Medical image processing method and device, computer device and storage medium

Abstract

Embodiments of the present application disclose a medical image processing method and device, computer equipment and a storage medium. The method comprises: generating at least one set of predicted deformation data based on a target motion model, the at least one set of predicted deformation data corresponding to a target time; determining an initial image corresponding to measurement data, the target time being within the collection time of the measurement data; performing deformation processing on the initial image based on the at least one set of predicted deformation data to obtain at least one reconstructed image, and taking the at least one reconstructed image as a reconstructed image sequence; determining predicted data corresponding to the reconstructed image sequence, and an error between the predicted data and the measurement data, and adjusting network parameters of the target motion model in the reverse direction according to the error; returning to the step of generating at least one set of predicted deformation data based on the target motion model until the error meets a set error condition. The technical effect of improving the time resolution of dynamic medical images is achieved.

Description

[0001] This application claims priority to patent application No. 202310458907.1, filed on April 21, 2023, the entire contents of which are incorporated herein by reference. Technical Field

[0002] The embodiments of the present invention relate to the field of medical image processing, and to medical image processing methods, apparatus, computer equipment and storage media. Background Technology

[0003] Cardiac imaging has always been one of the most challenging topics in medical imaging. Studies have shown that the average velocity of the right coronary artery can reach up to 7 cm / s, requiring a temporal resolution of 20 ms and 40 ms for medical imaging systems to achieve motion artifact-free coronary artery imaging and cardiac function imaging, respectively. However, the temporal resolution of the most advanced medical imaging equipment currently available is between 140-175 ms, far below the required 20 ms or 40 ms time resolution.

[0004] To address this, many researchers have attempted to improve the temporal resolution of cardiac functional imaging using various image processing methods. However, these methods all require cardiac images based on multiple temporal phases or even multiple cardiac cycles, resulting in higher radiation doses received by patients during the imaging process. Summary of the Invention

[0005] This invention provides a medical image processing method, apparatus, computer equipment, and storage medium, which solves the problem that existing dynamic medical imaging cannot simultaneously achieve low radiation dose and high temporal resolution.

[0006] In a first aspect, embodiments of the present invention provide a medical image processing method, comprising:

[0007] At least one set of predicted deformation data is generated based on the target motion model, and each set of predicted deformation data corresponds to a target time.

[0008] Determine the initial image corresponding to the measurement data, wherein the target time is within the acquisition time of the measurement data;

[0009] Based on the at least one set of predicted deformation data, the initial image is deformed to obtain at least one reconstructed image, and the at least one reconstructed image is used as a sequence of reconstructed images;

[0010] Determine the predicted data corresponding to the reconstructed image sequence, and the error between the predicted data and the measured data, and adjust the network parameters of the target motion model in reverse according to the error;

[0011] Return to the step of generating at least one set of predicted deformation data based on the target motion model until the error meets the set error condition.

[0012] Secondly, embodiments of the present invention also provide a medical image processing apparatus, comprising:

[0013] The deformation data prediction module is used to generate at least one set of predicted deformation data based on the target motion model, wherein each set of predicted deformation data corresponds to a target time.

[0014] An initial image determination module is used to determine the initial image corresponding to the measurement data, wherein the target time is within the acquisition time of the measurement data;

[0015] The deformation processing module is used to perform deformation processing on the initial image based on the at least one set of predicted deformation data to obtain at least one reconstructed image, and to use the at least one reconstructed image as a sequence of reconstructed images;

[0016] The model network parameter adjustment module is used to determine the predicted data corresponding to the reconstructed image sequence, and the error between the predicted data and the measurement data, and to adjust the network parameters of the target motion model in reverse according to the error.

[0017] An iterative module is used to return to the step of generating at least one set of predicted deformation data based on the target motion model until the error meets the set error condition.

[0018] Thirdly, embodiments of the present invention also provide a computer device, the computer device comprising:

[0019] One or more processors;

[0020] Storage device for storing one or more programs;

[0021] When the one or more programs are executed by the one or more processors, the one or more processors implement the medical image processing method as described in any embodiment.

[0022] Fourthly, embodiments of the present invention also provide a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform the medical image processing method as described in any embodiment.

[0023] Compared to existing technologies, the medical image processing method provided in this embodiment uses measurement data collected by a medical imaging system to reconstruct the initial image. Since the target time is within the acquisition time of the measurement data, the network parameters of the target motion model are adjusted based on the sum of the aforementioned errors. This ensures that the predicted deformation data output by the target model is consistent with the temporal phase of the target moving organ at the target time. In fact, the network parameters of the target motion model are adjusted based on the measurement data, achieving the goal of adjusting the network parameters of the target motion model according to specific circumstances. This improves the flexibility, accuracy, and generalizability of the network parameter settings of the target motion model, ensuring that it outputs a reconstructed image with high image quality when receiving different types of input images. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1A This is a flowchart of the medical image processing method provided in the embodiments of the present invention;

[0026] Figure 1B This is a schematic diagram of the target motion model provided in an embodiment of the present invention;

[0027] Figure 1C This is a flowchart illustrating another medical image processing method provided in an embodiment of the present invention;

[0028] Figure 2A This is a flowchart of the method for generating predicted deformation data provided in an embodiment of the present invention;

[0029] Figure 2B This is a flowchart illustrating another medical image processing method provided in an embodiment of the present invention;

[0030] Figure 3A This is a flowchart illustrating another medical image processing method provided in an embodiment of the present invention;

[0031] Figure 3B This is a flowchart illustrating another medical image processing method provided in an embodiment of the present invention;

[0032] Figure 4 This is a flowchart illustrating another medical image processing method provided in an embodiment of the present invention;

[0033] Figure 5 This is a flowchart of the network parameter adjustment method for the target motion model provided in this embodiment of the invention;

[0034] Figure 6 This is a structural block diagram of the medical image processing device provided in the embodiments of the present invention;

[0035] Figure 7 This is a schematic diagram of the structure of the C-arm CT imaging system provided in an embodiment of the present invention;

[0036] Figure 8A This is a schematic diagram of the structure of the diagnostic CT imaging system provided in an embodiment of the present invention;

[0037] Figure 8B This is a schematic diagram of the structure of another diagnostic CT imaging system provided in an embodiment of the present invention;

[0038] Figure 9 This is a schematic diagram of the structure of the computer device in the CT imaging system provided in an embodiment of the present invention. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention. Obviously, the described embodiments are only some embodiments of this invention, not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0040] Figure 1A This is a flowchart of a medical image processing method provided in an embodiment of the present invention. The technical solution of this embodiment is applicable to improving the temporal resolution of medical images including moving organs. This method can be executed by a medical image processing apparatus provided in this embodiment of the present invention, which can be implemented in software and / or hardware and configured in the processor of a computer device. Figure 1A and Figure 1B As shown, the method specifically includes the following steps:

[0041] S110. Generate at least one set of predicted deformation data based on the target motion model, wherein each set of predicted deformation data corresponds to a target time.

[0042] In one embodiment, the predicted deformation data is a motion vector field.

[0043] The target motion model is a deep learning model that generates at least one set of predicted deformation data for an initial image. In one embodiment, the target motion model is... Figure 1BThe neural network shown has 24 convolutional neural network layers, which include three types of convolutional network layers. The network parameters in these convolutional network layers are all learnable. The first type of convolutional network layer uses 3×3 convolutional kernels with a spacing of 1. Figure 1B The first layer is labeled "Conv, 3×3, S1", followed by the batch normalization (Bnorm) operation and the rectified linear unit (ReLU) activation function. The second type of convolutional network layer uses 3×3 convolutional kernels with a spacing of 2, and... Figure 1B The first layer is labeled "Conv, 3×3, S2", followed by Bnorm and ReLU. The third type of convolutional network layer uses 1×1 convolutional kernels with a spacing of 1, and... Figure 1B The values ​​are labeled "Conv, 1×1, S1", followed by the linear activation function. All convolutional layers have corresponding learnable bias terms. Each convolutional layer maintains the same spatial dimension for its input and output. Sampling layers use 2×2 convolutional kernels and... Figure 1B The sample is labeled as Up-sample 2×2. All sampling layers use bilinear interpolation. (Skip + Concatenate) Figure 1B The solid black arrows (in the center) are used to facilitate the network training process. The convolution kernels in the network parameters are initialized using Glorot uniformly distributed random numbers, and the bias term is initialized with 0. All other network parameter settings and initialization values ​​are set to their default values.

[0044] S120. Determine the initial image corresponding to the measurement data, with the target time being within the acquisition time of the measurement data.

[0045] The initial images are clinical images that include musculoskeletal organs. These organs can be the heart, lungs, abdomen, etc.

[0046] The measurement data is acquired by a clinical imaging system. This system can be a CT (Computer Tomography) system or an MRI (Magnetic Resonance Imaging) system. The CT system can be a diagnostic CT system, a C-arm CT system, etc.

[0047] Regarding the target time within the acquisition time of the measurement data corresponding to the initial image. For example, if the acquisition time corresponding to the initial image is [time A, time B], then at least one set of target times is distributed within [time A, time B].

[0048] In one embodiment, after the measurement data is determined, image reconstruction is performed on the measurement data of the target object to obtain an initial image. The target object is the patient.

[0049] Figure 1C This is a schematic flowchart of another medical image processing method provided by an embodiment of the present invention. For ease of explanation, in this drawing, the measurement data can only be used to reconstruct one initial image. It is understood that, if the measurement data can be used to reconstruct at least two initial images, the measurement data can be divided into at least two groups of partial measurement data, and additional processing can be performed on each group of partial measurement data. Figure 1C This can be achieved using medical image processing methods. Specifically, each measurement data segment can reconstruct an initial image; that is, each measurement data segment is equivalent to an attached image. Figure 1C Measurement data.

[0050] S130. Based on the at least one set of predicted deformation data, the initial image is subjected to deformation processing to obtain at least one reconstructed image, and the at least one reconstructed image is used as a sequence of reconstructed images.

[0051] Among them, deformation processing refers to performing at least one of the deformation processing methods, including translation, rotation, scaling, etc., on each pixel position of the image.

[0052] The initial image is deformed based on the predicted deformation data to obtain a reconstructed image of the target organ in the target temporal phase. Specifically,

[0053] Since each predicted deformation data corresponds to a different target time, the initial image can be deformed into a reconstructed image corresponding to a specific time by performing deformation processing based on the predicted deformation data, so as to achieve the purpose of capturing the temporal state of the moving organ at different target times.

[0054] In one embodiment, an existing deformation processing method is used to deform the initial image based on each predicted deformation data to obtain the corresponding reconstructed image.

[0055] It is understood that, since the reconstructed image sequence in this embodiment is determined based on an initial image, the number of reconstructed images included in the reconstructed image sequence is equal to the temporal resolution ratio, which is the ratio of the temporal resolution of the reconstructed image to the temporal resolution of the initial image.

[0056] S140. Determine the error between the predicted data and the measured data corresponding to the reconstructed image sequence, and adjust the network parameters of the target motion model in reverse according to the error.

[0057] After the predicted data corresponding to the reconstructed image sequence is determined, the error between the predicted data and the measurement data corresponding to the initial image is determined. Based on the error, the network parameters of the target motion model are adjusted in reverse so that some or all of the predicted deformation data in at least one set of predicted deformation data output by the target motion model after the network parameters are adjusted are changed.

[0058] In this embodiment, the initial image is reconstructed based on measurement data, and the predicted data is determined based on the reconstructed image sequence. Since the measurement data is directly acquired and accurate scan data, the error between the predicted data and the measurement data can accurately reflect the magnitude of the error in the reconstructed image sequence, thereby reflecting the accuracy of the reconstructed image sequence. Therefore, by adjusting the network parameters of the target motion model, at least one set of predicted deformation data can be output, thereby adjusting the reconstructed image after deforming the initial image based on any predicted deformation data.

[0059] S150, Return to the step of generating at least one set of predicted deformation data based on the target motion model until the error meets the set error condition.

[0060] If at least one new set of predicted deformation data is detected from the target motion model output, return to step S110 until the error meets the set error condition.

[0061] Compared to the original image, the reconstructed image sequence has a higher temporal resolution, thus providing richer and more accurate information on organ motion.

[0062] In one embodiment, the error condition is an error threshold, which can be configured to be modified. That is, the user can adjust the size of the error threshold within a set adjustable range according to actual needs.

[0063] Since the temporal resolution of the reconstructed image sequence is higher than that of the initial image, and the reconstructed image sequence can be used for clinical diagnosis, embodiments of this disclosure allow users to increase the frequency range of measurement data by reducing the temporal resolution of the initial image when acquiring MRI measurement data. Specifically, the temporal resolution of the MRI image used for clinical diagnosis is determined and used as the target temporal resolution, i.e., the temporal resolution of the reconstructed image sequence; based on the ratio of the target temporal resolution to the temporal resolution improvement corresponding to the target motion model, the temporal resolution of the initial image is determined and used as the initial temporal resolution, at which the MRI measurement data of the subject is acquired.

[0064] In one embodiment, the target temporal resolution is set to be equal to the temporal resolution of existing clinical diagnostic MRI images, and the temporal resolution improvement ratio is 5. Then, the initial temporal resolution is 1 / 6 of the target temporal resolution. Compared to the prior art, when acquiring MRI measurement data of a subject based on this initial temporal resolution, the acquisition time allocated to the measurement data used to reconstruct any initial image is 6 times that of existing data acquisition time. Therefore, it allows the user to increase the frequency range of the measurement data by increasing the acquisition time for the measurement data corresponding to the reconstruction of a single initial image. It is understood that for MRI images, measurement data with a larger frequency range corresponds to an initial image with higher spatial resolution. The spatial resolution of the reconstructed image is the same as the spatial resolution of the initial image. That is, the embodiments of this disclosure can indirectly improve the spatial resolution of MRI images without reducing the temporal resolution of clinical diagnostic MRI images.

[0065] Compared to existing technologies, the medical image processing method provided in this embodiment uses measurement data collected by a medical imaging system to reconstruct the initial image. Since the target time is within the acquisition time of the measurement data, the network parameters of the target motion model are adjusted based on the sum of the aforementioned errors. This ensures that the predicted deformation data output by the target model is consistent with the temporal phase of the target moving organ at the target time. In fact, the network parameters of the target motion model are adjusted based on the measurement data, achieving the goal of adjusting the network parameters of the target motion model according to specific circumstances. This improves the flexibility, accuracy, and generalizability of the network parameter settings of the target motion model, ensuring that it outputs a reconstructed image with high image quality when receiving different types of input images.

[0066] Figure 2A This is a flowchart of the method for generating predicted deformation data provided in an embodiment of the present invention. This embodiment refines the method for generating predicted deformation data based on a target motion model as described in the previous embodiments. Figure 2A and Figure 2B As shown, the method includes:

[0067] S2101. Determine the model input data, wherein the model input data is random deformation data or a coordinate set, wherein the coordinate set includes the coordinates of the at least one target time and the measurement data corresponding to the projection angle or azimuth angle corresponding to each target time, or includes the pixel coordinates of the at least one target time and the reconstructed image corresponding to each target time.

[0068] In one embodiment, random deformation data corresponding to a set data distribution pattern is determined based on a uniformly distributed random function, such as a random vector field corresponding to the initial image pixel coordinate distribution pattern. This embodiment provides a fast and convenient method for generating random deformation data.

[0069] In one embodiment, while determining the random deformation data, the number of predicted deformation data points that the target motion model is expected to generate based on the predicted deformation data is also determined. As can be seen from the foregoing embodiments, this number is the same as the number of reconstructed images included in the reconstructed image sequence. The number of reconstructed images included in the reconstructed image sequence is equal to the temporal resolution ratio, which is the ratio of the temporal resolution of the reconstructed image to the temporal resolution of the initial image. Therefore, the user can use the ratio of the desired reconstructed image temporal resolution to the temporal resolution of the initial image as the number of predicted deformation data points.

[0070] In one embodiment, the model input data includes a set of coordinates, which includes at least one target time and the pixel coordinates of the reconstructed image corresponding to each target time. The pixel coordinates of the reconstructed image can be understood as the row and column distribution data of pixels, such as 512×512 or 1024×1024.

[0071] In one embodiment, the coordinate set includes at least one target time and the coordinates of the measurement data corresponding to the projection angle or azimuth angle for each target time. The data distribution of the measurement data can be represented as A×B, where A is the number of rows in the matrix, B is the number of columns in the matrix, and both A and B are positive integers. Optionally, the data distribution of the predicted deformation data output by the target motion model is the same as the data distribution of the measurement data, achieving the technical effect of using the coordinate set as the model input data to limit the data distribution of the predicted deformation data output by the target motion model.

[0072] Different types of model input data correspond to different specific networks for the target motion model.

[0073] S2102. Input the model input data into the target motion model so that the target motion model generates at least one set of predicted deformation data.

[0074] Input the model input data into the target motion model so that the target motion model generates at least one set of predicted deformation data based on the model input data.

[0075] In one embodiment, while inputting random deformation data into the target motion model, the number of predicted deformation data points is also input into the target motion model. In this embodiment, the number of predicted deformation data points is configured as a modifiable item. Users can set the number of predicted deformation data points according to their actual needs. It should be noted that the number of predicted deformation data points is greater than 1, and less than or equal to the number of azimuth angles (for MRI scan data) or projection angles (for CT scan data) corresponding to the measured data.

[0076] The target motion model in this embodiment does not require pre-training, thus its application scenarios are not limited by the training samples; it directly generates at least one set of predicted deformation data based on the model input data, which improves the ease of use of the model.

[0077] Figure 3A This is a flowchart of another medical image processing method provided by an embodiment of the present invention. This embodiment adds a step of determining the initial image based on predicted deformation data and measurement data, based on the aforementioned embodiments. Figure 3A and Figure 3B As shown, the method includes:

[0078] S310. Generate at least one set of predicted deformation data based on the target motion model, wherein each set of predicted deformation data corresponds to a target time.

[0079] S3201. Based on the number of groups of predicted deformation data, the measurement data is divided using a set data division strategy to obtain measurement data groups corresponding to each group of predicted deformation data. Each measurement data group corresponds to a combination of projection angles or azimuth angles, and there is no overlap in the data acquisition time periods corresponding to different measurement data groups.

[0080] In this context, the absence of overlapping data acquisition time periods for different measurement data groups can be understood as meaning that for any projection angle or azimuth angle, a portion of the corresponding measurement data can be considered as one measurement data group. Specifically, the portion of measurement data corresponding to the projection angle or azimuth angle at adjacent acquisition times can be divided into at least one measurement data group, while the measurement data corresponding to the projection angle or azimuth angle at non-adjacent acquisition times cannot be divided into at least one measurement data group. Alternatively, the portion of measurement data corresponding to the projection angle or azimuth angle of the same motion state of a musculoskeletal organ can be divided into at least one measurement data group; the measurement data corresponding to the projection angle or azimuth angle of different motion states of a musculoskeletal organ cannot be divided into at least one measurement data group.

[0081] It is understandable that during CT or MRI data acquisition, if the data acquisition method is not based on an increasing or decreasing sequence of projection angles or azimuth angles, then when a combination of projection angles or azimuth angles includes at least one projection angle or azimuth angle, that at least one projection angle or azimuth angle is distributed dispersedly rather than clustered within the combination of projection angles or azimuth angles. In one embodiment, all projection angles or azimuth angles corresponding to the measurement data are taken as a set of projection angles or azimuth angles.

[0082] In one embodiment, the number of projection angles included in the projection angle combinations corresponding to different measurement data groups is the same, or the number of azimuth angles included in the azimuth angle combinations corresponding to different measurement data groups is the same.

[0083] S3202, an image reconstruction algorithm based on motion compensation, the at least one set of predicted deformation data, the correspondence between the at least one set of predicted deformation data and the corresponding measurement data set, and the determination of the initial image corresponding to the measurement data.

[0084] Based on the at least one set of predicted deformation data and the correspondence between the at least one set of predicted deformation data and the corresponding measurement data set, motion-compensated image reconstruction is performed on the measurement data to obtain the initial image.

[0085] S330. Based on the at least one set of predicted deformation data, the initial image is deformed to obtain at least one reconstructed image, and the at least one reconstructed image is used as a sequence of reconstructed images.

[0086] S340. Determine the predicted data corresponding to the reconstructed image sequence, as well as the error between the predicted data and the measured data, and adjust the network parameters of the target motion model in reverse according to the error.

[0087] S350, Return to the step of generating at least one set of predicted deformation data based on the target motion model until the error meets the set error condition.

[0088] This embodiment performs motion-compensated image reconstruction on the measurement data based on at least one set of predicted deformation data to obtain an initial image. This helps improve the quality of the initial image, as well as the quality of each reconstructed image in the reconstructed image sequence determined based on the initial image.

[0089] Figure 4 This is a flowchart of another medical image processing method provided by an embodiment of the present invention. This embodiment is used to further refine the predicted data corresponding to the reconstructed image sequence based on the foregoing embodiments. For example... Figure 4 As shown, the method includes:

[0090] S410. Generate at least one set of predicted deformation data based on the target motion model, wherein each set of predicted deformation data corresponds to a target time.

[0091] S420. Determine the initial image corresponding to the measurement data, with the target time within the measurement data acquisition time.

[0092] S430. Based on the at least one set of predicted deformation data, the initial image is subjected to deformation processing to obtain at least one reconstructed image, and the at least one reconstructed image is used as a sequence of reconstructed images.

[0093] S440. Determine the image type corresponding to the measurement data, use the prediction data determination strategy corresponding to the image type to determine the prediction data corresponding to the reconstructed image sequence, and determine the error between the prediction data and the measurement data. Adjust the network parameters of the target motion model in reverse according to the error.

[0094] In one embodiment, when the image type corresponding to the measurement data is an MRI image, a prediction data determination strategy corresponding to the MRI image is used to determine the prediction data corresponding to the reconstructed image sequence, including:

[0095] Step a1: When the reconstructed image is an MRI image, determine the azimuth combination corresponding to each reconstructed image based on the azimuth combination corresponding to the predicted deformation data corresponding to the reconstructed image.

[0096] Step a2: Calculate at least one radial line data group corresponding to each reconstructed image based on the first set signal model and the azimuth angle combination corresponding to each reconstructed image.

[0097] Step a3: Take the union of the radial line data groups corresponding to all reconstructed images in the reconstructed image sequence as the prediction data group corresponding to the reconstructed image sequence.

[0098] This embodiment determines the azimuth combination corresponding to each reconstructed image based on the azimuth combination corresponding to the predicted deformation data corresponding to the reconstructed image. Based on the first set signal model and the azimuth combination corresponding to each reconstructed image, it determines at least one radial line data group corresponding to each reconstructed image, thereby determining the predicted data group corresponding to the reconstructed image sequence. This ensures the accuracy of determining the predicted data corresponding to the reconstructed image sequence when the initial image is an MRI image.

[0099] In one embodiment, when the image type corresponding to the measurement data is a CT image, the prediction data corresponding to the reconstructed image sequence is determined using a prediction data determination strategy for CT images, including:

[0100] Step b1: When the reconstructed image is a CT image, determine the projection angle combination corresponding to each reconstructed image based on the projection angle combination corresponding to the predicted deformation data corresponding to the reconstructed image.

[0101] Step b2: Calculate the prediction data set corresponding to the reconstructed image along the projection angle combination based on the second set signal model and the projection angle combination corresponding to each reconstructed image.

[0102] Step b3: Take the union of the prediction data groups corresponding to all reconstructed images in the reconstructed image sequence as the prediction data.

[0103] In one embodiment, the projection angle is the midpoint of the rotation angle range of the gantry when the X-ray source of the CT imaging system outputs X-rays. For example, if the X-ray source outputs the beam at 125-126 degrees, then 125-126 degrees is a projection angle.

[0104] The second set signal model is an existing projection model, such as a forward projection model.

[0105] This embodiment determines the prediction data set for each reconstructed image by combining the second preset signal model with the projection angle corresponding to each reconstructed image, thereby determining the prediction data corresponding to the reconstructed image sequence. This ensures the accuracy of determining the prediction data corresponding to the reconstructed image sequence when the initial image is a CT image.

[0106] S450, Return to the step of generating at least one set of predicted deformation data based on the target motion model until the error meets the set error condition.

[0107] This embodiment uses a prediction data determination strategy corresponding to image type to determine the prediction data corresponding to the reconstructed image sequence, which improves the flexibility and accuracy of prediction data determination.

[0108] Figure 5 This is a flowchart of a network parameter adjustment method for a target motion model provided in an embodiment of the present invention. This embodiment further refines the network parameter adjustment method based on the aforementioned embodiments. The method includes:

[0109] S5301. Determine the first error corresponding to each reconstructed image in the reconstructed image sequence based on the error.

[0110] Since the predicted data is determined based on the reconstructed image sequence, if there is an error between the predicted data and the measured data, each reconstructed image in the reconstructed image sequence will also have a corresponding error. In this embodiment, the corresponding error is taken as the first error.

[0111] S5302. Based on the first error corresponding to each reconstructed image, determine the second error corresponding to each predicted deformation data corresponding to each reconstructed image.

[0112] Since the reconstructed image is determined based on the initial image and the corresponding predicted deformation data, if the reconstructed image has a first error, then the corresponding predicted deformation data will also have a corresponding error, assuming the initial image is accurate. In this embodiment, the error in the predicted deformation data is considered the second error.

[0113] S5303. Adjust the network parameters of the target motion model in reverse according to the second error corresponding to each predicted deformation data.

[0114] The network parameters of the target motion model are adjusted in reverse according to the second error corresponding to each predicted deformation data, so that the target motion model after network parameter adjustment can output at least one new set of predicted deformation data.

[0115] In this embodiment, the error between the predicted data corresponding to the reconstructed image sequence and the measurement data corresponding to the initial image is passed layer by layer to the target motion model, so that the target motion model can adjust its network parameters according to the error. Since the measurement data is real data, describing the real motion of the moving organ within the time range of data collection at each azimuth angle or projection angle, this network error adjustment method has high accuracy and also makes the target motion model have high generalizability.

[0116] Figure 6 This is a structural block diagram of a medical image processing apparatus according to another embodiment of the present invention. The apparatus is used to execute the medical image processing method provided in any of the above embodiments, and the apparatus may be implemented in software or hardware. The apparatus includes:

[0117] The deformation data prediction module 61 is used to generate at least one set of predicted deformation data based on the target motion model, wherein each set of predicted deformation data corresponds to a target time.

[0118] The initial image determination module 62 is used to determine the initial image corresponding to the measurement data, wherein the target time is within the acquisition time of the measurement data;

[0119] Image processing module 63 is used to perform deformation processing on the initial image based on the at least one set of predicted deformation data to obtain at least one reconstructed image, and to use the at least one reconstructed image as a sequence of reconstructed images;

[0120] The model network parameter adjustment module 64 is used to determine the predicted data corresponding to the reconstructed image sequence, and the error between the predicted data and the measurement data, and adjust the network parameters of the target motion model in reverse according to the error;

[0121] The iteration module 65 is used to return to the step of generating at least one set of predicted deformation data based on the target motion model until the error meets the set error condition.

[0122] In one embodiment, the deformation data prediction module 61 is used for:

[0123] Determine the model input data, which is random deformation data or a coordinate set. The coordinate set includes the coordinates of the measurement data corresponding to the at least one target time and the projection angle or azimuth angle corresponding to each target time, or includes the pixel coordinates of the at least one target time and the reconstructed image corresponding to each target time.

[0124] The model input data is input into the target motion model so that the target motion model generates at least one set of predicted deformation data.

[0125] In one embodiment, the deformation data prediction module 61 is used for:

[0126] While inputting random deformation data into the target motion model, the number of predicted deformation data points is also input into the target motion model.

[0127] In one embodiment, the predicted deformation data is a motion vector field.

[0128] In one embodiment, the initial medical image is a medical image including musculoskeletal organs.

[0129] In one embodiment, the image processing module 62 is used for:

[0130] Based on the number of groups of the predicted deformation data, the measurement data is divided using a set data division strategy to obtain measurement data groups corresponding to each group of predicted deformation data. Each measurement data group corresponds to a projection angle combination or azimuth angle combination, and the data acquisition time periods corresponding to different measurement data groups do not overlap.

[0131] Based on the at least one set of predicted deformation data and the correspondence between the at least one set of predicted deformation data and the corresponding measurement data set, motion-compensated image reconstruction is performed on the measurement data to obtain the initial image.

[0132] In one embodiment, the model network parameter adjustment module 64 is used for:

[0133] When the reconstructed image is an MRI image, the azimuth combination corresponding to each reconstructed image is determined based on the azimuth combination corresponding to the predicted deformation data corresponding to the reconstructed image.

[0134] Based on the first set signal model and the azimuth angle combination corresponding to each reconstructed image, at least one radial line data group corresponding to each reconstructed image is calculated.

[0135] The union of the radial line data sets corresponding to all reconstructed images in the reconstructed image sequence is taken as the prediction data set corresponding to the reconstructed image sequence.

[0136] In one embodiment, the model network parameter adjustment module 64 is used for:

[0137] When the reconstructed image is a CT image, the projection angle combination corresponding to each reconstructed image is determined based on the projection angle combination corresponding to the predicted deformation data corresponding to the reconstructed image.

[0138] Based on the second set signal model and the combination of projection angles corresponding to each reconstructed image, the prediction data group corresponding to the reconstructed image is calculated along the combination of projection angles.

[0139] The union of the predicted data sets corresponding to all reconstructed images in the reconstructed image sequence is used as the predicted data.

[0140] Compared to existing technologies, the technical solution of the medical image processing device provided in this embodiment, since the measurement data used to reconstruct the initial image is the scanning data collected by the medical imaging system, and the target time is within the acquisition time of the measurement data, adjusts the network parameters of the target motion model based on the sum of the aforementioned errors, so that the predicted deformation data output by the model is consistent with the temporal phase of the target moving organ of the target object at the target time. In fact, the network parameters of the target motion model are adjusted based on the measurement data, which realizes the purpose of adjusting the network parameters of the target motion model according to the specific situation, improves the flexibility, accuracy and generalizability of the network parameter settings of the target motion model, and can ensure that it will output a reconstructed image with high image quality when receiving different types of input images.

[0141] The medical image processing apparatus provided in the embodiments of the present invention can execute the medical image processing method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0142] Figure 7 This is a schematic diagram of a C-arm CT imaging system according to another embodiment of the present invention. The system includes a gantry 1211, a detector 1212, a bed 1214, an X-ray tube 1215, a C-arm drive shaft 1216, a rotating shaft 1217, and a base 1219. The X-ray tube 1215 and the detector 1212 are mounted at both ends of the C-gantry 1211, with their center connection line perpendicular to its rotation axis 1218. The C-gantry 1211 rotates around the rotation axis 1218, thereby capturing image data of the patient 1213 on the bed at different projection angles. The X-ray tube 1215 is controlled by an X-ray generator 123, which controls its current, voltage, and exposure time. The projection data acquired by the detector 1212 is transmitted to a computer via a communication system 126. The gantry 1211 is connected to the C-arm drive shaft 1216, whose power is provided by the rotating shaft 1217. The base 1219 bears the weight. The C-arm control unit 121 controls the rotational speed, angle, and position of the gantry 1211. The spindle control unit 122 connects to the base 1219 and provides power to the entire C-arm system. The X-ray generator 123 controls the current, voltage, and exposure time of the X-ray tube 1215. The data acquisition system 124 coordinates the gantry 1211, detector 1212, and X-ray generator 1215, and collects the acquired data. The bed board control system 125 controls the position and movement speed of the bed board 1214 to achieve different scanning paths for the patient 1213. The communication system 126 connects the C-arm control unit 121, spindle control unit 122, X-ray generator 124, data acquisition system 124, and bed board control system 125, and transmits the acquired projection data to the memory of the computer device 2.

[0143] Figure 8A and Figure 8B A schematic diagram of another CT imaging system is shown. This CT imaging system is a diagnostic CT. Compared with C-arm CT, its gantry 1211 is ring-shaped. The detector 1212 and X-ray tube 1215 are both mounted on the gantry and are relatively distributed. The bed plate 1214 moves in and out of the gantry aperture under the control of the bed plate controller 125. The gantry drives the detector 1212 and X-ray tube 1215 to move around the bed plate 1214.

[0144] Figure 9 This is a schematic diagram of the structure of a computer device provided in another embodiment of the present invention, as shown below. Figure 9 As shown, the computer device 2 includes a processor 201, a memory 202, an input device 203, and an output device 204; the number of processors 201 in the device can be one or more. Figure 9 Taking a processor 201 as an example; the processor 201, memory 202, input device 203, and output device 204 in the device can be connected via a bus or other means. Figure 9 Taking the example of a connection between China and Israel via a bus.

[0145] The memory 202, as a computer-readable storage medium, can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the medical image processing method in this embodiment of the invention (e.g., deformation data prediction module 61, initial image determination module 62, deformation processing module 63, model network parameter adjustment module 65, and iteration module 65). The processor 201 executes various functional applications and data processing of the device by running the software programs, instructions, and modules stored in the memory 202, thereby realizing the aforementioned medical image processing method.

[0146] The memory 202 may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function; the data storage area may store data created based on terminal usage. Furthermore, the memory 202 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 202 may further include memory remotely located relative to the processor 201, which can be connected to the device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0147] Input device 203 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the device. This input device can be configured in an operator workstation, through which the operator controls the operation of the CT imaging system.

[0148] The output device 204 may include a display device such as a display screen, for example, the display screen of an operation workstation.

[0149] Another embodiment of the present invention provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform a medical image processing method, the method comprising:

[0150] At least one set of predicted deformation data is generated based on the target motion model, and each set of predicted deformation data corresponds to a target time.

[0151] Determine the initial image corresponding to the measurement data, wherein the target time is within the acquisition time of the measurement data;

[0152] Based on the at least one set of predicted deformation data, the initial image is deformed to obtain at least one reconstructed image, and the at least one reconstructed image is used as a sequence of reconstructed images;

[0153] Determine the predicted data corresponding to the reconstructed image sequence, and the error between the predicted data and the measured data, and adjust the network parameters of the target motion model in reverse according to the error;

[0154] Return to the step of generating at least one set of predicted deformation data based on the target motion model until the error meets the set error condition.

[0155] Of course, the computer-executable instructions provided in the embodiments of the present invention are not limited to the method operations described above, but can also perform related operations in the medical image processing method provided in any embodiment of the present invention.

[0156] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the medical image processing methods described in the various embodiments of the present invention.

[0157] It is worth noting that in the embodiments of the above-mentioned medical image processing device, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of the present invention.

[0158] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A medical image processing method, characterized in that, include: At least one set of predicted deformation data is generated based on the target motion model, and each set of predicted deformation data corresponds to a target time. Determine the initial image corresponding to the measurement data, wherein the target time is within the acquisition time of the measurement data; Based on the at least one set of predicted deformation data, the initial image is deformed to obtain at least one reconstructed image, and the at least one reconstructed image is used as a sequence of reconstructed images; Determine the predicted data corresponding to the reconstructed image sequence, and the error between the predicted data and the measured data, and adjust the network parameters of the target motion model in reverse according to the error; Return to the step of generating at least one set of predicted deformation data based on the target motion model until the error meets the set error condition; The reconstructed image sequence is determined based on one of the initial images, and the number of reconstructed images included in the reconstructed image sequence is equal to the ratio of the temporal resolution of the reconstructed images to the temporal resolution of the initial image. The determination of the initial image corresponding to the measurement data includes: Based on the number of groups of the predicted deformation data, the measurement data is divided using a set data division strategy to obtain measurement data groups corresponding to each group of predicted deformation data. Each measurement data group corresponds to a projection angle combination or azimuth angle combination, and the data acquisition time periods corresponding to different measurement data groups do not overlap. Based on the at least one set of predicted deformation data and the correspondence between the at least one set of predicted deformation data and the corresponding measurement data set, motion-compensated image reconstruction of the measurement data is performed to obtain an initial image; Wherein, the set of predicted deformation data corresponds to a target time; the projection angle or azimuth angle corresponding to the same set of measurement data belongs to adjacent acquisition time or the same motion state of the corresponding locomotor organ; each part of the measurement data is a set of measurement data, and each part of the measurement data reconstructs an initial image.

2. The method according to claim 1, characterized in that, The generation of at least one set of predicted deformation data based on the target motion model includes: Determine the model input data, which is random deformation data or a set of coordinates. The set of coordinates includes at least one target time and the coordinates of the measurement data corresponding to the projection angle or azimuth angle corresponding to each target time, or includes the at least one target time and the pixel coordinates of the reconstructed image corresponding to each target time. The model input data is input into the target motion model so that the target motion model generates at least one set of predicted deformation data.

3. The method according to claim 2, characterized in that, While inputting random deformation data into the target motion model, the number of predicted deformation data points is also input into the target motion model.

4. The method according to claim 1, characterized in that, The predicted deformation data is a motion vector field.

5. The method according to claim 1, characterized in that, The initial image is a medical image that includes locomotor organs.

6. The method according to claim 1, characterized in that, Determining the prediction data corresponding to the reconstructed image sequence includes: When the reconstructed image is an MRI image, the azimuth combination corresponding to each reconstructed image is determined based on the azimuth combination corresponding to the predicted deformation data corresponding to the reconstructed image. Based on the first set signal model and the azimuth angle combination corresponding to each reconstructed image, at least one radial line data group corresponding to each reconstructed image is calculated. The union of the radial line data sets corresponding to all reconstructed images in the reconstructed image sequence is taken as the prediction data set corresponding to the reconstructed image sequence.

7. The method according to claim 1, characterized in that, Determining the prediction data corresponding to the reconstructed image sequence includes: When the reconstructed image is a CT image, the projection angle combination corresponding to each reconstructed image is determined based on the projection angle combination corresponding to the predicted deformation data corresponding to the reconstructed image. Based on the second set signal model and the combination of projection angles corresponding to each reconstructed image, the prediction data group corresponding to the reconstructed image is calculated along the combination of projection angles. The union of the predicted data sets corresponding to all reconstructed images in the reconstructed image sequence is used as the predicted data.

8. A medical image processing device, characterized in that, include: The deformation data prediction module is used to generate at least one set of predicted deformation data based on the target motion model, wherein each set of predicted deformation data corresponds to a target time. An initial image determination module is used to determine the initial image corresponding to the measurement data, wherein the target time is within the acquisition time of the measurement data; The deformation processing module is used to perform deformation processing on the initial image based on the at least one set of predicted deformation data to obtain at least one reconstructed image, and to use the at least one reconstructed image as a sequence of reconstructed images; The model network parameter adjustment module is used to determine the predicted data corresponding to the reconstructed image sequence, and the error between the predicted data and the measurement data, and to adjust the network parameters of the target motion model in reverse according to the error; The iteration module is used to return to the step of generating at least one set of predicted deformation data based on the target motion model until the error meets the set error condition; The reconstructed image sequence is determined based on one of the initial images, and the number of reconstructed images included in the reconstructed image sequence is equal to the ratio of the temporal resolution of the reconstructed images to the temporal resolution of the initial image. The initial image determination module is specifically used for: Based on the number of groups of the predicted deformation data, the measurement data is divided using a set data division strategy to obtain measurement data groups corresponding to each group of predicted deformation data. Each measurement data group corresponds to a projection angle combination or azimuth angle combination, and the data acquisition time periods corresponding to different measurement data groups do not overlap. Based on the at least one set of predicted deformation data and the correspondence between the at least one set of predicted deformation data and the corresponding measurement data set, motion-compensated image reconstruction of the measurement data is performed to obtain an initial image; Wherein, the set of predicted deformation data corresponds to a target time; the projection angle or azimuth angle corresponding to the same set of measurement data belongs to adjacent acquisition time or the same motion state of the corresponding locomotor organ; each part of the measurement data is a set of measurement data, and each part of the measurement data reconstructs an initial image.

9. A computer device, characterized in that, The computer device includes: One or more processors; Storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the medical image processing method as described in any one of claims 1-7.

10. A storage medium containing computer-executable instructions, characterized in that, The computer-executable instructions, when executed by a computer processor, are used to perform the medical image processing method as described in any one of claims 1-7.

Images