion artifacts; deep learning technology enables precise organ segmentation and improves the accuracy of respiratory motion analysis; respiratory gating technology effectively eliminates the impact of respiratory motion on image quality. The resulting four-dimensional CT images not only have high resolution but also clearly show the dynamic changes of organs throughout the entire respiratory cycle, thus solving the problems of spatial mismatch and artifact introduction in traditional CT and PET / SPECT images under respiratory motion, significantly improving the accuracy of diagnosis and the precision of target delineation in radiotherapy.

[0042] Among them, the four-dimensional energy spectrum material image that resolves different substances is a specific type of "four-dimensional CT image with different respiratory phases" obtained in the energy spectrum reconstruction mode, while the four-dimensional energy spectrum anatomical image is the anatomical image in the "four-dimensional CT image with different respiratory phases" obtained in the standard reconstruction mode. That is, the four-dimensional energy spectrum material image and the four-dimensional energy spectrum anatomical image together constitute the four-dimensional CT image.

[0043] In some implementations, in the energy spectrum reconstruction mode, step A2 includes: The CT imaging control system is controlled to perform a rotational scan. During the scan, the two X-ray sources operate at the first tube voltage and the second tube voltage, respectively. Based on the breathing signal, the CT imaging control system is controlled to perform a 360-degree rotational scan in less than 15 seconds. The two X-ray sources are triggered to expose at different rotation angles, and the corresponding flat panel detectors are read to obtain the first dual-angle projection data with two different energies.

[0044] Specifically, in energy spectrum reconstruction mode, the CT imaging control system is controlled to perform rotational scanning to acquire the patient's projection data. Two X-ray sources are configured to operate at a first tube voltage and a second tube voltage, respectively. The first and second tube voltages are two different tube voltage values; for example, the first tube voltage can be set to 80-100 kV, while the second tube voltage can be set to 120-140 kV. In this way, projection data of different energies can be acquired simultaneously in a single scan. Furthermore, based on the patient's respiratory signals, the CT imaging control system controls the scan time to within 15 seconds during 360-degree rotational scanning. This scan time limit aims to "freeze" the patient's respiratory movements as much as possible, thereby reducing image artifacts caused by respiratory motion. During the scan, the two X-ray sources are triggered sequentially at different rotation angles. "Sequential triggering" here means that the two X-ray sources are exposed sequentially within a very short time interval to ensure that dual-angle projection data is acquired at almost the same anatomical location. Subsequently, the corresponding flat panel detector reads the acquired projection data, which constitutes the first dual-angle projection data at two different energies. During the 360-degree rotation scan, the maximum number of sampling angles can be no less than 150, and the exposure time for each exposure can be 1 millisecond.

[0045] The proposed solution significantly shortens the data acquisition cycle by limiting the 360-degree rotation scanning time of the CT imaging control system in spectral reconstruction mode to within 15 seconds. This reduced scanning time allows for more data acquisition within a single respiratory cycle, or a complete scan to be completed in a shorter time, effectively minimizing the impact of respiratory motion on projection data and reducing motion artifacts. Furthermore, by operating the two X-ray sources at their respective tube voltages, projection data of different energies can be acquired simultaneously in a single scan, avoiding the time delays and potential motion artifacts associated with traditional dual-energy CT requiring two scans or rapid kVp switching. This rapid, synchronous dual-energy data acquisition method, combined with a respiratory signal triggering mechanism, ensures that projection data acquired at different rotation angles better reflects the patient's anatomical structure at specific respiratory phases, laying the foundation for subsequent high-quality spectral reconstruction.

[0046] In some implementations, under the standard reconstruction mode, step A2 includes: Two X-ray sources operate with the same third tube voltage. Based on the respiratory signal, the CT imaging control system is controlled to perform a 360-degree rotation scan in less than 15 seconds. The two X-ray sources are triggered to expose at different rotation angles, and the corresponding flat panel detectors are read to obtain two second dual-angle projection data with the same energy.

[0047] Specifically, the two X-ray sources mentioned above are configured to operate at the same third tube voltage in standard reconstruction mode. This third tube voltage can be understood as a tube voltage value commonly used in standard CT imaging, such as 120kVp or 140kVp. Its purpose is to ensure that the acquired projection data has consistent energy characteristics, thereby simplifying subsequent image processing and reconstruction. The rotational scanning based on respiratory signal control of the CT imaging control system refers to the system monitoring the patient's respiratory signal in real time during the scan and coordinating the scanning process accordingly. For example, respiratory signals acquired by a 3D optical camera can be used for synchronous scanning to ensure data acquisition at specific stages of the respiratory cycle. In practical applications, the CT imaging control system is controlled to complete data acquisition within less than 15 seconds of a 360-degree rotational scan. This rapid scanning significantly shortens the total data acquisition time, thereby minimizing artifacts introduced by the patient's respiratory movements during the scan. For example, in a complete 360-degree rotational scan, the maximum number of sampling angles can be no less than 150, and each exposure time can be as short as 1ms to capture instantaneous images. Furthermore, the two X-ray sources are triggered sequentially at different rotation angles. This dual-angle exposure method allows projection data from two different viewpoints to be acquired at each rotation position. Since the two X-ray sources operate with the same third tube voltage, the acquired second dual-angle projection data has the same energy.

[0048] The solution presented in this application ensures energy consistency in the acquired second dual-angle projection data by operating two X-ray sources with the same third tube voltage in standard reconstruction mode. This is crucial for subsequent image reconstruction and processing, avoiding complex corrections caused by energy differences. Simultaneously, the CT imaging control system, based on respiratory signals, completes a 360-degree rotation scan within a short time (e.g., less than 15 seconds), significantly shortening the data acquisition window and effectively reducing the impact of respiratory motion on the projection data, thus minimizing motion artifacts. Furthermore, triggering the two X-ray sources sequentially at different rotation angles allows for the acquisition of projection data from two slightly different angles, providing richer information for subsequent image reconstruction and contributing to improved image quality and reduced artifacts. It is precisely due to these synergistic effects that the solution presented in this application can obtain high-quality anatomical CT images under free breathing conditions.

[0049] In some implementations, in step A3, the input to the pre-trained deep neural network is dual-angle projection data, and the output is an organ mask image, where the pixel values ​​in the organ mask image are predefined organ index values.

[0050] The pre-trained deep neural network is configured to receive dual-angle projection data as its input. This dual-angle projection data is generated during rotational scanning in the CT imaging control system, based on respiratory signals, by triggering exposures from two X-ray sources at different rotation angles, and acquired by corresponding flat-panel detectors. The output of the deep neural network is an organ mask image, which reflects the contours and location information of organs within the patient's body. Specifically, each pixel value in the organ mask image is set as a predefined organ index value. For example, the index value of the heart can be defined as 1, the liver as 2, the spleen as 3, the lungs as 4, the kidneys as 5, the bladder as 6, the ribs as 7, the spine as 8, and other bones as 9. In this way, the organ mask image not only provides geometric information of the organs but also directly includes organ category information.

[0051] The solution in this application uses dual-angle projection data directly as input to a pre-trained deep neural network, and the network outputs organ mask images with predefined organ index values. For example, taking the lungs as an example, a sequence of organ (lung) mask images for one respiratory cycle, such as... Figure 4 As shown, this method effectively addresses the challenge of accurately extracting organ mask images from raw projection data in traditional methods. During training, the deep neural network learns the complex mapping relationship from projection data to organ mask images, enabling it to directly identify and segment different organs from dual-angle projection data. Pixel values ​​in the organ mask image are set to predefined organ index values, ensuring that each pixel carries the category information of its respective organ. This greatly simplifies the subsequent extraction and analysis of organ contour and location information, ensuring that the organ mask image effectively reflects the true contour and location information of the organ.

[0052] The above technical solution avoids the complex preprocessing and feature extraction steps of traditional image segmentation methods, significantly improving the efficiency and accuracy of organ segmentation. The pixel values ​​of the organ mask image are predefined organ index values, allowing direct access to organ category information, which facilitates subsequent respiratory motion analysis and gating data frame generation. This enables more accurate determination of organ boundary changes, and more precise quantification of the amplitude and phase of respiratory motion, providing a more reliable data foundation for subsequent four-dimensional CT image reconstruction, ultimately contributing to higher-quality four-dimensional CT images.

[0053] In some implementations, in the energy spectrum reconstruction mode, each of the first dual-angle projection data includes low-energy projection data and high-energy projection data, wherein the low-energy projection data is obtained by one of the X-ray sources operating at a first tube voltage, and the high-energy projection data is obtained by another X-ray source operating at a second tube voltage. Step A4 includes: By statistically analyzing the boundary changes of lung or diaphragm organs in organ mask image sequences, the amplitude of respiratory motion is determined. A complete respiratory cycle is divided into a preset number of phases, and respiratory phase values ​​are assigned to low-energy projection data and high-energy projection data corresponding to different rotation angles. Low-energy projection data with the same phase value are grouped together, and high-energy projection data with the same phase value are grouped together to obtain the corresponding low-energy gated data frames and high-energy gated data frames.

[0054] Specifically, in the energy spectrum reconstruction mode, the first dual-angle projection data refers to the projection data of two different energies acquired by the corresponding flat panel detector when the two X-ray sources operate at the first tube voltage and the second tube voltage, respectively, in step A2. The low-energy projection data can be understood as the projection data acquired by one X-ray source operating at a lower first tube voltage, while the high-energy projection data is the projection data acquired by the other X-ray source operating at a higher second tube voltage. This distinction ensures the different energy information required for subsequent energy spectrum reconstruction.

[0055] Furthermore, in step A4, to accurately match these projection data of different energies with the patient's respiratory motion state, a statistical analysis is first performed on the organ mask image sequence, focusing, for example, on the boundary changes of organs such as the lungs or diaphragm. The organ mask image sequence is reconstructed based on dual-angle projection data using a pre-trained deep neural network, which clearly reflects the morphology and positional information of the organs. By analyzing the changes in these organ boundaries over time, the amplitude of respiratory motion, i.e., the maximum displacement range of the organ during the respiratory cycle, can be accurately determined.

[0056] Based on this, a complete respiratory cycle is divided into a preset number of phases. For example, a respiratory cycle can be divided into 5, 8, or 10 discrete phases, each representing a specific state of respiratory motion. Subsequently, corresponding respiratory phase values ​​are assigned to low-energy projection data and high-energy projection data acquired at different rotation angles. This means that for low-energy projection data and high-energy projection data acquired at a specific rotation angle, a precise respiratory phase value is assigned based on the respiratory state at the time of acquisition.

[0057] Therefore, all low-energy projection data corresponding to the same phase value are grouped together to form a corresponding low-energy gated data frame. Similarly, all high-energy projection data corresponding to the same phase value are grouped together to form a corresponding high-energy gated data frame. In this way, it is ensured that the data in each gated data frame corresponds to the patient's projection information at the same respiratory phase, thus providing a highly consistent data foundation with minimized motion artifacts for subsequent energy spectrum image reconstruction.

[0058] This application's solution effectively solves the problem of inaccurate matching between different energy projection data and respiratory phases by independently assigning and gating respiratory phases to low-energy and high-energy projection data in energy spectrum reconstruction mode. Specifically, firstly, by utilizing two X-ray sources operating at the first and second tube voltages respectively, low-energy and high-energy projection data with different energy characteristics can be acquired simultaneously. Secondly, by statistically analyzing the boundary changes of lung or diaphragm organs in organ mask image sequences, the amplitude and phase of respiratory motion can be accurately quantified, providing a reliable basis for subsequent refined gating. It is precisely because a complete respiratory cycle is divided into a preset number of phases, and respiratory phase values ​​are assigned to low-energy and high-energy projection data corresponding to different rotation angles, that projection data of different energies can correspond to their respective precise respiratory states. Ultimately, by grouping low-energy projection data corresponding to the same phase value into a low-energy gated data frame and high-energy projection data corresponding to the same phase value into a high-energy gated data frame, the data within each gated data frame is highly consistent in terms of respiratory state, thus providing high-quality, low-artifact input data for subsequent energy spectrum image reconstruction.

[0059] Through the above technical solution, this application can significantly improve the image quality and accuracy of spectral CT imaging in the presence of respiratory motion. Compared with uniform or coarse respiratory phase processing of different energy projection data, this application, through independent, refined respiratory phase allocation and gating based on organ mask image sequence analysis of low-energy and high-energy projection data, can more accurately match projection data of different energies with the patient's actual respiratory state. This effectively reduces misalignment and artifacts between different energy projection data caused by respiratory motion. Especially in spectral reconstruction, this accurate phase matching is crucial for subsequent material separation, quantitative analysis, and lesion identification. By ensuring that the data within each gated data frame corresponds to a consistent respiratory phase, the final reconstructed four-dimensional spectral material image will have higher spatial resolution and lower motion artifacts, thus providing a more reliable basis for clinical diagnosis and treatment.

[0060] In some implementations, in the energy spectrum reconstruction mode, step A5 includes: All low-energy gated data frames for each phase are parsed and reconstructed to obtain the low-energy reconstructed image of the corresponding phase; All high-energy gated data frames for each phase are parsed and reconstructed to obtain the high-energy reconstructed image of the corresponding phase. AI algorithms were applied to all low-energy and high-energy reconstructed images for noise reduction and artifact removal. Then, a pixel-by-pixel linear merging method was used to obtain four-dimensional energy spectrum material images of different substances.

[0061] Specifically, in the aforementioned energy spectrum reconstruction mode, after acquiring low-energy gated data frames and high-energy gated data frames for different respiratory phases, all low-energy gated data frames for each phase are first analyzed and reconstructed to obtain the low-energy reconstructed image corresponding to that respiratory phase. Simultaneously, all high-energy gated data frames for each phase are analyzed and reconstructed to obtain the high-energy reconstructed image corresponding to that respiratory phase. The analytical reconstruction method can employ, for example, filtered back projection (FBP) or iterative reconstruction algorithms.

[0062] Furthermore, to improve the quality of the reconstructed images, AI algorithms were applied to all low-energy and high-energy reconstructed images for noise reduction and artifact removal. These AI algorithms can be understood as deep learning-based image processing models, such as Convolutional Neural Networks (CNNs) and Generative Adversarial Networks (GANs), whose purpose is to identify and eliminate noise points and artifacts caused by breathing movements, metal implants, etc., thereby improving image clarity and realism.

[0063] Based on this, a pixel-by-pixel linear merging method is used to fuse the low-energy reconstructed image and the high-energy reconstructed image after noise reduction and artifact removal processing to obtain a four-dimensional energy spectrum image of different substances. Specifically, the pixel-by-pixel linear merging method refers to linearly superimposing each pixel in the image, and its calculation formula is expressed as: HU mix =w*HU 低能 +(1-w)*HU 高能 Among them, HU 低能 HU represents a low-energy reconstructed image. 高能 Represents a high-energy reconstructed image, HU mix represents the pixel value of the merged image, and w represents the weighting coefficient, which is determined by the values ​​of the first and second tube voltages of the two X-ray sources during operation and the composition of the substance to be analyzed. Specifically, the attenuation characteristic coefficients (a1, b1, a2, b2) of two sets of basic substances (e.g., iodine and water) under the first and second tube voltages are obtained in advance through calibration. For any pixel in the image, its low-energy CT value (HU) 低能 ) and high-energy CT value (HU) 高能The density (ρ_A, ρ_B) of the two basic substances in that pixel forms a system of two linear equations. HU 低能 =a1*ρ_A+b1*ρ_B;HU 高能 =a²*ρ_A + b²*ρ_B. By solving this system of equations, the density of various basic substances in a pixel can be calculated. Having the density of the substances, to generate a "virtual flat scan image" (i.e., removing iodine and retaining only the tissue background), the density of HU... mix =HU_VNC (virtual flat scan image), where the virtual flat scan image HU_VNC corresponds to retaining only the contribution of the basic substance B (water) and eliminating the contribution of substance A (iodine). Then HU_VNC = (b1·ρ_B + b2·ρ_B) / 2. For example, HU... 低能 =200, HU 高能 =120, HU_VNC=80, then w * 200 + (1-w) *120 = 80, solving the equation gives w = -0.5. Therefore, by adjusting the weighting coefficient w, energy spectrum CT images highlighting different material components, such as iodine maps, water maps, or virtual plain scans, can be analyzed and generated. This merging method effectively fuses information from different energy images, thereby more accurately identifying different material components in tissues or lesions.

[0064] This application's solution effectively addresses the problem of noise and artifacts introduced by directly reconstructing projection data in energy spectrum reconstruction mode, thus affecting the accuracy of material analysis. This is achieved through a combination of phase-by-phase reconstruction, AI algorithm denoising and artifact removal, and pixel-by-pixel linear merging. First, by reconstructing low-energy and high-energy gated data frames separately for each phase, artifacts caused by respiratory motion between different phases are effectively reduced, ensuring high internal consistency within each phase. Second, applying AI algorithms to denoise and remove artifacts from the reconstructed low-energy and high-energy images further improves the signal-to-noise ratio and clarity, eliminating various interferences introduced by the acquisition process or the reconstruction algorithm itself, providing high-quality input images for subsequent material analysis. Finally, the pixel-by-pixel linear merging method accurately resolves the distribution information of different substances based on the pixel values ​​of images from different energies through weighted fusion, thereby obtaining a four-dimensional energy spectrum material image with high material resolution.

[0065] Through the aforementioned technical solutions, this application significantly improves the quality of energy-spectral CT images and the accuracy of material analysis. Phase-splitting reconstruction effectively suppresses respiratory motion artifacts, while the introduction of AI algorithms further optimizes image noise reduction and artifact removal, resulting in clearer and more realistic reconstructed images. Therefore, combined with a pixel-by-pixel linear merging method, different material components in tissues can be more accurately identified and quantified, such as distinguishing between water, fat, and calcium. This has significant clinical implications for early tumor diagnosis, qualitative analysis of lesions, and evaluation of treatment efficacy. Compared to traditional methods that directly reconstruct and decompose materials, the solution presented in this application provides higher-quality four-dimensional energy-spectral material images, thereby offering physicians richer and more reliable diagnostic information.

[0066] In some implementations, in the standard reconstruction mode, step A4 includes: By statistically analyzing the boundary changes of lung or diaphragm organs in organ mask image sequences, the amplitude of respiratory motion is determined. A complete respiratory cycle is divided into a preset number of phases, and respiratory phase values ​​are assigned to all second dual-angle projection data corresponding to different rotation angles. All second dual-angle projection data corresponding to the same phase value are grouped together to obtain a standard gated data frame.

[0067] Specifically, in the standard reconstruction mode, the execution method of A4 above is further optimized. The amplitude of respiratory motion is determined by statistically analyzing the boundary changes of the lungs or diaphragm organs in the organ mask image sequence. This method can accurately capture the actual range of respiratory motion. A complete respiratory cycle is divided into a preset number of phases, for example, 5 to 10 phases, to meticulously reflect each stage of respiratory motion. Subsequently, respiratory phase values ​​are assigned to all second dual-angle projection data corresponding to different rotation angles, these phase values ​​corresponding to the previously divided respiratory cycle phases. Finally, all second dual-angle projection data corresponding to the same phase value are grouped together, thus obtaining a standard gated data frame. This gating process aims to aggregate projection data under the same respiratory state, providing a consistent data foundation for subsequent image reconstruction.

[0068] This application's solution accurately identifies and quantifies the boundary changes of the lungs or diaphragm organs through statistical analysis of organ mask image sequences, thereby precisely determining the amplitude of respiratory movements. This amplitude determination method based on actual physiological movements avoids potential errors in traditional methods. Furthermore, a complete respiratory cycle is divided into a preset number of phases, allowing the dynamic process of respiratory movements to be captured and described in detail. By assigning corresponding respiratory phase values ​​to all second dual-angle projection data acquired at different rotation angles and grouping them into the same set, standard gated data frames are formed. This gating mechanism ensures that during subsequent image reconstruction, the projection data within each data frame corresponds to the same respiratory phase of the patient, effectively synchronizing the CT scan with the patient's respiratory movements and significantly reducing artifacts caused by respiratory movements.

[0069] The aforementioned technical solution enables precise quantification and allocation of respiratory motion amplitude and phase in standard reconstruction mode. This refined gating process allows for more accurate alignment of projection data with the patient's respiratory state during CT scans performed under free breathing conditions. This significantly reduces the impact of respiratory motion on image quality, minimizes artifacts, and improves the clarity and diagnostic value of reconstructed images. This solution provides high-quality, highly consistent gated data frames for subsequent image reconstruction, resulting in more accurate and reliable four-dimensional CT images.

[0070] In some preferred embodiments, specifically in the standard reconstruction mode, when executing A4 above, the boundaries of the lungs or diaphragm organs are first statistically analyzed using the organ mask image sequence reconstructed from A3 above. For example, the amplitude of respiratory motion is determined by calculating the range of positional changes at different time points. Subsequently, a complete respiratory cycle is divided into, for example, 10 phases. For all second bi-angle projection data acquired at different rotation angles, a respiratory phase value from 1 to 10 is assigned based on the corresponding respiratory signal and the analysis results of the organ mask image sequence at the time of acquisition. For example, if a projection data is acquired during the peak inspiratory phase of the respiratory cycle, it may be assigned phase 1; if it is acquired during the trough in expiration phase, it may be assigned phase 5. Finally, all second bi-angle projection data assigned to the same phase value (e.g., phase 1) are grouped together to form a standard gated data frame, and so on, to obtain multiple standard gated data frames corresponding to different respiratory phases.

[0071] In some implementations, under the standard reconstruction mode, step A5 includes: Each standard gated data frame is analyzed and reconstructed using a filtered back projection method or an iterative reconstruction algorithm. The reconstructed image is then processed with an AI algorithm for noise reduction and artifact removal to obtain a four-dimensional energy spectrum anatomical image.

[0072] Specifically, filtered backprojection is a classic CT image reconstruction algorithm. It reconstructs the image by filtering the projection data and then backprojecting and superimposing it along the projection direction. This method is computationally fast and widely used in clinical practice. Iterative reconstruction algorithms, on the other hand, are reconstruction methods that approximate the real image through repeated iterations. Starting from an initial image, they continuously refine the image to make its projection more consistent with the actual measured projection data. They typically provide higher quality images than filtered backprojection, especially under low-dose or sparse sampling conditions.

[0073] AI algorithms can be understood as image processing models developed based on artificial intelligence technologies such as deep learning. Their purpose is to identify and remove noise and various artifacts in images, such as motion artifacts, metal artifacts, and beam hardening artifacts. By applying such AI algorithms to reconstructed images, the signal-to-noise ratio and contrast of the images can be significantly improved, making anatomical structures clearer.

[0074] In practical applications, four-dimensional energy spectrum anatomical images refer to high-quality image sequences that reflect the anatomical structure of organs at different respiratory phases, obtained through the above processing in standard reconstruction mode. Although the X-ray source operates at the same third tube voltage in standard reconstruction mode, images with high anatomical detail and diagnostic value can still be obtained through fine reconstruction and AI post-processing, providing a more reliable basis for subsequent attenuation correction and image fusion.

[0075] This application's solution first performs analytical reconstruction on standard gated data frames that have been gated according to respiratory phase, ensuring that blurring and artifacts caused by respiratory motion are effectively suppressed in the initial stage of image reconstruction. Subsequently, a pre-trained AI algorithm is applied to these initially reconstructed images for noise reduction and artifact removal. This AI algorithm can learn and identify noise and artifact patterns in the image caused by various factors (such as scattering, hardware imperfections, and residual patient motion), and intelligently remove or suppress them. It is precisely this two-stage processing approach—first obtaining preliminary motion-corrected images through gating and analytical reconstruction, and then using AI algorithms to refine the quality of these images—that results in the final four-dimensional energy dispersive anatomical images with higher clarity, lower noise, and fewer artifacts, effectively solving the problem of insufficient image quality that may exist in traditional methods under standard reconstruction modes.

[0076] The above technical solutions significantly improve the quality of four-dimensional CT images under standard reconstruction mode. Specifically, by employing filtered back-projection or iterative reconstruction algorithms to analyze and reconstruct gated data frames, artifacts caused by respiratory motion can be effectively reduced. Furthermore, applying AI algorithms to the reconstructed images for noise reduction and artifact removal further eliminates noise and residual artifacts, significantly improving the signal-to-noise ratio and contrast, and making anatomical details clearer. Therefore, the solution presented in this application provides high-quality four-dimensional energy spectrum anatomical images, thereby improving diagnostic accuracy, providing clinicians with more reliable diagnostic evidence, and laying a solid foundation for subsequent advanced applications such as attenuation correction and image fusion.

[0077] In some implementations, in step A2, the time difference between the exposure of the two X-ray sources is not less than the afterglow time of the corresponding flat panel detector.

[0078] Specifically, "time difference" refers to the time interval between two X-ray sources triggering exposure at the same rotation angle during rotational scanning in a CT imaging control system. "Afterglow time" can be understood as the time it takes for the signal from an X-ray received and generated by a flat panel detector to decay to below a preset threshold after receiving the X-ray and generating an electrical signal. Afterglow time is an inherent physical characteristic of flat panel detectors, and its length depends on factors such as the detector's materials, structure, and operating conditions.

[0079] The solution proposed in this application strictly controls the triggering sequence of the two X-ray sources, ensuring that the subsequent X-ray source exposure occurs after the afterglow of the corresponding flat panel detector has completely disappeared. Because sufficient time is allowed between the two exposures, the flat panel detector can effectively remove the residual signal generated by the previous exposure, thus avoiding the superposition of the residual signal with the signal of the next exposure and effectively preventing contamination of the projection data.

[0080] The above technical solution can significantly improve the signal-to-noise ratio and accuracy of projection data, effectively avoiding data interference caused by afterglow effects. This results in higher-quality reconstructed images, which is particularly important for dual-source CT systems requiring multiple exposures within a short timeframe. Implementing this solution is crucial for ensuring image quality, thereby improving the overall performance and diagnostic accuracy of CT imaging control methods.

[0081] Secondly, referring to Figure 5This application also provides a CT imaging control system, which includes a controller, a gantry 110, two X-ray sources 120, two flat panel detectors 130, and multiple 3D optical cameras. The two X-ray sources 120 and the two flat panel detectors 130 are arranged on the gantry 110, and the scanning center lines of the two X-ray sources 120 are at a preset angle to each other. Each X-ray source 120 corresponds to one flat panel detector 130. The X-ray source 120 is used to emit X-rays, and the flat panel detector 130 is used to receive the projection data of the corresponding X-rays. Multiple 3D optical cameras are positioned in multiple directions above the patient to acquire information about the entire upper surface of the patient's body. The controller is mounted on the gantry and is used to control the rotation of the gantry, the operating status of the two X-ray sources and multiple 3D optical cameras, receive all projection data from the two flat panel detectors and data from all the 3D optical cameras, and execute the steps in any of the CT imaging control methods described above.

[0082] Specifically, the gantry 110 of the CT imaging control system stably supports and rotates the X-ray source 120 and the flat panel detector 130. The two X-ray sources 120 and two flat panel detectors 130 are precisely mounted on the gantry 110, forming a preset angle, such as 90 degrees or 180 degrees, between their scanning centerlines to optimize the acquisition efficiency and coverage of dual-angle projection data. Each X-ray source 120 is equipped with a corresponding flat panel detector 130, forming two independent X-ray imaging chains for synchronous or alternating acquisition of projection data. The X-ray source 120 can be a conventional X-ray tube, capable of emitting X-rays of different or the same energy according to the controller's instructions. The flat panel detector 130 is a high-sensitivity digital detector used to convert the received X-rays into electrical signals, thereby forming projection data.

[0083] Multiple 3D optical cameras are positioned in multiple directions above the patient. For example, they can be mounted on the outer shell of the gantry or circumferentially above the CT imaging control system. The specific installation location is not limited, and they can be distributed around the patient's body surface to ensure comprehensive and seamless acquisition of all surface information. These 3D optical cameras can be based on stereoscopic vision principles, enabling real-time and high-precision capture of the patient's body surface movements, particularly the surface undulations caused by respiratory movements.

[0084] The controller is the core control unit of the entire CT imaging control system, and it can be a combination of one or more processors, memory, and corresponding control circuits. The controller can precisely control the rotation speed and angle of the gantry 110, coordinate the exposure timing and tube voltage settings of the two X-ray sources 120, and manage the data acquisition of multiple 3D optical cameras. Furthermore, the controller is responsible for receiving and processing all projection data from the two flat panel detectors 130 and body surface information data from all 3D optical cameras. Most importantly, the controller executes all steps of the aforementioned CT imaging control method, including but not limited to acquiring respiratory signals using 3D optical cameras, controlling the CT imaging control system to perform rotational scanning according to the reconstruction mode and triggering X-ray source 120 exposure at different rotation angles, reconstructing organ mask image sequences based on projection data, analyzing respiratory motion and assigning respiratory phase values, and performing image reconstruction on gated data frames to obtain four-dimensional CT images. The principle of the CT imaging control system provided in this embodiment is the same as the principle of the CT imaging control method provided in the first aspect above, and will not be discussed in detail here.

[0085] The CT imaging control system of this application effectively solves the motion artifacts and image mismatch problems existing in traditional CT imaging in free breathing mode by integrating the aforementioned hardware components and CT imaging control methods. Specifically, the arrangement of multiple 3D optical cameras along more than 110 directions of the gantry allows for the accurate and real-time acquisition of all surface information of the patient's upper body, thus providing a comprehensive and reliable data foundation for respiratory signal extraction. Compared with respiratory monitoring of a single or local area, this multi-angle, all-round surface information acquisition method can more accurately reflect the patient's overall respiratory motion state, improving the accuracy and robustness of respiratory signals.

[0086] As the core of the system, the controller not only coordinates the rotation of the gantry 110, the exposure of the X-ray source 120, and the data acquisition of the 3D optical camera, but more importantly, it tightly integrates these hardware operations with the steps in the aforementioned CT imaging control method. By executing step A1 in the method, the controller precisely extracts respiratory signals using the surface information acquired by the 3D optical camera. Subsequently, in step A2, the controller controls the CT imaging control system to perform rotational scanning according to the reconstruction mode (spectral or standard), and based on the real-time respiratory signal, triggers the exposure of the two X-ray sources 120 sequentially at different rotation angles, thereby efficiently acquiring dual-angle projection data in a short time. This configuration of dual X-ray sources 120 and dual flat-panel detectors 130, combined with precise exposure timing control, can minimize the impact of respiratory motion on single projection data while ensuring data acquisition efficiency.

[0087] Furthermore, in step A3, the controller uses a pre-trained deep neural network to process the dual-angle projection data, reconstructing an organ mask image sequence. In step A4, this sequence is analyzed to quantify the amplitude and phase of respiratory motion, assigning respiratory phase values ​​to the projection data, and generating gated data frames. Finally, in step A5, the controller performs image reconstruction on each gated data frame to obtain four-dimensional CT images with different respiratory phases. The entire process, through precise scheduling and algorithm execution by the controller, achieves end-to-end respiratory motion management from respiratory signal acquisition to final image reconstruction, ensuring high-quality, motion artifact-free CT images are obtained even when the patient is breathing freely.

[0088] Through the aforementioned CT imaging control system, this application can significantly improve the image quality and diagnostic accuracy of CT imaging under free breathing conditions. Specifically, the introduction of multiple 3D optical cameras makes the acquisition of respiratory signals more comprehensive and accurate, overcoming the local monitoring errors that may exist in traditional respiratory gating technology, thus providing a more reliable input for subsequent respiratory phase allocation and image reconstruction. As a result, the system can more effectively reduce artifacts caused by respiratory motion and obtain high-precision, motion artifact-free anatomical CT images.

[0089] Furthermore, the controller, as the core intelligent unit, seamlessly integrates advanced CT imaging control methods with the hardware system, achieving precise control over the entire imaging process. This integration not only optimizes the working efficiency of the dual X-ray sources 120 and dual flat panel detectors 130, but more importantly, it makes it possible to complete dual-angle projection data acquisition in a short time (e.g., 360-degree rotation scanning time less than 15 seconds). This is of great significance for reducing patient breath-holding time, improving patient comfort, and rapidly obtaining diagnostic information in emergency situations. By applying AI algorithms to the reconstructed images for noise reduction and artifact removal, the signal-to-noise ratio and clarity of the images are further improved. Therefore, the CT imaging control system of this application can provide more accurate imaging information for clinical diagnosis, especially in multimodal imaging, effectively solving the spatial mismatch problem between CT images and PET / SPECT images, thereby improving diagnostic accuracy and the precision of radiotherapy target delineation, and has significant clinical application value.

[0090] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0091] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A CT imaging control method, based on a CT imaging control system, characterized in that, The CT imaging control system includes two X-ray sources and two flat panel detectors, with the scanning centerlines of the two X-ray sources forming a preset angle with each other. Each X-ray source corresponds to one flat panel detector. The CT imaging control method includes the following steps: A1. Acquire the patient's respiratory signals using a 3D optical camera; A2. According to the reconstruction mode, the CT imaging control system is controlled to perform rotational scanning. During the scanning process, based on the respiratory signal, two X-ray sources are triggered to expose at different rotation angles, and the projection data obtained by the corresponding flat panel detector is read to obtain dual-angle projection data. The reconstruction modes include energy spectrum reconstruction mode and standard reconstruction mode. A3. Based on dual-angle projection data acquired at different rotation angles, an organ mask image sequence is reconstructed using a pre-trained deep neural network; A4. Analyze the organ mask image sequence to obtain the amplitude and phase of the quantified respiratory motion, and assign respiratory phase values ​​to the projection data at different rotation angles. At the same time, group the projection data corresponding to the same respiratory phase value to obtain the corresponding gated data frame. A5. Perform image reconstruction on each gated data frame to obtain four-dimensional CT images of different respiratory phases.

2. The CT imaging control method according to claim 1, characterized in that, In the energy spectrum reconstruction mode, step A2 includes: The CT imaging control system is controlled to perform a rotational scan. During the scan, the two X-ray sources operate at the first tube voltage and the second tube voltage, respectively. Based on the breathing signal, the CT imaging control system is controlled to perform a 360-degree rotational scan in less than 15 seconds. The two X-ray sources are triggered to expose at different rotation angles, and the corresponding flat panel detectors are read to obtain the first dual-angle projection data with two different energies.

3. The CT imaging control method according to claim 1, characterized in that, In the standard reconstruction mode, step A2 includes: Two X-ray sources operate with the same third tube voltage. Based on the respiratory signal, the CT imaging control system is controlled to perform a 360-degree rotation scan in less than 15 seconds. The two X-ray sources are triggered to expose at different rotation angles, and the corresponding flat panel detectors are read to obtain two second dual-angle projection data with the same energy.

4. The CT imaging control method according to claim 1, characterized in that, In step A3, the input of the pre-trained deep neural network is dual-angle projection data, and the output is a sequence of organ mask images, where the pixel values ​​in the organ mask images are predefined organ index values.

5. The CT imaging control method according to claim 2, characterized in that, In the energy spectrum reconstruction mode, each first dual-angle projection data includes low-energy projection data and high-energy projection data. The low-energy projection data is obtained by one of the X-ray sources operating at a first tube voltage, and the high-energy projection data is obtained by another X-ray source operating at a second tube voltage. Step A4 includes: By statistically analyzing the boundary changes of lung or diaphragm organs in organ mask image sequences, the amplitude of respiratory motion is determined. A complete respiratory cycle is divided into a preset number of phases, and respiratory phase values ​​are assigned to low-energy projection data and high-energy projection data corresponding to different rotation angles. Low-energy projection data with the same phase value are grouped together, and high-energy projection data with the same phase value are grouped together to obtain the corresponding low-energy gated data frames and high-energy gated data frames.

6. The CT imaging control method according to claim 5, characterized in that, In the energy spectrum reconstruction mode, step A5 includes: All low-energy gated data frames for each phase are parsed and reconstructed to obtain the low-energy reconstructed image of the corresponding phase; All high-energy gated data frames for each phase are parsed and reconstructed to obtain the high-energy reconstructed image of the corresponding phase. AI algorithms were applied to all low-energy and high-energy reconstructed images for noise reduction and artifact removal. Then, a pixel-by-pixel linear merging method was used to obtain four-dimensional energy spectrum material images of different substances.

7. The CT imaging control method according to claim 3, characterized in that, In the standard reconstruction mode, step A4 includes: By statistically analyzing the boundary changes of lung or diaphragm organs in organ mask image sequences, the amplitude of respiratory motion is determined. A complete respiratory cycle is divided into a preset number of phases, and respiratory phase values ​​are assigned to all second dual-angle projection data corresponding to different rotation angles. All second dual-angle projection data corresponding to the same phase value are grouped together to obtain a standard gated data frame.

8. The CT imaging control method according to claim 7, characterized in that, In the standard reconstruction mode, step A5 includes: Each standard gated data frame is analyzed and reconstructed using a filtered back projection method or an iterative reconstruction algorithm. The reconstructed image is then processed with an AI algorithm for noise reduction and artifact removal to obtain a four-dimensional energy spectrum anatomical image.

9. The CT imaging control method according to claim 1, characterized in that, In step A2, the time difference between the exposure of the two X-ray sources is not less than the afterglow time of the corresponding flat panel detector.

10. A CT imaging control system, characterized in that, The CT imaging control system includes a controller, a gantry, two X-ray sources, two flat panel detectors, and multiple 3D optical cameras. The two X-ray sources and two flat panel detectors are mounted on the gantry, and the scanning center lines of the two X-ray sources are at a preset angle to each other. Each X-ray source corresponds to one flat panel detector. The X-ray source is used to emit X-rays, and the flat panel detector is used to receive the projection data of the corresponding X-rays. Multiple 3D optical cameras are positioned in multiple directions above the patient to acquire information about the entire upper surface of the patient's body. The controller is mounted on the gantry and is used to control the rotation of the gantry, the operating status of the two X-ray sources and multiple 3D optical cameras, receive all projection data from the two flat panel detectors and data from all the 3D optical cameras, and execute the steps in the CT imaging control method as described in any one of claims 1-9.

Images