CT data processing method and device based on collaborative branch fusion and electronic equipment

Abstract

The application provides a CT data processing method and device based on collaborative branch fusion, and relates to the field of medical image generation and image processing. The method comprises the following steps: acquiring three-dimensional preoperative CT volume data; inputting the three-dimensional preoperative CT volume data into a denoising network; embedding three parallel axial global state space branches and one local convolution branch in each scale layer of the denoising network, and embedding a collaborative branch fusion (CBF) layer after the three axial global state space branches and the local convolution branch; and performing collaborative weighted fusion on the output data of the three axial global state space branches and the local convolution branch through the CBF layer, so that the three axial global state space branches and the local convolution branch are simultaneously processed and complementarily fused to obtain synthesized three-dimensional intraoperative CT volume data.

Description

Technical Field

[0001] This application relates to the field of medical image generation and image processing technology, and in particular to a CT data processing method, device and electronic device based on collaborative branch fusion. Background Technology

[0002] Currently, in orthopedic trauma and complex fracture surgeries, preoperative computed tomography (preCT) is typically used for preoperative planning, 3D reconstruction, and screw access assessment; intraoperative CT (mostly low-dose cone-beam computed tomography, CBCT) is used for real-time localization and navigation. However, existing intraoperative CBCT techniques are limited by low-dose acquisition, scattering, limited field of view, and artifacts caused by metallic implants, often resulting in increased noise, blurred cortical bone boundaries, and obvious streaking artifacts. Local areas are often overly smooth, easily creating illusions, leading to decreased intraoperative image usability and affecting downstream segmentation, registration, and navigation accuracy. To address the issues of insufficient intraoperative image quality and data scarcity, recent attempts have focused on using image generation / domain transfer methods to synthesize high-quality preCT into volumetric data with iCT appearance characteristics for intraoperative assistance or as training data augmentation. However, existing techniques often suffer from structural information loss in CT images because fracture fissures, oblique fracture surfaces, and other structures typically extend across multiple directions. Summary of the Invention

[0003] The purpose of this invention is to provide a CT data processing method, apparatus, and electronic device based on collaborative branch fusion to solve the technical problem of missing structural information in CT images.

[0004] In a first aspect, this application provides a CT data processing method based on collaborative branch fusion, the method comprising: Acquire 3D preoperative CT volume data; The three-dimensional preoperative CT body data is input into a denoising network; each scale layer in the denoising network is embedded with three parallel axial global state space branches and one local convolution branch, as well as a Cooperative Branch Fusion (CBF) layer after the three axial global state space branches and the local convolution branch. The output data of the three axial global state space branches and the local convolution branches are collaboratively weighted and fused through the CBF layer, so that the three axial global state space branches and the local convolution branches are processed simultaneously and fused complementaryly to obtain the synthesized three-dimensional intraoperative CT volume data.

[0005] In one possible implementation, the output data based on the three-axis global state space branches and the local convolutional branches are collaboratively weighted and fused through the CBF layer, so that the three-axis global state space branches and the local convolutional branches are processed simultaneously and complementaryly fused to obtain synthesized three-dimensional intraoperative CT volume data, including: Calculate the mean characteristics of the output data after gating the three axial global state space branches and the local convolutional branches; Based on the mean feature, a lightweight mapping method is used to generate the fusion weights of the three axial global state space branches and each branch of the local convolutional branch after expansion on each feature channel; The weighted summation result is obtained based on the mean feature and the fusion weight, and the weighted summation result is used as the collaborative weighted fusion result of the CBF layer, so that the three axial global state space branches and the local convolution branches are processed simultaneously and fused complementaryly to obtain the synthesized three-dimensional intraoperative CT body data.

[0006] In one possible implementation, the step of generating the fusion weights of the three axial global state space branches and each of the local convolution branches after expansion on each feature channel based on the mean feature through lightweight mapping includes: Based on the mean feature, initial fusion weights are generated for each feature channel of each of the three axial global state space branches and the local convolutional branches using a lightweight mapping method. The initial fusion weights are constrained to the range of 0 to 1 by using the Sigmoid function, and then expanded along the spatial dimension to match the feature size of the corresponding branch, resulting in the expanded fusion weights of each of the three axial global state space branches and the local convolutional branches on each feature channel.

[0007] In one possible implementation, obtaining the weighted summation result based on the mean feature and the fusion weights includes: The fusion weights are multiplied element by element by the feature data output by the corresponding branches, and the results of each branch are weighted and summed based on the mean feature to obtain the final fusion output feature.

[0008] In one possible implementation, the step of multiplying the fusion weights element-wise with the feature data output by the corresponding branches, and then performing a weighted summation of the results of each branch based on the mean feature to obtain the final fusion output feature includes: The fusion weights are multiplied element-wise with the feature data output from the corresponding branches using the following formula, and the results of each branch are weighted and summed based on the mean feature using the following formula to obtain the final fusion output feature: ; in, This represents the feature data output by the three axial global state space branches and the one local convolution branch; This represents the fusion weight of the three axial global state space branches and each of the local convolution branches on each channel; d This represents the spatial branch corresponding to the length of the three axial global state space branches. h This represents the height-corresponding spatial branch among the three axial global state space branches. This represents the width-corresponding spatial branch among the three axial global state space branches. loc This represents the local convolution branch; This represents element-wise matrix multiplication; represents the preset summation coefficient; F represents the mean characteristic of the output data after gating the three axial global state space branches and the local convolution branches; This represents the result of element-wise multiplication; This represents the final fused output characteristic of the CBF layer.

[0009] In one possible implementation, the CBF layer is used to enable the three axial global state space branches and the local convolution branches to jointly perform the reconstruction task and achieve complementary fusion of global and local operations, thereby improving the continuity between the oblique structure and the thin-layer structure through the CBF layer.

[0010] In one possible implementation, after the output data based on the three axial global state space branches and the local convolutional branches are collaboratively weighted and fused through the CBF layer, the following is also included: The CBF layer obtains the fused features output by co-weighting and fusing the output data of the three axial global state space branches with the local convolution branches. Based on the fused features, the residuals are added back to the input features to stabilize training and preserve the original information flow.

[0011] Secondly, this application provides a CT data processing device based on collaborative branch fusion, comprising: The acquisition module is used to acquire three-dimensional preoperative CT body data; The input module is used to input the three-dimensional preoperative CT body data into the denoising network; each scale layer in the denoising network is embedded with three parallel axial global state space branches and one local convolution branch, as well as a collaborative branch fusion CBF layer after the three axial global state space branches and the local convolution branch; The fusion module is used to perform collaborative weighted fusion of the output data of the three-axis global state space branches and the local convolution branches through the CBF layer, so that the three-axis global state space branches and the local convolution branches are processed simultaneously and fused complementaryly to obtain synthesized three-dimensional intraoperative CT body data.

[0012] Thirdly, this application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program that can run on the processor, and the processor executes the computer program to implement the method described in the first aspect above.

[0013] Fourthly, this application also provides a computer-readable storage medium storing computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method described in the first aspect above.

[0014] This application brings the following beneficial effects: This application provides a CT data processing method, apparatus, and electronic device based on cooperative branch fusion, capable of acquiring three-dimensional preoperative CT volume data; inputting the three-dimensional preoperative CT volume data into a denoising network; each scale layer of the denoising network embeds three parallel axial global state space branches and one local convolutional branch, and a cooperative branch fusion (CBF) layer following the three axial global state space branches and the local convolutional branch; based on the output data of the three axial global state space branches and the local convolutional branch, the data is cooperatively weighted and fused through the CBF layer, so that the three axial global state space branches and the local convolutional branch are processed simultaneously and complementaryly fused to obtain a fused CT volume data. The proposed solution generates three-dimensional intraoperative CT volume data. In this solution, the three-axis global state space branches and local convolution branches are collaboratively and weightedly fused through the collaborative branch fusion mechanism CBF layer. This allows the three-axis global branches and local branches to participate in feature expression and complement each other, achieving global / local complementarity to reduce oversmoothing and local illusions. This enables multi-directional information to contribute simultaneously, rather than through hard selection competition, which is more favorable for cross-directional structures such as oblique fractures. This reduces oversmoothing, local structural breaks, or artifacts, and avoids situations where only a single axial branch or competitive branch selection is used due to the multi-directional extension and distribution of structures such as fracture fissures and oblique fracture surfaces. This solves the technical problem of missing structural information in CT images.

[0015] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 A flowchart illustrating the CT data processing method based on collaborative branch fusion provided in this application embodiment; Figure 2 Another flowchart illustrating the CT data processing method based on collaborative branch fusion provided in this application embodiment; Figure 3 A schematic diagram of the overall architecture of the CT data processing method based on shared sub-band sensing gating provided in the embodiments of this application; Figure 4 A schematic diagram of three-axis state-space modeling in the CT data processing method based on shared sub-band sensing gating provided in the embodiments of this application; Figure 5 A schematic diagram of the structure of a CT data processing device based on collaborative branch fusion provided in this application embodiment; Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] The terms "comprising" and "having," and any variations thereof, used in the embodiments of this application, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the steps or units listed, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0020] Currently, existing methods for high-resolution 3D preCT to iCT synthesis have the following main shortcomings: high data resolution leads to unbearable computation; Transformer diffusion incurs huge memory overhead under voxel-level global attention, which can easily lead to memory overflow; even downsampling or block partitioning will destroy the continuity of the 3D structure.

[0021] In existing technologies, since structures such as fracture fissures and oblique fracture surfaces usually extend and are distributed across multiple directions, relying solely on a single axial branch or using competitive branch selection can easily lead to a loss of structural information in CT images.

[0022] Based on this, embodiments of this application provide a CT data processing method, apparatus, and electronic device based on collaborative branch fusion, which can solve the technical problem of missing structural information in CT images.

[0023] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0024] Figure 1 This is a flowchart illustrating a CT data processing method based on collaborative branch fusion, provided as an embodiment of this application. Figure 1 As shown, the method includes: Step S110: Obtain three-dimensional preoperative CT body data.

[0025] As one possible implementation method, such as Figure 2 and Figure 3 As shown, the input 3D preoperative CT volume data can be in nii format, such as 512×512×350 3D preoperative CT volume data. High-resolution 3D volume data such as 384×384×256 can be used as native training data.

[0026] Step S120: Input the three-dimensional preoperative CT body data into the denoising network.

[0027] In this denoising network, each scale layer is embedded with three parallel axial global state space branches and one local convolutional branch, as well as a Cooperative Branch Fusion (CBF) layer following the three axial global state space branches and the local convolutional branch.

[0028] For example, such as Figure 4 As shown, features F∈R are processed by three parallel axial global state space branches and one local convolution branch. B×C×H′×W′×D′The data is serialized along the D', H', and W' directions and input into the state-space model, resulting in three global branch outputs Fd, Fh, and Fw. Simultaneously, a local convolution branch Floc is introduced, using lightweight 3D depthwise separable convolution to refine local textures and bone boundary details. These four branches (three axial global state-space branches and one local convolution branch) output in parallel, and the data from this parallel output is input into the co-branch fusion CBF layer.

[0029] As one possible implementation, the CBF layer is used to achieve the joint execution of the reconstruction task by the three-axis global state space branches and local convolution branches through collaborative weighted fusion of the three-axis global state space branches and local convolution branches, and to achieve complementary fusion of global and local components, so as to improve the continuity of the oblique structure and the thin-layer structure through the CBF layer.

[0030] Step S130: The output data of the three-axis global state space branches and local convolution branches are collaboratively weighted and fused through the CBF layer so that the three-axis global state space branches and local convolution branches are processed simultaneously and fused complementaryly to obtain the synthesized three-dimensional intraoperative CT volume data.

[0031] As an example, the output data based on the three-axis global state space branches and local convolution branches are collaboratively weighted and fused through a CBF layer, so that the three-axis global state space branches and local convolution branches are processed simultaneously and fused complementaryly to obtain synthesized three-dimensional intraoperative CT volume data. Specifically, this may include the following steps: The mean features of the output data after gating the global state space branch and the local convolution branch along three axes are calculated. Based on the mean features, the fusion weights of each branch in the global state space branch and the local convolution branch along three axes are generated on each feature channel through a lightweight mapping method. The weighted summation result is obtained based on the mean features and the fusion weights, and the weighted summation result is used as the collaborative weighted fusion result of the CBF layer, so that the global state space branch and the local convolution branch along three axes are processed simultaneously and fused complementaryly to obtain the synthesized three-dimensional intraoperative CT body data.

[0032] For example, the above-mentioned fusion weights, which are the expanded weights of each of the three axial global state space branches and the local convolution branches based on the mean feature and generated through lightweight mapping, can specifically include the following steps: Based on the mean feature, initial fusion weights are generated for each feature channel of each branch in the three-axis global state space branch and local convolution branch using a lightweight mapping method. The initial fusion weights are constrained to the range of 0 to 1 by the Sigmoid function, and the initial fusion weights are expanded along the spatial dimension to be consistent with the feature size of the corresponding branch, so as to obtain the expanded fusion weights of each branch in the three-axis global state space branch and local convolution branch on each feature channel.

[0033] For example, the mean feature F after four-branch gating is first calculated, and then the fusion weight A of each branch in each channel is predicted by lightweight mapping, constrained to 0~1 by a sigmoid function, and broadcast to the spatial dimension; the final output is a weighted sum. This fusion method is collaborative rather than hard-selective, allowing multiple branches to jointly undertake the reconstruction task, which is particularly suitable for the representation of fracture structures in multiple directions and at multiple scales. CBF improves the continuity of oblique and thin-layer structures, reduces local artifacts and structural fractures, while maintaining overall morphological stability.

[0034] In one optional implementation, the weighted summation result obtained based on the mean feature and the fusion weight may specifically include the following steps: multiplying the fusion weight element by element with the feature data output by the corresponding branch, and performing a weighted summation of the results of each branch based on the mean feature to obtain the final fusion output feature.

[0035] For example, the above method involves multiplying the fusion weights element-wise with the feature data output by the corresponding branches, and then performing a weighted summation of the results from each branch based on the mean feature to obtain the final fusion output feature. Specifically, this may include the following steps: The fusion weights are multiplied element-wise with the feature data output from the corresponding branches using the following formula, and the results of each branch are weighted and summed based on the mean feature using the following formula to obtain the final fusion output feature: ; in, This represents the feature data output by three axial global state space branches and one local convolutional branch; This represents the fusion weight of each of the three-axis global state space branches and local convolution branches on each channel; d This represents the spatial branch corresponding to the length in the global state space branches along the three axes. h This represents the height-corresponding spatial branch in the global state space branches along the three axes. This represents the width-corresponding spatial branch in the global state space branches along the three axes. loc Indicates a local convolution branch; This represents element-wise matrix multiplication; Indicates the preset summation coefficient; F represents the mean characteristic of the output data after gating the global state space branch and the local convolution branch along the three axes; This represents the result of element-wise multiplication; This represents the final fused output characteristics of the CBF layer.

[0036] For example, the average of the features of the four branches after gating modulation is first calculated to obtain the mean feature. Then, based on the mean feature, a fusion weight A is generated for each feature channel of each branch through the weight generation module; the fusion weight is then constrained to the range of 0 to 1 by the Sigmoid function and expanded along the spatial dimension to be consistent with the feature size of the corresponding branch; finally, the expanded fusion weight is multiplied element by element with the feature of the corresponding branch, and the results of each branch are weighted and summed to obtain the fusion output feature.

[0037] By employing a collaborative branch fusion mechanism, the CBF layer performs collaborative weighted fusion of the three-axis global state space branches and local convolutional branches. This allows the three-axis global and local branches to participate in feature representation simultaneously and complement each other, achieving global / local complementarity to reduce oversmoothing and local illusions. It enables multi-directional information to contribute simultaneously, rather than through hard selection competition, making it more favorable for cross-directional structures such as oblique fractures. This reduces oversmoothing, local structural breaks, or artifacts, and solves the technical problem that structural information is easily lost in CT images when relying on a single axial branch or using competitive branch selection because structures such as fracture fissures and oblique fracture surfaces usually extend and distribute across multiple directions.

[0038] In this embodiment, robustness to complex scenes is demonstrated: for scenes with limited field of view and metallic interference, frequency modeling and collaborative fusion can reduce local "illusions" and oversmoothing phenomena, resulting in more stable generated results. While ensuring trainability, it simultaneously improves the global anatomical consistency and local bone boundary sharpness of the generated results, and enhances the performance of downstream tasks such as segmentation.

[0039] In some embodiments, after the output data based on the three-axis global state space branches and local convolution branches are collaboratively weighted and fused through a CBF layer, the method may further include the following steps: The CBF layer obtains the fused features of the CBF layer output by co-weighting and fusing the output data of the three-axis global state space branches and the local convolution branches. Based on the fused features, the residuals are added back to the input features to stabilize the training and preserve the original information flow.

[0040] The output data of the three axial global state space branches and the local convolutional branches can be fused with CBF through a collaborative branch to obtain the output features of this layer, and then added back to the input features in the form of residuals, thereby stabilizing the training and preserving the original information flow.

[0041] In this embodiment, the downstream task benefits are significant: when synthetic data is used for segmentation model training, it can more effectively bridge domain differences, improve Dice on real intraoperative CT and reduce HD95, demonstrating stronger engineering application value.

[0042] Figure 5 A schematic diagram of a CT data processing device based on collaborative branch fusion is provided. Figure 5 As shown, the CT data processing device 500 based on collaborative branch fusion includes: Acquisition module 501 is used to acquire three-dimensional preoperative CT body data; The input module 502 is used to input the three-dimensional preoperative CT body data into the denoising network; each scale layer in the denoising network is embedded with three parallel axial global state space branches and one local convolution branch, as well as a collaborative branch fusion CBF layer after the three axial global state space branches and the local convolution branch; The fusion module 503 is used to perform collaborative weighted fusion of the output data of the three-axis global state space branches and the local convolution branches through the CBF layer, so that the three-axis global state space branches and the local convolution branches are processed simultaneously and fused complementaryly to obtain synthesized three-dimensional intraoperative CT body data.

[0043] The CT data processing device based on collaborative branch fusion provided in this application has the same technical features as the CT data processing method based on collaborative branch fusion provided in the above embodiments, so it can also solve the same technical problems and achieve the same technical effects.

[0044] An electronic device provided in this application embodiment, such as Figure 6 As shown, the electronic device 600 includes a processor 602 and a memory 601. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the steps of the method provided in the above embodiments.

[0045] See Figure 6 The electronic device also includes a bus 603 and a communication interface 604. The processor 602, the communication interface 604 and the memory 601 are connected through the bus 603. The processor 602 is used to execute executable modules, such as computer programs, stored in the memory 601.

[0046] The memory 601 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 604 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0047] Bus 603 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 6 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0048] The memory 601 is used to store programs. After receiving an execution instruction, the processor 602 executes the program. The method executed by the apparatus defined by the process disclosed in any of the preceding embodiments of this application can be applied to the processor 602 or implemented by the processor 602.

[0049] The processor 602 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 602 or by instructions in software form. The processor 602 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 601, and processor 602 reads the information from memory 601 and, in conjunction with its hardware, completes the steps of the above method.

[0050] Corresponding to the above-described CT data processing method based on cooperative branch fusion, this application embodiment also provides a computer-readable storage medium storing computer-executable instructions. When the computer-executable instructions are invoked and executed by a processor, the computer-executable instructions cause the processor to perform the steps of the above-described CT data processing method based on cooperative branch fusion.

[0051] The CT data processing device based on collaborative branch fusion provided in this application embodiment can be specific hardware on the device or software or firmware installed on the device. The implementation principle and technical effects of the device provided in this application embodiment are the same as those in the foregoing method embodiments. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the foregoing method embodiments. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can all be referred to the corresponding processes in the above method embodiments, and will not be repeated here.

[0052] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0053] For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0054] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0055] In addition, the functional units in the embodiments provided in this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0056] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the CT data processing method based on collaborative branch fusion described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0057] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "first", "second", "third", etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0058] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application. All should be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A CT data processing method based on collaborative branch fusion, characterized in that, The method includes: Acquire 3D preoperative CT volume data; The three-dimensional preoperative CT volume data is input into a denoising network; each scale layer in the denoising network is embedded with three parallel axial global state space branches and one local convolution branch, as well as a CBF layer fused by a co-branch after the three axial global state space branches and the local convolution branch; The output data of the three axial global state space branches and the local convolution branches are collaboratively weighted and fused through the CBF layer, so that the three axial global state space branches and the local convolution branches are processed simultaneously and fused complementaryly to obtain the synthesized three-dimensional intraoperative CT volume data.

2. The method according to claim 1, characterized in that, The output data based on the three-axis global state space branches and the local convolution branches are collaboratively weighted and fused through the CBF layer, so that the three-axis global state space branches and the local convolution branches are processed simultaneously and complementaryly fused to obtain synthesized three-dimensional intraoperative CT volume data, including: Calculate the mean characteristics of the output data after gating the three axial global state space branches and the local convolutional branches; Based on the mean feature, a lightweight mapping method is used to generate the fusion weights of the three axial global state space branches and each branch of the local convolutional branch after expansion on each feature channel; The weighted summation result is obtained based on the mean feature and the fusion weight, and the weighted summation result is used as the collaborative weighted fusion result of the CBF layer, so that the three axial global state space branches and the local convolution branches are processed simultaneously and fused complementaryly to obtain the synthesized three-dimensional intraoperative CT body data.

3. The method according to claim 2, characterized in that, The process of generating fusion weights for each feature channel of the three axial global state space branches and each of the local convolution branches based on the mean feature using a lightweight mapping method includes: Based on the mean feature, initial fusion weights are generated for each feature channel of each of the three axial global state space branches and the local convolutional branches using a lightweight mapping method. The initial fusion weights are constrained to the range of 0 to 1 by using the Sigmoid function, and then expanded along the spatial dimension to match the feature size of the corresponding branch, resulting in the expanded fusion weights of each of the three axial global state space branches and the local convolutional branches on each feature channel.

4. The method according to claim 2, characterized in that, The weighted summation result obtained based on the mean feature and the fusion weight includes: The fusion weights are multiplied element by element by the feature data output by the corresponding branches, and the results of each branch are weighted and summed based on the mean feature to obtain the final fusion output feature.

5. The method according to claim 4, characterized in that, The step of multiplying the fusion weights element-wise with the feature data output by the corresponding branches, and then performing a weighted summation of the results of each branch based on the mean feature to obtain the final fusion output feature includes: The fusion weights are multiplied element-wise with the feature data output from the corresponding branches using the following formula, and the results of each branch are weighted and summed based on the mean feature using the following formula to obtain the final fusion output feature: ； in, This represents the feature data output by the three axial global state space branches and the one local convolution branch; This represents the fusion weight of the three axial global state space branches and each of the local convolution branches on each channel; d This represents the spatial branch corresponding to the length of the three axial global state space branches. h This represents the height-corresponding spatial branch among the three axial global state space branches. This represents the width-corresponding spatial branch among the three axial global state space branches. loc This represents the local convolution branch; This represents element-wise matrix multiplication; represents the preset summation coefficient; F represents the mean characteristic of the output data after gating the three axial global state space branches and the local convolution branches; This represents the result of element-wise multiplication; This represents the final fused output characteristic of the CBF layer.

6. The method according to claim 1, characterized in that, The CBF layer is used to achieve collaborative weighted fusion of the three axial global state space branches and the local convolution branches, enabling the three axial global state space branches and the local convolution branches to jointly perform the reconstruction task and achieve complementary fusion of global and local aspects, thereby improving the continuity of the oblique structure and the thin-layer structure through the CBF layer.

7. The method according to claim 1, characterized in that, After the output data based on the three axial global state space branches and the local convolution branches are collaboratively weighted and fused through the CBF layer, the following is also included: The CBF layer obtains the fused features output by co-weighting and fusing the output data of the three axial global state space branches with the local convolution branches. Based on the fused features, the residuals are added back to the input features to stabilize training and preserve the original information flow.

8. A CT data processing device based on collaborative branch fusion, characterized in that, include: The acquisition module is used to acquire three-dimensional preoperative CT body data; The input module is used to input the three-dimensional preoperative CT body data into the denoising network; Each scale layer in the denoising network embeds three parallel axial global state space branches and one local convolutional branch, as well as a collaborative branch fusion CBF layer following the three axial global state space branches and the local convolutional branch. The fusion module is used to perform collaborative weighted fusion of the output data of the three-axis global state space branches and the local convolution branches through the CBF layer, so that the three-axis global state space branches and the local convolution branches are processed simultaneously and fused complementaryly to obtain synthesized three-dimensional intraoperative CT body data.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions that, when invoked and executed by a processor, cause the processor to perform the method according to any one of claims 1 to 7.

Images