Method and apparatus for obtaining truncated partial prediction image

Abstract

The present application provides a method and apparatus for obtaining a predicted image of a truncated part, an imaging method and system, and a non-transitory computer-readable storage medium. The method for obtaining the predicted image of the truncated part includes preprocessing projection data to reconstruct an initial image of the truncated part, and calibrating the initial image based on a trained learning network to obtain the predicted image of the truncated part.

Description

Technical Field

[0001] The present invention relates to image processing, and in particular, to a method and apparatus for obtaining a predicted image of a truncated part, an imaging method and system, and a non-transitory computer-readable storage medium. Background Art

[0002] During the process of Computed Tomography (CT), a detector is used to collect data of X-rays that have passed through the object to be detected, and then the collected X-ray data is processed to obtain projection data. These projection data can be used to reconstruct CT images. Complete projection data can reconstruct accurate CT images for diagnosis.

[0003] However, if the object to be detected is large in size or in a special pose, then some parts of the object to be detected will exceed the scanning field, and the detector will not be able to collect complete projection data, which is called data truncation. Generally, some mathematical models, such as water models, etc., can be used to predict the projection data or image of the truncated part. However, the image quality of the truncated part predicted by these traditional methods will change with different actual situations, and the performance is not ideal enough. Summary of the Invention

[0004] The present invention provides a method and apparatus for obtaining a predicted image of a truncated part, an imaging method and system, and a non-transitory computer-readable storage medium.

[0005] An exemplary embodiment of the present invention provides a method for obtaining a predicted image of a truncated part, the method including preprocessing projection data to reconstruct an initial image of the truncated part; and calibrating the initial image based on a trained learning network to obtain a predicted image of the truncated part.

[0006] Optionally, the preprocessing of the projection data includes filling the truncated part of the projection data. Further, filling the truncated part of the projection data includes filling the projection data information of the boundary of the non-truncated part into the truncated part. Further still, filling the projection data information of the boundary of the non-truncated part into the truncated part includes filling the projection data information of the boundary of the non-truncated part of each channel into the truncated part of the corresponding channel.

[0007] Optionally, calibrating the initial image based on a trained learning network to obtain a predicted image of the truncated part includes converting the pixels of the initial image of the truncated part from polar coordinates to rectangular coordinates to obtain a pixel matrix of the initial image; calibrating the pixel matrix based on a trained learning network; and converting the calibrated pixel matrix from rectangular coordinates to polar coordinates to obtain a predicted image of the truncated part.

[0008] Optionally, the trained learning network is trained based on virtual distorted images and reference images. Further, the trained learning network is trained based on the pixel matrix obtained by coordinate transformation of the virtual distorted image and the pixel matrix obtained by coordinate transformation of the reference image. Further, the method for obtaining the virtual distorted image and the reference image includes receiving an original image without data truncation; virtually translating a part of the original image corresponding to the target object to partially move out of the scanning field to obtain a reference image; virtually scanning the reference image and performing virtual data acquisition to generate virtual truncated projection data; and performing image reconstruction on the virtual truncated projection data to obtain a virtual distorted image. Optionally, the method for obtaining the virtual distorted image and the reference image includes: receiving an original image without data truncation, and using the original image as the reference image; keeping the original image within the scanning field and performing forward projection on the original image to obtain the projection of the original image; filling the two-side channels of the projection to generate virtual truncated projection data; and performing image reconstruction on the virtual truncated projection data to obtain a virtual distorted image.

[0009] An exemplary embodiment of the present invention further provides an imaging method, the imaging method includes obtaining a predicted image of the truncated part, and splicing the predicted image of the truncated part with an untruncated part image reconstructed from the original projection data to obtain a medical image, wherein obtaining the predicted image of the truncated part includes preprocessing the projection data to reconstruct an initial image of the truncated part; and calibrating the initial image based on the trained learning network to obtain the predicted image of the truncated part.

[0010] An exemplary embodiment of the present invention further provides a non-transitory computer-readable storage medium, which is used to store a computer program, and when the computer program is executed by a computer, it causes the computer to execute the instructions of the above method for obtaining an image of a truncated part.

[0011] An exemplary embodiment of the present invention further provides a device for obtaining a prediction of an image of a truncated part, the device includes a preprocessing device and a control device. The preprocessing device is used to preprocess the projection data to reconstruct an initial image of the truncated part. The control device is used to calibrate the initial image based on the trained learning network to obtain the predicted image of the truncated part.

[0012] Optionally, the preprocessing device includes a filling module, and the filling module is used to fill the truncated part of the projection data. Further, the filling module is further configured to fill the projection data information at the boundary of the untruncated part into the truncated part.

[0013] Optionally, the control device includes a transformation module, a calibration module, and an inverse transformation module. The transformation module is configured to convert the pixels of the initial image of the truncated part from polar coordinates to rectangular coordinates to obtain a pixel matrix of the initial image. The calibration module is configured to calibrate the pixel matrix based on a trained learning network. The inverse transformation module is configured to convert the calibrated pixel matrix from rectangular coordinates to polar coordinates to obtain a predicted image of the truncated part.

[0014] An exemplary embodiment of the present invention further provides an imaging system, which includes the above-described device for obtaining a predicted image of a truncated part and a splicing device. The device for obtaining a predicted image of a truncated part includes a preprocessing device and a control device. The preprocessing device is configured to preprocess projection data to reconstruct an initial image of the truncated part. The control device is configured to calibrate the initial image based on a trained learning network to obtain a predicted image of the truncated part. The splicing device is configured to splice the predicted image of the truncated part with an untruncated part image reconstructed from the projection data to obtain a medical image.

[0015] Other features and aspects will become apparent through the following detailed description, the drawings, and the claims. Description of the Drawings

[0016] The present invention can be better understood by describing exemplary embodiments of the present invention in conjunction with the drawings. In the drawings:

[0017] Figure 1 A flowchart of an imaging method according to some embodiments of the present invention is shown;

[0018] Figure 2 Shown according to Figure 1 A flowchart of the step of calibrating an initial image based on a trained learning network in the method shown to obtain a predicted image of a truncated part;

[0019] Figure 3 Shown according to Figure 1 Some intermediate images of preprocessing the projection in the method shown, where image (1) shows the projection with data truncation, image (2) shows the simulated projection obtained by filling the truncated part of the projection in (1), and image (3) shows the CT image obtained by reconstructing the simulated projection in (2);

[0020] Figure 4 Shown according to Figure 2Some intermediate images in the steps shown, where image (1) shows the rectangular coordinate pixel matrix corresponding to the truncated part, image (2) shows the predicted pixel matrix obtained by passing the rectangular coordinate transformed image in (1) through the learning network, and image (3) shows the predicted image obtained by performing an inverse coordinate transformation on the predicted pixel matrix in (2);

[0021] Figure 5 Shows a flowchart of a method for preparing learning network training data according to some embodiments of the present invention;

[0022] Figure 6 Shows a flowchart of a method for preparing learning network training data according to other embodiments of the present invention;

[0023] Figure 7 Shows a schematic diagram of a CT system according to some embodiments of the present invention; and

[0024] Figure 8 Shows a schematic diagram of an imaging system including a device for obtaining a predicted image of a truncated part according to some embodiments of the present invention. Detailed implementation manners

[0025] The following will describe the detailed implementation manners of the present invention. It should be noted that in the specific description of these implementation manners, for the sake of concise description, this specification may not describe all features of the actual implementation manners in detail. It should be understood that in the actual implementation process of any implementation manner, just as in the process of any engineering project or design project, in order to achieve the specific goals of the developer and to meet system-related or business-related limitations, various specific decisions are often made, and these will also change from one implementation manner to another. In addition, it should also be understood that although the efforts made in this development process may be complex and lengthy, for those of ordinary skill in the art related to the content disclosed in the present invention, some design, manufacturing, or production changes based on the technical content disclosed in this disclosure are only conventional technical means and should not be understood as the content of this disclosure being insufficient.

[0026] Unless otherwise defined, technical terms or scientific terms used in the claims and the specification shall have the ordinary meanings understood by those of ordinary skill in the technical field to which the present invention pertains. The terms "first", "second" and similar terms used in the specification and claims of this patent application for invention do not denote any order, quantity or importance, but are only used to distinguish different components. The terms such as "a" or "an" do not denote a quantity limitation, but mean that there is at least one. The terms such as "comprising" or "including" mean that the elements or items appearing before "comprising" or "including" cover the elements or items listed after "comprising" or "including" and their equivalent elements, and do not exclude other elements or items. The terms such as "connected" or "coupled" are not limited to physical or mechanical connections, nor are they limited to direct or indirect connections.

[0027] As used in the present invention, the term "object to be detected" may include any object to be imaged. The terms "projection data" and "projection image" have the same meaning.

[0028] In some embodiments, during a CT scan, when the object to be detected is large in size or in a special posture, it may cause some parts to exceed the scan field of view (SFOV) of the CT, and the collected projection data will be truncated, and the reconstructed image will also be distorted. The method and system for obtaining a predicted image of the truncated part in some embodiments of the present invention can more accurately predict the image of the truncated part based on artificial intelligence, and better provide a basis for doctors' diagnosis and / or treatment. It should be noted that from the perspective of those of ordinary skill in the relevant art or the related art, such a description should not be understood as limiting the present invention only to CT systems. In fact, the method and system for obtaining a predicted image of the truncated part described here can be reasonably applied to other imaging fields in the medical and non-medical fields, such as X-ray systems, PET systems, SPECT systems, MR systems or any combination thereof.

[0029] As discussed herein, artificial intelligence (which includes deep learning techniques, also known as deep machine learning, hierarchical learning, or deep structured learning, etc.) employs artificial neural networks for learning. Deep learning methods are characterized by using one or more network architectures to extract or model a class of data of interest. Deep learning methods can be accomplished using one or more processing layers (e.g., convolutional layers, input layers, output layers, normalization layers, etc., which can have different functional layers according to different deep learning network models), where the configuration and number of layers allow the deep learning network to handle complex information extraction and modeling tasks. Typically, specific parameters of the network (which can also be referred to as "weights" and "biases") are estimated through a so-called learning process (or training process), although in some embodiments, the learning process itself can also be extended to the learning elements of the network architecture. The parameters that are learned or trained typically result in (or output) a network corresponding to different levels of layers, so different aspects of the initial data being extracted or modeled, or the output of the previous layer, can generally represent the hierarchical structure or cascade of the layers. In the process of image processing or reconstruction, this can be characterized as different layers corresponding to different feature levels or resolutions in the data. Thus, the processing can be carried out in layers, i.e., earlier or higher-level layers can correspond to extracting "simple" features from the input data, followed by layers that combine these simple features into features exhibiting higher complexity. In fact, each layer (or more specifically, each "neuron" in each layer) can employ one or more linear and / or non-linear transformations (so-called activation functions) to process the input data into an output data representation. The number of multiple "neurons" can be constant between multiple layers, or can vary from layer to layer.

[0030] As discussed herein, as part of the initial training of a deep learning process to solve a specific problem, a training data set with known input values (e.g., an input image or a pixel matrix of an image undergoing coordinate transformation) and known or expected values can be employed to obtain the final output of the deep learning process (e.g., a target image or a pixel matrix of an image undergoing coordinate transformation) or the individual layers of the deep learning process (assuming a multi-layer network architecture). In this way, the deep learning algorithm can process the known or training data set (in a supervised or guided manner or in an unsupervised or unguided manner) until the mathematical relationship between the initial data and the expected output and / or the mathematical relationship between the input and output of each layer is identified and characterized. The learning process typically utilizes (part of) the input data and creates a network output for that input data. The created output is then compared with the expected (target) output of the data set, and the difference between the generated and expected outputs is then used to iteratively update the parameters (weights and biases) of the network. One such update / learning mechanism uses the Stochastic Gradient Descent (SGD) method to update the parameters of the network, and of course those skilled in the art should understand that other methods known in the art can also be used. Similarly, a separate validation data set can be employed, where both the input and the expected target values are known, but only the initial values are provided to the trained deep learning algorithm, and then the output is compared with the output of the deep learning algorithm to verify the previous training and / or prevent over-training.

[0031] Figure 1 FIG. 4 shows a schematic diagram of an imaging method 100 according to some embodiments of the present invention. Figure 2 Shown according to Figure 1 FIG. 8 is a flowchart of step 112 of calibrating an initial image to obtain a predicted image of a truncated portion based on a training-based learning network in the method 100 shown. Figure 3 Shown according to Figure 1 Some intermediate images of preprocessing projection data in the method shown, wherein image (1) shows a projection with data truncation, image (2) shows a simulated projection obtained by filling the truncated portion of the projection in (1), and image (3) shows a CT image obtained by reconstructing the simulated projection in (2). Figure 4 Shown according to Figure 2 Some intermediate images in the method shown, wherein image (1) shows a rectangular coordinate pixel matrix corresponding to the truncated portion, image (2) shows a predicted pixel matrix obtained by passing the rectangular coordinate pixel matrix in (1) through a learning network, and image (3) shows a predicted image obtained by performing an inverse coordinate transformation on the predicted pixel matrix in (2).

[0032] As Figure 1-4As shown, the imaging method in some embodiments of the present invention includes method 110 for obtaining a predicted image of a truncated part, and method 110 for obtaining a predicted image of a truncated part includes steps 111 and 112.

[0033] In step 111, the projection data is preprocessed to reconstruct an initial image of the truncated part.

[0034] In some embodiments, preprocessing the projection data includes padding the truncated part of the projection data. The projection data (i.e., projection) 210 as shown in image (1) can be obtained by collecting X-ray data passing through the object to be detected. Since part of the object to be detected extends beyond the scanning field of the CT system, data truncation occurs in the obtained projection data 210. The truncated part of the projection data can be simulated through preprocessing of the projection data. Preferably, padding the truncated part of the projection data includes filling the projection data information of the boundary of the untruncated part (the outermost channel) into the truncated part. Further, the projection data information of the boundary of the untruncated part (the outermost channel) of each channel is filled into the truncated part of the corresponding channel, as shown in image (2) to obtain the simulated projection data 220 through padding. Figure 3 Although the method of padding is used to simulate the truncated part in some embodiments of the present invention, those skilled in the art should be aware that the truncated part can be simulated in other ways. For example, the projection data information of the truncated part can be calculated based on the projection data information of the untruncated part according to a mathematical model. In addition, although some embodiments of the present invention use the method of filling the projection data information of the boundary of the untruncated part into the truncated part, those skilled in the art should be aware that the embodiments of the present invention are not limited to this filling method. The truncated part can also be filled in other ways. For example, the truncated part can be entirely or partially filled with a certain specific tissue or a preset CT value (such as 0) or projection data information, etc. Further, if there is data loss in the obtained projection data, the same filling method can also be used to fill the data. In addition, although the preprocessing in some embodiments of the present invention includes padding the truncated part, those skilled in the art should understand that the preprocessing not only includes this step, but may also include any data preprocessing operations performed before image reconstruction. Figure 3 In some embodiments, after preprocessing the projection data, image reconstruction is performed on the padded projection data to obtain a CT image. As shown in image (3), the CT image 230 is

[0035] for

[0036] In some embodiments, after preprocessing the projection data, image reconstruction is performed on the padded projection data to obtain a CT image. As shown in image (3) below, the CT image 230 is Figure 3 for Figure 3The projection data of the image (2) 220 (including the untruncated part and the simulated truncated part) is reconstructed to obtain a CT image 230. The CT image 230 includes an untruncated part image 231 and a truncated part image 232. That is, the part within the scanning field 80 is the truncated part image 231, and the part outside the scanning field 80 is the truncated part image 232. The boundary of the scanning field 80 is the junction 234 between the untruncated part 231 and the truncated part 232. The image reconstruction algorithm may include, for example, the filtered back projection (FBP) reconstruction method, the adaptive statistical iterative reconstruction (ASIR) method, the conjugate gradient (CG) method, the maximum likelihood expectation maximization (MLEM) method, the model-based iterative reconstruction (MBIR) method, etc.

[0037] In some embodiments, step 111 further includes extracting the CT image obtained by image reconstruction to obtain an initial image of the truncated part, such as Figure 3 the part 232 shown in the image (3). In some embodiments, although the initial image of the truncated part in the present invention represents the part image outside the scanning field, however, those skilled in the art should know that image extraction is not limited to only extracting the image of the part within and outside the scanning field, and may also include at least a part of the untruncated part image. In actual operation, it can be adjusted according to needs.

[0038] Please continue to refer to Figure 1 as Figure 1 shown, the method 110 for obtaining a predicted image of the truncated part in some embodiments of the present invention further includes step 112.

[0039] In step 112, based on the trained learning network, the initial image of the truncated part is calibrated to obtain a predicted image of the truncated part.

[0040] By inputting a certain amount of virtual distortion images (known input values) of the truncated part and reference images (desired output values), or pixel matrices corresponding to the virtual distortion images of the truncated part (known input values) and pixel matrices corresponding to the reference images (desired output values), a learning network is constructed or trained based on deep learning methods to obtain the mathematical relationship between the known input values and the desired output values. Based on this, in actual operation, when a known image of the truncated part (such as the initial image 232 of the truncated part mentioned above) is input, the desired image of the truncated part (i.e., the desired output value - reference image) can be obtained based on the learning network. The training, construction, and data preparation of the learning network will be further described in conjunction with Figure 5 and Figure 6 In some embodiments, the training of the learning network in the present invention is based on some well-known network models in the art (such as Unet, etc.).

[0041] Please refer toFigure 2 As shown, in some embodiments, calibrating the initial image of the truncated part based on the trained learning network to obtain the predicted image of the truncated part (step 112) includes step 121, step 122, and step 123.

[0042] In step 121, the pixels of the initial image of the truncated part are converted from polar coordinates to rectangular coordinates to obtain the pixel matrix of the initial image. For example, by performing a coordinate transformation from polar coordinates to rectangular coordinates on the initial image of the truncated part (such as Figure 3 part 232 in image (3) in Figure 4 ), a rectangular coordinate pixel matrix 242 as shown in image (1) in

[0043] In step 122, based on the trained learning network, the above pixel matrix is calibrated. For example, by inputting the pixel matrix 242 in image (1) in Figure 4 into the learning network, a calibrated reference pixel matrix 252 as shown in image (2) in Figure 4 can be obtained.

[0044] In step 123, the calibrated pixel matrix is converted from rectangular coordinates to polar coordinates to obtain the predicted image of the truncated part. For example, by converting the calibrated pixel matrix (such as

[0045] Figure 4 252 shown in image (2) in Figure 4 from rectangular coordinates to polar coordinates, the predicted image 262 of the truncated part (as shown in image (3) in

[0046] Please continue to refer to Figure 1 , as Figure 1 shown, the imaging method of some embodiments of the present invention further includes step 120.

[0047] In step 120, the predicted image of the truncated part is stitched with the untruncated part image reconstructed from the original projection data to obtain a medical image. For example, by stitching or synthesizing with the untruncated part 231 in the CT image 230, a complete CT image 260 can be obtained, where the untruncated part image 231 is reconstructed based on the initial projection data.

[0048] Figure 5 shows a flowchart of a learning network training data preparation method 300 according to some embodiments of the present invention. As Figure 5 shown, the learning network training data preparation method 300 includes step 310, step 320, step 330, and step 340.

[0049] In step 310, obtain the original image without data truncation. In some embodiments, the initial image is reconstructed based on projection data. When all parts of the object to be detected are within the scanning field, there is no data truncation in the obtained projection data.

[0050] In step 320, offset the part of the original image corresponding to the target object to partially move it out of the scanning field, thereby obtaining a reference image. In some embodiments, the reference image represents a complete CT image without truncation.

[0051] In step 330, perform virtual scanning on the reference image and perform virtual data acquisition to generate virtual truncated projection data. In some embodiments, based on a virtual sampling system, the reference image is virtually scanned. Since part of the target object has moved out of the scanning field, this part of the image will not be scanned, which is equivalent to virtual truncation.

[0052] In step 340, perform image reconstruction on the virtual truncated projection data to obtain a virtual distorted image. In some embodiments, the virtual distorted image represents a distorted image with data truncation.

[0053] Optionally, the data preparation method further includes step 350. In step 350, perform coordinate transformation (e.g., from polar coordinates to rectangular coordinates) on the virtual distorted image and the reference image to obtain a virtual distorted pixel matrix and a virtual reference pixel matrix. The above pixel matrices are respectively used as the known input values and the expected output values of the learning network to construct or train the learning network to better obtain the predicted image of the truncated part.

[0054] Figure 6 The flowchart of the training data preparation method of the learning network according to some other embodiments of the present invention is shown. As Figure 6 shown, the training data preparation method 400 of the learning network includes step 410, step 420, step 430, and step 440.

[0055] In step 410, obtain the original image without data truncation, and the original image is used as the reference image.

[0056] In step 420, obtain the projection data corresponding to the original image. In some embodiments, keep the original image (i.e., the reference image) within the scanning field and perform forward projection on the original image to obtain the projection of the original image (i.e., the projection data). In some other embodiments, the original projection data obtained by the initial scan can be directly used as the projection (i.e., the projection data).

[0057] In step 430, padding is performed on both side channels of the projection data to generate virtual truncated projection data. The both side channels include the upper and lower side channels, and may also include the left and right side channels, which depends on the orientation of the projection image. In some embodiments, the padding in step 430 is the same as the padding method for preprocessing the initial image (step 110 described above). Similarly, when other non-padding methods are used for preprocessing the initial image, the same method is also used in step 430 for processing.

[0058] In step 440, image reconstruction is performed on the virtual truncated projection data to obtain a virtual distorted image.

[0059] Optionally, the data preparation method further includes step 450. In step 450, coordinate transformation (e.g., from polar coordinates to rectangular coordinates) is performed on the virtual distorted image and the reference image to obtain a virtual distorted pixel matrix and a virtual reference pixel matrix. The above pixel matrices are respectively used as the known input values and the expected output values of the learning network to construct or train the learning network, so as to better obtain the predicted image of the truncated part.

[0060] Although two embodiments of the training data preparation of the learning network are described above, this does not mean that the present invention can only simulate data truncation in these two ways. Other ways can also be used to simulate data truncation to simultaneously obtain the virtual distorted image (known input value) and the reference image (expected output value) required for AI network learning.

[0061] The method for obtaining the predicted image of the truncated part based on artificial intelligence proposed by the present invention can more accurately predict the projection data and / or image of the part beyond the scanning field (i.e., the truncated part), and more accurately restore the lines and boundaries of the detected object, so as to accurately determine the radiation dose that should be applied to the patient, providing a strong basis for radiotherapy. By padding the truncated part to obtain the initial image of the truncated part (preprocessing step 110) or the virtual truncated projection data of the truncated part (step 430 in data preparation), the learning network can be better and more accurately constructed, which is beneficial to improving the accuracy of predicting the truncated part. In addition, by performing coordinate transformation on the initial image of the truncated part, transforming the initial annular image into a matrix image in rectangular coordinates, the accuracy of the learning network prediction can also be improved, making the predicted image more accurate.

[0062] Figure 7 FIG. shows a schematic diagram of a CT system 10 according to some embodiments of the present invention. As Figure 7As shown, system 10 includes a gantry 12, on which an X-ray source 14 and a detector array 18 are oppositely arranged. The detector array 18 is composed of a plurality of detectors 20 and a data acquisition system (DAS) 26. The DAS 26 is used to convert the sampled analog data of the analog attenuation data received by the plurality of detectors 20 into digital signals for subsequent processing. In some embodiments, the system 10 is used to collect projection data of the object to be detected at different angles. Therefore, the components on the gantry 12 are used to rotate around the rotation center 24 to collect projection data. During rotation, the X-ray radiation source 14 is used to project X-rays 16 that penetrate the object to be detected towards the detector array 18. The attenuated X-ray beam data is used as projection data of the target volume of the object after preprocessing. Based on this projection data, an image of the object to be detected can be reconstructed. The reconstructed image can display the internal features of the object to be detected, and these features include, for example, lesions, sizes, shapes, etc. of the body tissue structure. The rotation center 24 of the gantry also defines the center of the scan field 80.

[0063] The system 10 further includes an image reconstruction module 50. As described above, the DAS 26 samples and digitizes the projection data collected by the plurality of detectors 20. Then, the image reconstruction module 50 performs high-speed image reconstruction based on the above sampled and digitized projection data. In some embodiments, the image reconstruction module 50 stores the reconstructed image in a storage device or a mass storage 46. Alternatively, the image reconstruction module 50 transmits the reconstructed image to a computer 40 to generate patient information for diagnosis and evaluation.

[0064] Although Figure 7 the image reconstruction module 50 is illustrated as a separate entity in [reference], in certain embodiments, the image reconstruction module 50 may form a part of the computer 40. Alternatively, the image reconstruction module 50 may not exist in the system 10, or the computer 40 may execute one or more functions of the image reconstruction module 50. In addition, the image reconstruction module 50 may be located at a local or remote location and may be connected to the system 10 using a wired or wireless network. In some embodiments, the computing resources in the cloud network can be used for the image reconstruction module 50.

[0065] In some embodiments, the system 10 includes a control mechanism 30. The control mechanism 30 may include an X-ray controller 34 for providing power and timing signals to the X-ray radiation source 14. The control mechanism 30 may also include a gantry controller 32 for controlling the rotation speed and / or position of the gantry 12 based on imaging requirements. The control mechanism 30 may also include a carrier bed controller 36 for driving the carrier bed 28 to move to a suitable position to position the object to be detected in the gantry 12 to collect projection data of a target volume of the object to be detected. Further, the carrier bed 28 includes a drive device, and the carrier bed controller 36 can control the carrier bed 28 by controlling the drive device.

[0066] In some embodiments, the system 10 further includes a computer 40, and the data sampled and digitized by the DAS 26 and / or the images reconstructed by the image reconstruction module 50 are transmitted to the computer or computer 40 for processing. In some embodiments, the computer 40 stores the data and / or images in a storage device such as a mass storage 46. The mass storage 46 may include a hard disk drive, a floppy disk drive, a compact disk read / write (CD-R / W) drive, a digital versatile disk (DVD) drive, a flash drive, and / or a solid-state storage device. In some embodiments, the computer 40 transmits the reconstructed image and / or other information to a display 42, which is communicatively connected to the computer 40 and / or the image reconstruction module 50. In some embodiments, the computer 40 can be connected to a local or remote display, printer, workstation, and / or similar device, for example, such equipment can be connected to a medical institution or hospital, or connected to a remote device via one or more configured wires or a wireless network such as the Internet and / or a virtual private network.

[0067] In addition, the computer 40 can provide commands and parameters to the DAS 26, and the control mechanism 30 (including the gantry controller 32, the X-ray controller 34 and the carrier bed controller 36) based on user-provided and / or system-defined commands to control system operations, such as data acquisition and / or processing. In some embodiments, the computer 40 controls system operations based on user input. For example, the computer 40 can receive user input, including commands, scan protocols and / or scan parameters, through an operator console 48 connected thereto. The operator console 48 may include a keyboard (not shown) and / or a touch screen to allow the user to input / select commands, scan protocols and / or scan parameters. Although Figure 7 Only one operator console 48 is shown by way of example, but the computer 40 may be connected to more operator consoles, for example, for inputting or outputting system parameters, requesting medical examinations and / or viewing images.

[0068] In some embodiments, system 10 may include or be connected to a Picture Archiving and Communication System (PACS) (not shown in the figures). In some embodiments, the PACS is further connected to remote systems such as, for example, a Radiology Information System, a Hospital Information System, and / or an internal or external network (not shown) to allow operators located at different locations to provide commands and parameters, and / or access image data.

[0069] The methods or processes described further below may be stored as executable instructions in a non-volatile memory on a computing device of system 10. For example, computer 40 may include executable instructions in non-volatile memory and may apply the methods described herein to automatically perform some or all of the scanning process, such as selecting a suitable protocol, determining suitable parameters, etc. Also, for example, image reconstruction module 50 may include executable instructions in non-volatile memory and may apply the methods described herein to perform image reconstruction tasks.

[0070] Computer 40 may be set up and / or arranged to be used in different ways. For example, in some implementations, a single computer 40 may be used; in other implementations, multiple computers 40 are configured to work together (e.g., based on a distributed processing configuration) or individually, with each computer 40 being configured to handle a particular aspect and / or function, and / or process data for generating models only for a particular medical imaging system 10. In some implementations, computer 40 may be local (e.g., co-located with one or more medical imaging systems 10, such as within the same facility and / or the same local network); in other implementations, computer 40 may be remote and thus accessible only via a remote connection (e.g., via the Internet or other available remote access technologies). In a particular implementation, computer 40 may be configured in a cloud-like manner and may be accessed and / or used in a manner that is substantially similar to the way other cloud-based systems are accessed and used.

[0071] Once data (e.g., a trained learning network) is generated and / or configured, the data can be copied and / or loaded into the medical imaging system 10, which can be done in different ways. For example, the model can be loaded via a directed connection or link between the medical imaging system 10 and the computer 40. In this regard, available wired and / or wireless connections and / or communication between different components can be accomplished according to any suitable communication (and / or network) standard or protocol. Alternatively or additionally, the data can be loaded into the medical imaging system 10 indirectly. For example, the data can be stored on a suitable machine-readable medium (e.g., a flash card, etc.), and then the medium is used to load the data into the medical imaging system 10 (on-site, such as by a user or authorized personnel of the system), or the data can be downloaded to an electronic device capable of local communication (e.g., a laptop computer, etc.), and then the device is used on-site (e.g., by a user or authorized personnel of the system) to upload the data to the medical imaging system 10 via a direct connection (e.g., a USB connector, etc.).

[0072] Figure 8 An imaging system 500 including a device for acquiring a truncated part prediction image according to some embodiments of the present invention is shown. As Figure 8 shown, the imaging system 500 includes a device 501 for acquiring a truncated part prediction image.

[0073] The device 501 for acquiring a truncated part prediction image includes a preprocessing device 510 and a control device 520. The preprocessing device 510 is used to preprocess the projection data to reconstruct an initial image of the truncated part, and the control device 520 is used to calibrate the initial image based on a trained learning network to obtain a prediction image of the truncated part.

[0074] In some embodiments, the preprocessing device 510 includes a filling module (not shown in the figure). The filling module is used to fill the truncated part of the projection data. Further, the filling module is further used to fill the projection data information of the boundary of the non-truncated part (the outermost channel) into the truncated part. Preferably, the projection data information of the boundary of the non-truncated part (the outermost channel) of each channel is filled into the truncated part of the corresponding channel.

[0075] In some embodiments, the preprocessing device 510 further includes an image reconstruction module (not shown in the figure). The image reconstruction module is used to perform image reconstruction on the preprocessed (e.g., filled) projection data. In some embodiments, the image reconstruction module is as Figure 7The image reconstruction module 50 in the CT system 10 shown. In some other embodiments, the preprocessing device 510 and the image reconstruction module 50 of the CT system are connected by wire or wirelessly (including direct connection or indirect connection through a computer, etc.). After the filling module fills the truncated part of the projection data, the filling module (or the preprocessing device 510 or the device 501 for obtaining the predicted image of the truncated part) can send the filled projection data to the image reconstruction module 50. After the image reconstruction module 50 completes the reconstruction, it sends the reconstructed CT image to the image extraction module (or the preprocessing device 510 or the device 501 for obtaining the predicted image of the truncated part).

[0076] Optionally, the preprocessing device 510 further includes an image extraction module (not shown in the figure). The image extraction module is used to extract the reconstructed CT image to obtain the initial image of the truncated part.

[0077] In some embodiments, the control device 520 includes a transformation module 521, a calibration module 522, and an inverse transformation module 523.

[0078] The transformation module 521 is used to convert the pixels of the initial image of the truncated part from polar coordinates to rectangular coordinates to obtain the pixel matrix of the initial image.

[0079] The calibration module 522 is used to calibrate the above pixel matrix based on the trained learning network. In some embodiments, the calibration module 522 and the training module 540 are connected by wire or wirelessly (including direct connection or indirect connection through a computer, etc.). The training module 540 is configured to be based on Figure 5 and Figure 6 the method shown to prepare the virtual distortion image and the reference image in the learning network, and learn or train based on the existing deep learning network model (such as Unet) to construct the learning network. In some embodiments, although Figure 8 the training module 540 in is a separate module, those skilled in the art can understand that the training module 540 can be integrated in the control device, or other modules or devices.

[0080] The inverse transformation module 523 is used to convert the calibrated pixel matrix from rectangular coordinates to polar coordinates to obtain the predicted image of the truncated part.

[0081] The present invention can also provide a non-transitory computer-readable storage medium for storing an instruction set and / or a computer program, which, when executed by a computer, causes the computer to execute the above method for obtaining a predicted image of a truncated part. The computer executing the instruction set and / or the computer program can be the computer of a CT system or other devices / modules of the CT system. In one embodiment, the instruction set and / or the computer program can be programmed in the processor / controller of the computer.

[0082] Specifically, when the instruction set and / or the computer program is executed by a computer, it causes the computer to:

[0083] Preprocess the projection data to reconstruct an initial image of the truncated part; and

[0084] Calibrate the initial image based on a trained learning network to obtain a predicted image of the truncated part.

[0085] The instructions as described above can be combined into one instruction for execution, and any instruction can also be split into multiple instructions for execution. In addition, it is not limited to the above instruction execution order.

[0086] In some embodiments, calibrating the initial image based on a trained learning network to obtain a predicted image of the truncated part may include:

[0087] Convert the pixels of the initial image of the truncated part from polar coordinates to rectangular coordinates to obtain a pixel matrix of the initial image;

[0088] Calibrate the pixel matrix based on a trained learning network; and

[0089] Convert the calibrated pixel matrix from rectangular coordinates to polar coordinates to obtain a predicted image of the truncated part.

[0090] In some embodiments, the instruction set and / or the computer program further causes the computer to execute the training of the learning network. Further, it includes the following instructions:

[0091] Obtain an original image without data truncation;

[0092] Virtual-translate the part of the original image corresponding to the target object to partially move it out of the scanning field to obtain a reference image;

[0093] Perform virtual scanning on the reference image and perform virtual data acquisition to generate virtual truncated projection data; and

[0094] Perform image reconstruction on the virtual truncated projection data to obtain a virtual distorted image.

[0095] Optionally, it further includes:

[0096] Perform coordinate transformation on the virtual distorted image and the reference image to obtain a virtual distorted pixel matrix and a virtual reference pixel matrix.

[0097] In some other embodiments, to cause a computer to perform training of a learning network, it further includes the following instructions:

[0098] Obtain an original image without data truncation, and the original image is used as a reference image;

[0099] Obtain projection data corresponding to the original image;

[0100] Pad both channels of the projection data to generate virtual truncated projection data; and perform image reconstruction on the virtual truncated projection data to obtain a virtual distorted image.

[0101] Optionally, it further includes:

[0102] Perform coordinate transformation on the virtual distorted image and the reference image to obtain a virtual distorted pixel matrix and a virtual reference pixel matrix.

[0103] As used herein, the term "computer" may include any processor - based or microprocessor - based system, including systems using microcontrollers, reduced instruction set computers (RISC), application - specific integrated circuits (ASIC), logic circuits, and any other circuits or processors capable of performing the functions described herein. The above examples are merely exemplary and are thus not intended to limit the definition and / or meaning of the term "computer" in any way.

[0104] The instruction set may include various commands that direct a computer or processor, which is a processing unit, to perform specific operations, such as the methods and processes of various embodiments. The instruction set may be in the form of a software program, and the software program may form part of one or more tangible non - transitory computer - readable media. The software may take various forms, such as system software or application software. In addition, the software may take the form of a collection of independent programs or modules, program modules within a larger program, or a part of a program module. The software may also include modular programming in the form of object - oriented programming. The processing of the input data by the processing unit may be in response to an operator's command, or in response to a previous processing result, or in response to a request made by another processing unit.

[0105] Some exemplary embodiments have been described above. However, it should be understood that various modifications can be made. For example, suitable results can be achieved if the described techniques are performed in a different order and / or if the components in the described systems, architectures, devices, or circuits are combined in a different manner and / or are replaced or supplemented by other components or their equivalents. Accordingly, other embodiments also fall within the scope of the claims.

Claims

1. A method for obtaining a predicted image of a truncated part, comprising: Preprocess the projection data to reconstruct the initial image of the truncated part; and Based on a trained learning network, calibrate the initial image to obtain a predicted image of the truncated part, where calibrating the initial image based on the trained learning network to obtain a predicted image of the truncated part includes: Convert the pixels of the initial image of the truncated part from polar coordinates to rectangular coordinates to obtain the pixel matrix of the initial image; Based on a trained learning network, calibrate the pixel matrix; and Convert the calibrated pixel matrix from rectangular coordinates to polar coordinates to obtain the predicted image of the truncated part.

2. The method according to claim 1, wherein, The preprocessing of the projection data includes filling the truncated part of the projection data.

3. The method according to claim 2, wherein, Filling the truncated part of the projection data includes filling the projection data information at the boundary of the untruncated part into the truncated part.

4. The method according to claim 3, wherein, Filling the projection data information at the boundary of the untruncated part into the truncated part includes filling the projection data information at the boundary of the untruncated part of each channel into the truncated part of the corresponding channel.

5. The method according to claim 1, wherein, The trained learning network is trained based on a virtual distorted image and a reference image.

6. The method according to claim 5, wherein, The trained learning network is trained based on the pixel matrix obtained by coordinate transformation of the virtual distorted image and the pixel matrix obtained by coordinate transformation of the reference image.

7. The method according to claim 5, wherein, The method for obtaining the virtual distorted image and the reference image includes: Obtain an original image without data truncation; Virtual-translate the part of the original image corresponding to the target object to partially move it out of the scanning field to obtain a reference image; Perform virtual scanning on the reference image and perform virtual data acquisition to generate virtual truncated projection data; and Perform image reconstruction on the virtual truncated projection data to obtain a virtual distorted image.

8. The method according to claim 5, wherein, The method for obtaining the virtual distorted image and the reference image includes: Obtain an original image without data truncation, and the original image is used as the reference image; Obtain the projection data corresponding to the original image; Fill the two side channels of the projection data to generate virtual truncated projection data; and Perform image reconstruction on the virtual truncated projection data to obtain a virtual distorted image.

9. An imaging method, comprising: The method for obtaining a predicted image of the truncated part according to any one of claims 1-8; and Stitch the predicted image of the truncated part with the image of the untruncated part reconstructed from the original projection data to obtain a medical image.

10. An apparatus for obtaining a predicted image of a truncated part, comprising: A preprocessing device for preprocessing projection data to reconstruct the initial image of the truncated part; and A control device for calibrating the initial image based on a trained learning network to obtain a predicted image of the truncated part, where the control device includes: A transformation module for converting the pixels of the initial image of the truncated part from polar coordinates to rectangular coordinates to obtain the pixel matrix of the initial image; A calibration module for calibrating the pixel matrix based on a trained learning network; and An inverse transformation module for converting the calibrated pixel matrix from rectangular coordinates to polar coordinates to obtain the predicted image of the truncated part.

11. The apparatus according to claim 10, wherein, The preprocessing device includes a filling module, which is used to fill the truncated part of the projection data.

12. The apparatus according to claim 11, wherein, The filling module is further used to fill the projection data information at the boundary of the untruncated part into the truncated part.

13. An imaging system, comprising: The device for obtaining a predicted image of the truncated part according to any one of claims 10-12; and a splicing device, which is used to splice the predicted image of the truncated part and the image of the untruncated part reconstructed from the projection data to obtain a medical image.

14. A non-transitory computer-readable storage medium for storing a computer program, which when executed by a computer causes the computer to execute the method for obtaining a predicted image of a truncated portion according to any one of claims 1-8.

Images