A method, device and equipment for automatically generating peripheral nerve courses

Abstract

The embodiments of this specification disclose a method, device and equipment for automatically generating the course of peripheral nerves, including: obtaining a CT image to be processed of a designated inspection site obtained by CT scanning; obtaining CT scanning parameters corresponding to the CT image to be processed; obtaining a magnetic resonance image of the designated inspection site by magnetic resonance imaging; extracting peripheral nerve features from the magnetic resonance image to obtain peripheral nerve feature information of the designated inspection site; based on the peripheral nerve feature information of the designated inspection site, marking the peripheral nerves in the CT image to be processed to obtain a marked CT image; inputting the marked CT image and the CT scanning parameters into a pre-trained peripheral nerve automatic generation model to obtain a result image of the course of peripheral nerves of the designated inspection site. The result image of the course of peripheral nerves generated by this application can be applied to the field of radiotherapy, which is of great benefit to the disease analysis during radiotherapy.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method, device and equipment for automatically generating the course of peripheral nerves. Background Art

[0002] In the medical field, many treatments require precise delineation of nerves in specific locations. However, there are some problems with the current methods of peripheral nerve delineation, which brings many problems to the follow-up.

[0003] Traditional peripheral nerve delineation mainly relies on the physician's experience and is performed based on the approximate direction of the anatomical structure. This method has the following shortcomings:

[0004] Since it depends on the personal experience of the physician, the delineation results between different physicians may differ, affecting the accuracy and consistency of treatment. At the same time, the physician needs to have rich anatomical knowledge and clinical experience to accurately delineate the peripheral nerves, which places high demands on the physician's professional level. In addition, it is difficult for beginners to master this experience-based method of peripheral nerve delineation, which requires a long period of learning and practice. Summary of the invention

[0005] One or more embodiments of the present specification provide a method, device and apparatus for automatically generating the course of peripheral nerves, which are used to solve the technical problems raised by the background technology.

[0006] One or more embodiments of this specification adopt the following technical solutions:

[0007] One or more embodiments of this specification provide a method for automatically generating a peripheral nerve course, the method comprising:

[0008] Acquire a CT image to be processed of a designated examination part obtained through CT scanning;

[0009] Acquiring CT scanning parameters corresponding to the CT image to be processed;

[0010] Obtaining a magnetic resonance image of the designated examination site by magnetic resonance imaging;

[0011] Extracting peripheral nerve features from the magnetic resonance image to obtain peripheral nerve feature information of the designated inspection site;

[0012] Based on the peripheral nerve characteristic information of the designated inspection site, the peripheral nerves in the CT image to be processed are marked to obtain a marked CT image;

[0013] The marked CT image and the CT scanning parameters are input into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site, wherein the peripheral nerve automatic generation model is a neural network model.

[0014] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0015] Improve accuracy and consistency: By automatically generating the course of peripheral nerves, it reduces the reliance on the physician’s personal experience and reduces the differences in delineation results between different physicians, thereby improving the accuracy and consistency of subsequent treatment.

[0016] Lowering the requirements for physicians’ professional level: Traditional methods require physicians to have rich anatomical knowledge and clinical experience, while automatic generation methods can reduce the burden on physicians, enabling more physicians to perform peripheral nerve delineation and improving the utilization efficiency of medical resources.

[0017] Shorten the learning curve: For beginners, it is difficult to master the traditional experience-based peripheral nerve delineation method, while the automatic generation method can provide a more intuitive and easier to learn way, shortening the learning curve and helping to improve the skill level of physicians.

[0018] Improve work efficiency: Automatic generation of peripheral nerve routes can save physicians’ time and energy, allowing them to focus more on other important clinical work and improve work efficiency.

[0019] Provide better support for clinical treatment: Accurate information on the course of peripheral nerves is of great significance for formulating treatment plans and evaluating treatment effects. Automatic generation methods can provide a more reliable basis for clinical treatment and help improve the effectiveness and safety of treatment.

[0020] Promote medical research: The method of automatically generating the course of peripheral nerves can provide more data and information for medical research, help to gain a deeper understanding of the anatomical structure and physiological function of peripheral nerves, and promote the development of the medical field.

[0021] Furthermore, the extracting peripheral nerve features from the magnetic resonance image to obtain peripheral nerve feature information of the designated inspection site includes:

[0022] Determining a suitable segmentation algorithm according to the peripheral nerve characteristics of the designated examination site;

[0023] Segmenting the magnetic resonance image according to the adapted segmentation algorithm to obtain a segmented peripheral nerve image;

[0024] Based on the characteristics of the peripheral nerves, peripheral nerve features are extracted from the segmented peripheral nerve image to obtain peripheral nerve feature information of the designated inspection site.

[0025] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0026] Improve the accuracy of peripheral nerve feature extraction: By selecting an appropriate segmentation algorithm based on the characteristics of the peripheral nerves in the specified examination site, the peripheral nerves can be segmented from the magnetic resonance image more accurately. This helps to reduce the influence of background noise and other interference factors, thereby improving the accuracy of peripheral nerve feature extraction.

[0027] Enhanced adaptability to different inspection sites: Peripheral nerves at different inspection sites may have different characteristics, such as shape, size, texture, etc. Selecting an appropriate segmentation algorithm based on these characteristics can better adapt to different inspection sites and improve the effect of feature extraction.

[0028] Improve work efficiency: Feature extraction based on the characteristics of peripheral nerves can reduce unnecessary calculations and processing and improve work efficiency. At the same time, accurate feature extraction also helps reduce the time and cost of subsequent analysis and diagnosis.

[0029] Furthermore, the peripheral nerves in the CT image to be processed are marked based on the peripheral nerve feature information of the designated inspection site to obtain a marked CT image, including:

[0030] Registering the CT image to be processed with the magnetic resonance image to obtain a registration result;

[0031] Extracting peripheral nerve characteristic information of the designated inspection site;

[0032] According to the registration result, the peripheral nerve feature information of the designated inspection site is marked on the CT image to be processed to obtain the marked CT image.

[0033] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0034] Improve the accuracy of marking: By registering the CT image to be processed with the MRI image, the characteristic information of the peripheral nerve can be more accurately marked on the CT image. This helps to reduce the error of marking and improve the accuracy of marking.

[0035] Enhanced image visualization: Marking CT images can make peripheral nerves appear more clearly in the image, enhancing the visualization of the image. This helps doctors observe the location and morphology of peripheral nerves more intuitively, thereby better diagnosing and treating them.

[0036] Improve the accuracy of diagnosis: Accurate peripheral nerve marking can provide doctors with more detailed and accurate diagnostic information, which helps to improve the accuracy of diagnosis. This is of great significance for the early detection and treatment of diseases.

[0037] Furthermore, registering the to-be-processed CT image with the magnetic resonance image to obtain a registration result includes:

[0038] An image registration algorithm is used to register the spatial positions of the CT image to be processed and the magnetic resonance image to obtain the registration result.

[0039] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0040] Improve diagnostic accuracy: The registered image can more accurately reflect the spatial relationship between different tissues and structures, and can more accurately mark the peripheral nerve characteristic information of the specified examination site to the CT image to be processed to obtain a marked CT image.

[0041] Furthermore, before inputting the marked CT image and the CT scanning parameters into a pre-trained peripheral nerve automatic generation model, the method further includes:

[0042] Acquire a plurality of first CT images of different examination parts obtained by CT scanning, and acquire a first magnetic resonance image corresponding to each first CT image obtained by magnetic resonance imaging;

[0043] Acquire CT scanning parameters corresponding to each first CT image;

[0044] Extracting peripheral nerve features from the plurality of first magnetic resonance images respectively to obtain peripheral nerve feature information of each first magnetic resonance image;

[0045] Based on the peripheral nerve feature information of each first magnetic resonance image, the peripheral nerves in the corresponding first CT image are marked respectively to obtain a corresponding first marked CT image;

[0046] Acquire a first image with clear peripheral nerve courses obtained by other examination methods at the different examination sites;

[0047] According to the clear peripheral nerve courses in each first image, the courses of the peripheral nerves are respectively marked on the first marked CT images to obtain a control CT image with clear courses of the peripheral nerves;

[0048] Constructing a data set by using the CT scanning parameters, the first labeled CT images and the control CT images corresponding to each first CT image;

[0049] The peripheral nerve automatic generation model is trained based on the data set to obtain a peripheral nerve automatic generation model that meets preset requirements.

[0050] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0051] Improve the accuracy of the automatic generation model of peripheral nerves: By acquiring a large number of CT images and magnetic resonance images of different examination sites, extracting and labeling peripheral nerve features, and building a data set for training, the automatic generation model of peripheral nerves can learn more features and patterns, thereby improving its accuracy and reliability.

[0052] Enhance the generalization ability of the model: Using images of different examination parts for training can enable the model to learn the characteristics of peripheral nerves in different situations, enhance its generalization ability, and better handle various complex situations.

[0053] Improve image quality and repeatability: By extracting peripheral nerve features from magnetic resonance images and marking and annotating them on CT images, the quality and repeatability of images can be improved, providing a more reliable basis for subsequent diagnosis and treatment.

[0054] Promoting multimodal image fusion: Combining CT images and magnetic resonance images for analysis and processing can promote multimodal image fusion and provide doctors with more comprehensive and accurate information.

[0055] Improve work efficiency: The automatic generation model of peripheral nerves can quickly and accurately generate peripheral nerve images, greatly improving work efficiency and reducing the workload of doctors.

[0056] Furthermore, the automatic generation model of peripheral nerves is trained based on the data set to obtain the automatic generation model of peripheral nerves that meets the preset requirements, including:

[0057] Inputting the CT scanning parameters corresponding to each first CT image and the first marked CT image into the initial peripheral nerve automatic generation model to obtain a result image to be verified of the peripheral nerve course of the corresponding examination site;

[0058] Comparing each result image to be verified with the corresponding control CT image to obtain a comparison result;

[0059] If the comparison results all meet the preset requirements, a peripheral nerve automatic generation model that meets the preset requirements is trained.

[0060] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0061] Improve the accuracy of the model: By comparing the result image to be verified with the control CT image, the difference between the peripheral nerve course generated by the model and the actual situation can be found, so as to continuously adjust the model parameters and improve the accuracy of the model.

[0062] Enhance the reliability of the model: The comparison results meet the preset requirements, indicating that the peripheral nerve course generated by the model has high reliability and can provide doctors with more accurate diagnostic information.

[0063] Promote model optimization: Based on the comparison results, the model can be optimized in a targeted manner to further improve its performance and expression.

[0064] Improve medical efficiency: Accurate automatic generation model of peripheral nerves can quickly generate peripheral nerve course images, reduce doctors' workload and improve medical efficiency.

[0065] Furthermore, the step of comparing each result image to be verified with a corresponding control CT image to obtain a comparison result includes:

[0066] Extracting the first preset key point of the peripheral nerve course in each result image to be verified;

[0067] Extracting a second key point corresponding to the first preset key point of the peripheral nerve course in the corresponding control CT image;

[0068] The comparison result is obtained based on comparing the first preset key point of each result image to be verified with the second key point of the corresponding control CT image.

[0069] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0070] Improve the accuracy of key points: By extracting the first preset key points and the corresponding second key points of the peripheral nerve course, the results generated by the model can be compared with the actual situation more accurately, thereby discovering possible differences.

[0071] Enhance the comparability of results: Using key points for comparison can make the comparison between different images more objective and comparable, and reduce the interference of other factors.

[0072] Improve the interpretability of the model: By analyzing the differences in key points, we can better understand the decision-making process and output results of the model, providing a basis for improving and optimizing the model.

[0073] Discover potential problems and errors: Comparing key points can help discover potential problems and errors in the peripheral nerve routing generated by the model, and make timely adjustments and improvements.

[0074] Furthermore, the step of inputting the marked CT image and the CT scanning parameters into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site includes:

[0075] Acquire the positional relationship between the designated inspection site and the corresponding one or more organs, wherein the positional relationship includes spatial distance, angle and relative position;

[0076] The positional relationship between the designated inspection site and the corresponding one or more organs, the marked CT image and the CT scanning parameters are input into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site.

[0077] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0078] Improve diagnostic accuracy: By obtaining the spatial distance, angle and relative position between the designated examination site and the corresponding one or more organs, the anatomical relationship between the peripheral nerves and organs can be determined more accurately, thereby improving the accuracy of diagnosing peripheral nerve diseases.

[0079] Reduce misdiagnosis and missed diagnosis: The model automatically generates images of peripheral nerves, which can reduce misdiagnosis caused by complex anatomical structures or subjective judgment errors of doctors. At the same time, the model can comprehensively analyze the relationship between peripheral nerves and organs, which helps to discover potential missed diagnoses.

[0080] Optimize treatment plans: Accurate images of peripheral nerves are essential for surgical planning, which can reduce surgical risks and optimize surgical pathways. At the same time, for patients who need drug treatment, understanding the precise relationship between peripheral nerves and organs helps to choose the most appropriate drugs.

[0081] Improve work efficiency: Using pre-trained models can greatly reduce the time of manual analysis and improve work efficiency. At the same time, the model can process a large number of CT images and is suitable for large-scale medical research and clinical applications.

[0082] One or more embodiments of this specification provide a device for automatically generating a peripheral nerve path, including:

[0083] A CT image acquisition unit, which acquires a CT image to be processed of a designated inspection part obtained through CT scanning;

[0084] A scanning parameter acquisition unit, which acquires CT scanning parameters corresponding to the CT image to be processed;

[0085] A magnetic resonance image acquisition unit, which obtains a magnetic resonance image of the designated examination site through magnetic resonance imaging;

[0086] A peripheral nerve characteristic information acquisition unit is used to extract peripheral nerve characteristics from the magnetic resonance image to obtain peripheral nerve characteristic information of the designated inspection site;

[0087] a marked CT image determining unit, which marks the peripheral nerves in the CT image to be processed based on the characteristic information of the peripheral nerves at the designated inspection site to obtain a marked CT image;

[0088] The peripheral nerve course determination unit inputs the marked CT image and the CT scanning parameters into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site, wherein the peripheral nerve automatic generation model is a neural network model.

[0089] One or more embodiments of this specification provide a peripheral nerve routing automatic generation device, including:

[0090] at least one processor; and,

[0091] a memory communicatively connected to the at least one processor; wherein,

[0092] The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0093] Acquire a CT image to be processed of a designated examination part obtained through CT scanning;

[0094] Acquiring CT scanning parameters corresponding to the CT image to be processed;

[0095] Obtaining a magnetic resonance image of the designated examination site by magnetic resonance imaging;

[0096] Extracting peripheral nerve features from the magnetic resonance image to obtain peripheral nerve feature information of the designated inspection site;

[0097] Based on the peripheral nerve characteristic information of the designated inspection site, the peripheral nerves in the CT image to be processed are marked to obtain a marked CT image;

[0098] The marked CT image and the CT scanning parameters are input into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site, wherein the peripheral nerve automatic generation model is a neural network model.

[0099] One or more embodiments of this specification provide a non-volatile computer storage medium storing computer executable instructions, which can achieve the following when executed by a computer:

[0100] Acquire a CT image to be processed of a designated examination part obtained through CT scanning;

[0101] Acquiring CT scanning parameters corresponding to the CT image to be processed;

[0102] Obtaining a magnetic resonance image of the designated examination site by magnetic resonance imaging;

[0103] Extracting peripheral nerve features from the magnetic resonance image to obtain peripheral nerve feature information of the designated inspection site;

[0104] Based on the peripheral nerve characteristic information of the designated inspection site, the peripheral nerves in the CT image to be processed are marked to obtain a marked CT image;

[0105] The marked CT image and the CT scanning parameters are input into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site, wherein the peripheral nerve automatic generation model is a neural network model.

[0106] At least one of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects:

[0107] Improved accuracy and consistency: By automatically generating peripheral nerve courses, it reduces reliance on physicians’ personal experience and reduces the variability of outline results between different physicians, thereby improving the accuracy and consistency of treatment.

[0108] Lowering the requirements for physicians’ professional level: Traditional methods require physicians to have rich anatomical knowledge and clinical experience, while automatic generation methods can reduce the burden on physicians, enabling more physicians to perform peripheral nerve delineation and improving the utilization efficiency of medical resources.

[0109] Shorten the learning curve: For beginners, it is difficult to master the traditional experience-based peripheral nerve delineation method, while the automatic generation method can provide a more intuitive and easier to learn way, shortening the learning curve and helping to improve the skill level of physicians.

[0110] Improve work efficiency: Automatic generation of peripheral nerve routes can save physicians’ time and energy, allowing them to focus more on other important clinical work and improve work efficiency.

[0111] Provide better support for clinical treatment: Accurate information on the course of peripheral nerves is of great significance for formulating treatment plans and evaluating treatment effects. Automatic generation methods can provide a more reliable basis for clinical treatment and help improve the effectiveness and safety of treatment.

[0112] Promote medical research: The method of automatically generating the course of peripheral nerves can provide more data and information for medical research, help to gain a deeper understanding of the anatomical structure and physiological function of peripheral nerves, and promote the development of the medical field. BRIEF DESCRIPTION OF THE DRAWINGS

[0113] In order to more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art description. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. In the drawings:

[0114] Figure 1 A schematic flow chart of a method for automatically generating a peripheral nerve course provided in one or more embodiments of this specification;

[0115] Figure 2 A schematic diagram of the structure of a peripheral nerve course automatic generation device provided by one or more embodiments of this specification;

[0116] Figure 3 A schematic diagram of the structure of a peripheral nerve routing automatic generation device provided in one or more embodiments of this specification. DETAILED DESCRIPTION

[0117] The embodiments of this specification provide a method, device, equipment and medium for automatically generating the course of peripheral nerves.

[0118] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0119] Figure 1 A schematic diagram of a flow chart of a method for automatically generating a peripheral nerve course provided in one or more embodiments of this specification, the flow chart can be executed by a system for automatically generating a peripheral nerve course. Certain input parameters or intermediate results in the flow chart allow manual intervention and adjustment to help improve accuracy.

[0120] The method steps of the embodiment of this specification are as follows:

[0121] S101, obtaining a CT image to be processed of a designated examination part obtained by CT scanning.

[0122] In the embodiments of the present specification, the designated inspection part for CT scanning can be determined according to preset requirements, and a to-be-processed CT image of the designated inspection part can be obtained through CT scanning.

[0123] S102: Acquire CT scanning parameters corresponding to the CT image to be processed.

[0124] In the embodiments of this specification, the above-mentioned acquisition of CT scanning parameters can be implemented by the following specific implementation scheme:

[0125] Reading from image files: CT images are usually stored in specific file formats, such as DICOM (Digital Imaging and Communications in Medicine). These files may contain information about the scan parameters. Appropriate image processing libraries or tools can be used to read DICOM files and extract the scan parameters.

[0126] Communicate with the CT device: If the CT device supports a communication interface, the scan parameters can be obtained by communicating with the device. This may require the use of specific communication protocols and software to interact with the device and request the scan parameters.

[0127] S103, obtaining a magnetic resonance image of the designated examination site through magnetic resonance imaging.

[0128] In the embodiments of this specification, the acquisition of magnetic resonance images can be implemented by the following specific implementation schemes:

[0129] MRI (nuclear magnetic resonance imaging) sequence selection: Select the appropriate MRI sequence according to the examination site and clinical needs. Common sequences include T1-weighted, T2-weighted, proton density-weighted, etc. Special sequences such as fat suppression and diffusion weighted can be added as needed to obtain more information.

[0130] Parameter setting: Set the parameters of MRI scanning, such as magnetic field strength, pulse sequence, repetition time, echo time, etc. Adjust the parameters to optimize the image quality according to the characteristics of the examination area.

[0131] Scan execution: Start the MRI device and scan according to the set parameters. If necessary, multiple scans may be required to obtain images at different angles or levels.

[0132] Image reconstruction and post-processing: After the MRI scan is completed, the acquired data is reconstructed into a magnetic resonance image using an image reconstruction algorithm. The image can be post-processed to improve image quality, such as enhancing contrast, removing noise, and adjusting brightness.

[0133] S104, extracting peripheral nerve features from the magnetic resonance image to obtain peripheral nerve feature information of the designated examination site.

[0134] In an embodiment of the present specification, a suitable segmentation algorithm can be determined according to the peripheral nerve characteristics of the designated inspection site; the magnetic resonance image can be segmented according to the suitable segmentation algorithm to obtain a segmented peripheral nerve image; based on the characteristics of the peripheral nerves, peripheral nerve features can be extracted from the segmented peripheral nerve image to obtain peripheral nerve feature information of the designated inspection site.

[0135] It should be noted that the above content can be implemented through the following specific implementation plans:

[0136] Peripheral nerve characteristic analysis: Analyze the anatomical structure, morphological characteristics and signal characteristics of the peripheral nerves at the designated examination site to obtain peripheral nerve characteristics such as the contrast and texture between the peripheral nerves and surrounding tissues.

[0137] Segmentation algorithm selection: Select a suitable segmentation algorithm based on the characteristics of peripheral nerves. Common segmentation algorithms include threshold segmentation, region growing, edge detection, model-based segmentation, etc. You can consider using multiple segmentation algorithms for comparison and verification, and select the algorithm with the best effect.

[0138] Magnetic resonance image preprocessing: Preprocess the magnetic resonance images, such as denoising and contrast enhancement, to improve image quality and segmentation effect. Technologies such as image filtering and histogram equalization can be used.

[0139] Application of segmentation algorithm: Use the selected segmentation algorithm to segment the preprocessed magnetic resonance image to obtain the segmented peripheral nerve image. According to the requirements of the algorithm, set appropriate parameters and thresholds.

[0140] Segmentation result evaluation: Evaluate the segmentation results and use quantitative indicators such as precision, recall, Dice coefficient, etc. to evaluate the accuracy of segmentation. The performance of the segmentation algorithm can be verified by comparing with the manual segmentation results.

[0141] Peripheral nerve feature extraction: Based on the characteristics of peripheral nerves, select appropriate feature extraction methods. For example, the shape, size, position, texture and other features of peripheral nerves can be extracted. Use image processing and computer vision techniques, such as morphological operations, feature descriptors, etc. to extract peripheral nerve features.

[0142] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0143] Improve the accuracy of peripheral nerve feature extraction: By selecting an appropriate segmentation algorithm based on the characteristics of the peripheral nerves in the specified examination site, the peripheral nerves can be segmented from the magnetic resonance image more accurately. This helps to reduce the influence of background noise and other interference factors, thereby improving the accuracy of peripheral nerve feature extraction.

[0144] Enhanced adaptability to different inspection sites: Peripheral nerves at different inspection sites may have different characteristics, such as shape, size, texture, etc. Selecting an appropriate segmentation algorithm based on these characteristics can better adapt to different inspection sites and improve the effect of feature extraction.

[0145] Improve work efficiency: Feature extraction based on the characteristics of peripheral nerves can reduce unnecessary calculations and processing and improve work efficiency. At the same time, accurate feature extraction also helps reduce the time and cost of subsequent analysis and diagnosis.

[0146] S105, based on the peripheral nerve feature information of the designated inspection site, the peripheral nerves in the CT image to be processed are marked to obtain a marked CT image.

[0147] In an embodiment of the present specification, the CT image to be processed can be registered with the magnetic resonance image to obtain a registration result; the peripheral nerve characteristic information of the designated examination site can be extracted; and based on the registration result, the peripheral nerve characteristic information of the designated examination site can be marked on the CT image to be processed to obtain the marked CT image.

[0148] It should be noted that the above content can be implemented through the following specific implementation plans:

[0149] Image registration: Select an appropriate image registration algorithm, such as feature point-based registration, grayscale-based registration, or deformation model-based registration. Preprocess the CT images and MRI images to be processed, such as denoising and contrast enhancement, to improve the accuracy of registration. Select common anatomical structures or feature points in the CT image and MRI image as reference points for registration. Use the registration algorithm to align the CT image and MRI image to obtain the registration result.

[0150] Peripheral nerve feature extraction: Analyze the characteristics of peripheral nerves at the specified examination site in the magnetic resonance image, such as shape, size, location, signal strength, etc. Use image processing and computer vision techniques, such as threshold segmentation, edge detection, morphological operations, etc. to extract the characteristic information of peripheral nerves. Deep learning algorithms or model-based methods can be used to improve the accuracy and automation of feature extraction.

[0151] Feature information marking: Based on the registration results, the extracted peripheral nerve feature information is mapped to the CT image to be processed. Use marking tools or algorithms to mark the location, shape and other features of the peripheral nerves in the CT image. Different colors, symbols or marking methods can be used to distinguish different peripheral nerves or features.

[0152] Result evaluation: Evaluate the labeled CT images to check the accuracy and completeness of the peripheral nerve feature information. Communicate with doctors or experts to ensure that the labeling results meet clinical needs and diagnostic requirements. Adjust and optimize the registration algorithm, feature extraction method or labeling process based on the evaluation results.

[0153] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0154] Improve the accuracy of marking: By registering the CT image to be processed with the MRI image, the characteristic information of the peripheral nerve can be more accurately marked on the CT image. This helps to reduce the error of marking and improve the accuracy of marking.

[0155] Enhanced image visualization: Marking CT images can make peripheral nerves appear more clearly in the image, enhancing the visualization of the image. This helps doctors observe the location and morphology of peripheral nerves more intuitively, thereby better diagnosing and treating them.

[0156] Improve the accuracy of diagnosis: Accurate peripheral nerve marking can provide doctors with more detailed and accurate diagnostic information, which helps to improve the accuracy of diagnosis. This is of great significance for the early detection and treatment of diseases.

[0157] Furthermore, when registering the CT image to be processed with the magnetic resonance image to obtain the registration result, an image registration algorithm may be used to register the spatial positions of the CT image to be processed with the magnetic resonance image to obtain the registration result.

[0158] It should be noted that the above content can be implemented through the following specific implementation plans:

[0159] Registration process: Use the selected registration algorithm to register the CT image and the MRI image. During the registration process, the spatial position of the image is continuously adjusted according to the similarity metric until the best registration result is achieved. Iterative optimization methods can be used to improve the accuracy of the registration.

[0160] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0161] Improve diagnostic accuracy: The registered image can more accurately reflect the spatial relationship between different tissues and structures, and can more accurately mark the peripheral nerve characteristic information of the specified examination site to the CT image to be processed to obtain a marked CT image.

[0162] S106, inputting the marked CT image and the CT scanning parameters into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site, wherein the peripheral nerve automatic generation model is a neural network model.

[0163] It should be noted that the peripheral nerves in the embodiments of this specification may be cranial nerves and other important peripheral nerves, and other important peripheral nerves may include superior laryngeal nerve, laryngeal reflex nerve, paralysis nerve, phrenic nerve, intercostal nerve, etc. Head and neck radiotherapy requires clear identification of the pores and courses of cranial nerves entering and exiting the skull, one is to avoid target area residues and the other is to avoid nerve damage after high-dose irradiation. There are 12 pairs of cranial nerves, and important brain parenchyma parts can be expanded, including the hippocampus, thalamus, brainstem, etc.

[0164] At the same time, for peripheral nerves, dose calculation and injury assessment of the brachial plexus, superior laryngeal nerve, recurrent laryngeal nerve, phrenic nerve, and pudendal nerve are very important.

[0165] Furthermore, before the marked CT image and the CT scanning parameters are input into the pre-trained automatic generation model of peripheral nerves, a plurality of first CT images of different inspection parts obtained by CT scanning can be obtained, and a first magnetic resonance image corresponding to each first CT image obtained by magnetic resonance imaging can be obtained; the CT scanning parameters corresponding to each first CT image can be obtained; peripheral nerve feature extraction is performed on the plurality of first magnetic resonance images respectively to obtain peripheral nerve feature information of each first magnetic resonance image; based on the peripheral nerve feature information of each first magnetic resonance image, the peripheral nerves in the corresponding first CT image are marked respectively to obtain the corresponding first marked CT image; the first image with clear peripheral nerve course obtained by other inspection methods of the different inspection parts is obtained; according to the clear peripheral nerve course in each first image, the peripheral nerve course is marked on the first marked CT image respectively to obtain a control CT image with clear peripheral nerve course; the CT scanning parameters corresponding to each first CT image, the first marked CT image and the control CT image are used to construct a data set; the automatic generation model of peripheral nerves is trained based on the data set to obtain the automatic generation model of peripheral nerves that meets the preset requirements.

[0166] It should be noted that the above content can be implemented through the following specific implementation plans:

[0167] Data collection: Collect multiple first CT images of different examination parts obtained through CT scanning. At the same time, obtain the first magnetic resonance image corresponding to each first CT image obtained through magnetic resonance imaging. Record the CT scanning parameters corresponding to each first CT image, such as scanning layer thickness, tube voltage, tube current, etc.

[0168] Peripheral nerve feature extraction: Peripheral nerve feature extraction is performed on multiple first magnetic resonance images respectively. Image processing and computer vision techniques, such as threshold segmentation, edge detection, morphological operations, etc., can be used to extract feature information of peripheral nerves. The extracted peripheral nerve feature information may include the location, shape, size, direction, etc. of the peripheral nerves.

[0169] Peripheral nerve marking: Based on the peripheral nerve feature information of each first magnetic resonance image, the peripheral nerves in the corresponding first CT image are marked respectively. Manual marking or semi-automatic marking methods can be used to mark the position and contour of the peripheral nerves on the first CT image to obtain the corresponding first marked CT image.

[0170] Obtain control images: Obtain first images with clear peripheral nerve courses obtained by other examination methods at different examination sites. These images can be used as controls to annotate the courses of peripheral nerves. Control images can be obtained by other imaging techniques (such as magnetic resonance neurography, ultrasound imaging, etc.), or images annotated by professional doctors.

[0171] Marking the course of peripheral nerves: According to the clear course of peripheral nerves in each first image, mark the course of peripheral nerves on the first marked CT image. Manual marking or semi-automatic marking method can be used to draw the course of peripheral nerves on the first marked CT image. Obtain a control CT image with clear course of peripheral nerves.

[0172] Dataset construction: The CT scanning parameters corresponding to each first CT image, the first marked CT image and the control CT image are constructed into a data set. Each sample in the data set should contain CT scanning parameters, marked peripheral nerve images and annotated peripheral nerve course images.

[0173] Model training: Use the constructed data set to train the peripheral nerve automatic generation model. You can choose a suitable deep learning model, such as convolutional neural network (CNN), recurrent neural network (RNN), etc. The data set is divided into training set, validation set and test set for model training, evaluation and verification. During the training process, adjust the parameters of the model according to the preset requirements (such as accuracy, recall rate, etc.) to improve the performance of the model.

[0174] Model evaluation and optimization: Use the validation set to evaluate the trained model and calculate the model's accuracy, recall, F1 value and other indicators. Based on the evaluation results, optimize and improve the model, such as adjusting the model structure, adding training data, using more complex features, etc. Repeat the training and evaluation process until the model meets the preset requirements.

[0175] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0176] Improve the accuracy of the automatic generation model of peripheral nerves: By acquiring a large number of CT images and magnetic resonance images of different examination sites, extracting and labeling peripheral nerve features, and building a data set for training, the automatic generation model of peripheral nerves can learn more features and patterns, thereby improving its accuracy and reliability.

[0177] Enhance the generalization ability of the model: Using images of different examination parts for training can enable the model to learn the characteristics of peripheral nerves in different situations, enhance its generalization ability, and better handle various complex situations.

[0178] Improve image quality and repeatability: By extracting peripheral nerve features from magnetic resonance images and marking and annotating them on CT images, the quality and repeatability of images can be improved, providing a more reliable basis for subsequent diagnosis and treatment.

[0179] Promoting multimodal image fusion: Combining CT images and magnetic resonance images for analysis and processing can promote multimodal image fusion and provide doctors with more comprehensive and accurate information.

[0180] Improve work efficiency: The automatic generation model of peripheral nerves can quickly and accurately generate peripheral nerve images, greatly improving work efficiency and reducing the workload of doctors.

[0181] Furthermore, when the automatic generation model of peripheral nerves is trained based on the data set and the automatic generation model of peripheral nerves that meets the preset requirements is obtained through training, the CT scanning parameters corresponding to each first CT image and the first marked CT image can be input into the initial automatic generation model of peripheral nerves to obtain the result image to be verified of the peripheral nerve course of the corresponding examination site; each result image to be verified is compared with the corresponding control CT image to obtain a comparison result; if the comparison results all meet the preset requirements, the automatic generation model of peripheral nerves that meets the preset requirements is obtained through training.

[0182] It should be noted that the above content can be implemented through the following specific implementation plans:

[0183] Model training: The CT scan parameters corresponding to each first CT image and the first labeled CT image are input into the initial peripheral nerve automatic generation model. The model is trained using a suitable training algorithm, such as stochastic gradient descent (SGD) or adaptive moment estimation (Adam). During the training process, the parameters of the model are adjusted to minimize the loss function and improve the performance of the model.

[0184] Generate the result image to be verified: Use the trained peripheral nerve automatic generation model to predict each first CT image to obtain the result image to be verified of the peripheral nerve course of the corresponding examination site.

[0185] Comparison and evaluation: Compare each result image to be verified with the corresponding control CT image. Image similarity metrics, such as mean square error (MSE) and structural similarity index (SSIM), can be used to evaluate the similarity between the result image to be verified and the control CT image. According to the preset requirements, determine whether the comparison results meet the requirements. If all comparison results meet the preset requirements, it is considered that the training has obtained a peripheral nerve automatic generation model that meets the preset requirements.

[0186] Model adjustment and optimization: If the comparison results do not meet the preset requirements, you can adjust and optimize the model. You can try to improve the performance of the model by adjusting the model architecture, increasing the amount of training data, using more complex features, etc. Repeat the above steps until the comparison results meet the preset requirements.

[0187] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0188] Improve the accuracy of the model: By comparing the result image to be verified with the control CT image, the difference between the peripheral nerve course generated by the model and the actual situation can be found, so as to continuously adjust the model parameters and improve the accuracy of the model.

[0189] Enhance the reliability of the model: The comparison results meet the preset requirements, indicating that the peripheral nerve course generated by the model has high reliability and can provide doctors with more accurate diagnostic information.

[0190] Promote model optimization: Based on the comparison results, the model can be optimized in a targeted manner to further improve its performance and expression.

[0191] Improve medical efficiency: Accurate automatic generation model of peripheral nerves can quickly generate peripheral nerve course images, reduce doctors' workload and improve medical efficiency.

[0192] Furthermore, when comparing each result image to be verified with the corresponding control CT image to obtain a comparison result, the first preset key point of the peripheral nerve course in each result image to be verified is extracted; the second key point corresponding to the first preset key point of the peripheral nerve course in the corresponding control CT image is extracted; and the comparison result is obtained based on comparing the first preset key point of each result image to be verified with the second key point of the corresponding control CT image.

[0193] It should be noted that the above content can be implemented through the following specific implementation plans:

[0194] Key point extraction: For each result image to be verified, use image processing technology to extract the first preset key point of the peripheral nerve course. These key points can be representative locations such as the starting point, end point, and turning point of the peripheral nerve. Similarly, for the corresponding control CT image, extract the second key point corresponding to the first preset key point.

[0195] Key point matching: Match the first preset key point of each result image to be verified with the second key point of the corresponding control CT image. Feature matching algorithms, such as SIFT (Scale-Invariant Feature Transform) or SURF (Speeded-Up Robust Features), can be used to find the correspondence between key points.

[0196] Comparison and evaluation: Comparison and evaluation are performed based on the matching results of key points. The distance, angle and other differences between key points can be calculated, and the comparison results can be judged based on the preset threshold or standard to see whether they meet the requirements. For example, a distance threshold can be set. If the distance between key points is less than the threshold, the comparison result is considered to meet the requirements; otherwise, it is considered not to meet the requirements.

[0197] Comprehensive evaluation: In addition to the comparison of key points, other factors can be considered to comprehensively evaluate the comparison results. For example, the overall shape of the peripheral nerve course, the smoothness of the curve, etc. can be compared. Image similarity metrics, such as mean square error (MSE) and structural similarity index (SSIM), can be used to evaluate the overall similarity between the result image to be verified and the control CT image.

[0198] Result determination: Based on the results of key point comparison and comprehensive evaluation, determine whether the comparison results meet the preset requirements. If the comparison results of all key points meet the requirements and the comprehensive evaluation results are also good, it can be considered that the comparison results meet the preset requirements. Otherwise, further analysis and adjustment of the model or data are required to improve the accuracy of the comparison results.

[0199] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0200] Improve the accuracy of key points: By extracting the first preset key points and the corresponding second key points of the peripheral nerve course, the results generated by the model can be compared with the actual situation more accurately, thereby discovering possible differences.

[0201] Enhance the comparability of results: Using key points for comparison can make the comparison between different images more objective and comparable, and reduce the interference of other factors.

[0202] Improve the interpretability of the model: By analyzing the differences in key points, we can better understand the decision-making process and output results of the model, providing a basis for improving and optimizing the model.

[0203] Discover potential problems and errors: Comparing key points can help discover potential problems and errors in the peripheral nerve routing generated by the model, and make timely adjustments and improvements.

[0204] It should be noted that there is a mutual connection between peripheral nerves and various organs, and the positional relationship between peripheral nerves and organs is crucial for accurately generating the result image of the peripheral nerve course. If this factor is not fully considered when generating the result image of the peripheral nerve course, it may lead to inaccurate result images of the peripheral nerve course, or even wrong conclusions. Therefore, when generating the result image of the peripheral nerve course, it is necessary to carefully analyze the positional relationship between the peripheral nerves and organs, and process and analyze according to this positional relationship to ensure the accuracy and reliability of the result image of the peripheral nerve course.

[0205] Based on the above content, in the embodiment of this specification, when the marked CT image and the CT scanning parameters are input into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site, the positional relationship between the designated inspection site and the corresponding one or more organs can be first obtained, and the positional relationship includes spatial distance, angle and relative position; the positional relationship between the designated inspection site and the corresponding one or more organs, the marked CT image and the CT scanning parameters are input into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site.

[0206] It should be noted that the above content can be implemented through the following specific implementation plans:

[0207] 1. Data Collection and Preprocessing

[0208] Patient data: Collect the patient's CT scan data to ensure that the data contains the peripheral nerves and the corresponding one or more organs at the designated examination site.

[0209] Image marking: Manual or semi-automatic marking of CT images to mark the contours of peripheral nerves and corresponding organs.

[0210] 2. Position relationship analysis

[0211] Spatial distance calculation: Use image processing algorithms (such as convolutional neural networks (CNNs) in deep learning) to extract the boundaries of peripheral nerves and organs from labeled CT images. Calculate spatial distance parameters such as the shortest distance and closest point between peripheral nerves and organs.

[0212] Angle calculation: Determine the angle relationship between peripheral nerves and organs through geometric algorithms or computer vision technology.

[0213] Relative position analysis: Analyze the direction of peripheral nerves relative to organs, such as passing through, bypassing or running parallel.

[0214] 3. Model Preparation

[0215] Model selection: Select or develop a deep learning model suitable for peripheral nerve course prediction, such as a 3D convolutional neural network based on U-Net.

[0216] Model training: The model is trained using a large amount of labeled data, including the path of peripheral nerves and the corresponding organ position relationship. CT scan parameters are used as additional input to adapt to different scanning conditions.

[0217] 4. Model Input and Prediction

[0218] Data integration: The calculated positional relationship parameters, labeled CT images, and CT scan parameters are integrated as model input.

[0219] Predicting the course of peripheral nerves: The integrated data is input into a pre-trained peripheral nerve automatic generation model. The model outputs the predicted peripheral nerve course image.

[0220] It should be noted that the embodiments of this specification have the following beneficial effects through the above contents:

[0221] Improve diagnostic accuracy: By obtaining the spatial distance, angle and relative position between the designated examination site and the corresponding one or more organs, the anatomical relationship between the peripheral nerves and organs can be determined more accurately, thereby improving the accuracy of diagnosing peripheral nerve diseases.

[0222] Reduce misdiagnosis and missed diagnosis: The model automatically generates images of peripheral nerves, which can reduce misdiagnosis caused by complex anatomical structures or subjective judgment errors of doctors. At the same time, the model can comprehensively analyze the relationship between peripheral nerves and organs, which helps to discover potential missed diagnoses.

[0223] Optimize treatment plans: Accurate images of peripheral nerves are essential for surgical planning, which can reduce surgical risks and optimize surgical pathways. At the same time, for patients who need drug treatment, understanding the precise relationship between peripheral nerves and organs helps to choose the most appropriate drugs.

[0224] Improve work efficiency: Using pre-trained models can greatly reduce the time of manual analysis and improve work efficiency. At the same time, the model can process a large number of CT images and is suitable for large-scale medical research and clinical applications.

[0225] Figure 2 A schematic diagram of the structure of a peripheral nerve course automatic generation device provided for one or more embodiments of the present specification includes: a CT image acquisition unit 201, a scanning parameter acquisition unit 202, a magnetic resonance image acquisition unit 203, a peripheral nerve characteristic information acquisition unit 204, a marked CT image determination unit 205 and a peripheral nerve course determination unit 206.

[0226] A CT image acquisition unit 201 acquires a to-be-processed CT image of a designated inspection part obtained by CT scanning;

[0227] A scanning parameter acquisition unit 202 is used to acquire CT scanning parameters corresponding to the CT image to be processed;

[0228] A magnetic resonance image acquisition unit 203, which obtains a magnetic resonance image of the designated examination part through magnetic resonance imaging;

[0229] The peripheral nerve characteristic information acquisition unit 204 extracts peripheral nerve characteristics from the magnetic resonance image to obtain peripheral nerve characteristic information of the designated examination site;

[0230] A marked CT image determining unit 205 is configured to mark peripheral nerves in the CT image to be processed based on the peripheral nerve feature information of the designated inspection site to obtain a marked CT image;

[0231] The peripheral nerve course determination unit 206 inputs the marked CT image and the CT scanning parameters into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site, wherein the peripheral nerve automatic generation model is a neural network model.

[0232] Figure 3 A schematic diagram of a peripheral nerve routing automatic generation device provided for one or more embodiments of this specification includes:

[0233] at least one processor; and,

[0234] a memory communicatively connected to the at least one processor; wherein,

[0235] The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0236] Acquire a CT image to be processed of a designated examination part obtained through CT scanning;

[0237] Acquiring CT scanning parameters corresponding to the CT image to be processed;

[0238] Obtaining a magnetic resonance image of the designated examination site by magnetic resonance imaging;

[0239] Extracting peripheral nerve features from the magnetic resonance image to obtain peripheral nerve feature information of the designated inspection site;

[0240] Based on the peripheral nerve characteristic information of the designated inspection site, the peripheral nerves in the CT image to be processed are marked to obtain a marked CT image;

[0241] The marked CT image and the CT scanning parameters are input into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site, wherein the peripheral nerve automatic generation model is a neural network model.

[0242] One or more embodiments of this specification provide a non-volatile computer storage medium storing computer executable instructions, which can achieve the following when executed by a computer:

[0243] Acquire a CT image to be processed of a designated examination part obtained through CT scanning;

[0244] Acquiring CT scanning parameters corresponding to the CT image to be processed;

[0245] Obtaining a magnetic resonance image of the designated examination site by magnetic resonance imaging;

[0246] Extracting peripheral nerve features from the magnetic resonance image to obtain peripheral nerve feature information of the designated inspection site;

[0247] Based on the peripheral nerve characteristic information of the designated inspection site, the peripheral nerves in the CT image to be processed are marked to obtain a marked CT image;

[0248] The marked CT image and the CT scanning parameters are input into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site, wherein the peripheral nerve automatic generation model is a neural network model.

[0249] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment, and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

Claims

1. A method for automatically generating the course of peripheral nerves, characterized in that: include: Acquire a CT image to be processed of a designated examination part obtained through CT scanning; Acquiring CT scanning parameters corresponding to the CT image to be processed; Obtaining a magnetic resonance image of the designated examination site by magnetic resonance imaging; Extracting peripheral nerve features from the magnetic resonance image to obtain peripheral nerve feature information of the designated inspection site; Based on the peripheral nerve characteristic information of the designated inspection site, the peripheral nerves in the CT image to be processed are marked to obtain a marked CT image; Inputting the marked CT image and the CT scanning parameters into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site, wherein the peripheral nerve automatic generation model is a neural network model; Before inputting the marked CT image and the CT scanning parameters into a pre-trained peripheral nerve automatic generation model, the method further includes: Acquire a plurality of first CT images of different examination parts obtained by CT scanning, and acquire a first magnetic resonance image corresponding to each first CT image obtained by magnetic resonance imaging; Acquire CT scanning parameters corresponding to each first CT image; Extracting peripheral nerve features from the plurality of first magnetic resonance images respectively to obtain peripheral nerve feature information of each first magnetic resonance image; Based on the peripheral nerve feature information of each first magnetic resonance image, the peripheral nerves in the corresponding first CT image are marked respectively to obtain a corresponding first marked CT image; Acquire a first image with clear peripheral nerve courses obtained by other examination methods at the different examination sites; According to the clear peripheral nerve courses in each first image, the courses of the peripheral nerves are respectively marked on the first marked CT images to obtain a control CT image with clear courses of the peripheral nerves; Constructing a data set by using the CT scanning parameters, the first labeled CT images and the control CT images corresponding to each first CT image; The peripheral nerve automatic generation model is trained based on the data set to obtain a peripheral nerve automatic generation model that meets preset requirements.

2. The method according to claim 1, characterized in that The step of extracting peripheral nerve features from the magnetic resonance image to obtain peripheral nerve feature information of the designated inspection site includes: Determining a suitable segmentation algorithm according to the peripheral nerve characteristics of the designated examination site; Segmenting the magnetic resonance image according to the adapted segmentation algorithm to obtain a segmented peripheral nerve image; Based on the characteristics of the peripheral nerves, peripheral nerve features are extracted from the segmented peripheral nerve image to obtain peripheral nerve feature information of the designated inspection site.

3. The method according to claim 1, characterized in that The step of marking the peripheral nerves in the CT image to be processed based on the peripheral nerve feature information of the designated inspection site to obtain a marked CT image includes: Registering the CT image to be processed with the magnetic resonance image to obtain a registration result; Extracting peripheral nerve characteristic information of the designated inspection site; According to the registration result, the peripheral nerve feature information of the designated inspection site is marked on the CT image to be processed to obtain the marked CT image.

4. The method according to claim 3, characterized in that The registering the to-be-processed CT image with the magnetic resonance image to obtain a registration result includes: An image registration algorithm is used to register the spatial positions of the CT image to be processed and the magnetic resonance image to obtain the registration result.

5. The method according to claim 1, characterized in that The method of training the automatic generation model of peripheral nerves based on the data set to obtain the automatic generation model of peripheral nerves that meets preset requirements includes: Inputting the CT scanning parameters corresponding to each first CT image and the first marked CT image into the initial peripheral nerve automatic generation model to obtain a result image to be verified of the peripheral nerve course of the corresponding examination site; Comparing each result image to be verified with the corresponding control CT image to obtain a comparison result; If the comparison results all meet the preset requirements, a peripheral nerve automatic generation model that meets the preset requirements is trained.

6. The method according to claim 5, characterized in that The step of comparing each result image to be verified with the corresponding control CT image to obtain a comparison result includes: Extracting the first preset key point of the peripheral nerve course in each result image to be verified; Extracting a second key point corresponding to the first preset key point of the peripheral nerve course in the corresponding control CT image; The comparison result is obtained based on comparing the first preset key point of each result image to be verified with the second key point of the corresponding control CT image.

7. The method according to claim 1, characterized in that The step of inputting the marked CT image and the CT scanning parameters into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site includes: Acquire the positional relationship between the designated inspection site and the corresponding one or more organs, wherein the positional relationship includes spatial distance, angle and relative position; The positional relationship between the designated inspection site and the corresponding one or more organs, the marked CT image and the CT scanning parameters are input into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site.

8. A device for automatically generating the course of peripheral nerves, characterized in that: include: A CT image acquisition unit, which acquires a CT image to be processed of a designated inspection part obtained through CT scanning; A scanning parameter acquisition unit, which acquires CT scanning parameters corresponding to the CT image to be processed; A magnetic resonance image acquisition unit, which obtains a magnetic resonance image of the designated examination site through magnetic resonance imaging; A peripheral nerve characteristic information acquisition unit is used to extract peripheral nerve characteristics from the magnetic resonance image to obtain peripheral nerve characteristic information of the designated inspection site; a marked CT image determining unit, which marks the peripheral nerves in the CT image to be processed based on the characteristic information of the peripheral nerves at the designated inspection site to obtain a marked CT image; A peripheral nerve course determination unit inputs the marked CT image and the CT scanning parameters into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site, wherein the peripheral nerve automatic generation model is a neural network model; Before inputting the marked CT image and the CT scanning parameters into the pre-trained peripheral nerve automatic generation model, the method further includes: Acquire a plurality of first CT images of different examination parts obtained by CT scanning, and acquire a first magnetic resonance image corresponding to each first CT image obtained by magnetic resonance imaging; Acquire CT scanning parameters corresponding to each first CT image; Extracting peripheral nerve features from the plurality of first magnetic resonance images respectively to obtain peripheral nerve feature information of each first magnetic resonance image; Based on the peripheral nerve feature information of each first magnetic resonance image, the peripheral nerves in the corresponding first CT image are marked respectively to obtain a corresponding first marked CT image; Acquire a first image with clear peripheral nerve courses obtained by other examination methods at the different examination sites; According to the clear peripheral nerve courses in each first image, the courses of the peripheral nerves are respectively marked on the first marked CT images to obtain a control CT image with clear courses of the peripheral nerves; Constructing a data set by using the CT scanning parameters, the first labeled CT images and the control CT images corresponding to each first CT image; The peripheral nerve automatic generation model is trained based on the data set to obtain a peripheral nerve automatic generation model that meets preset requirements.

9. A peripheral nerve course automatic generation device, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: Acquire a CT image to be processed of a designated examination part obtained through CT scanning; Acquiring CT scanning parameters corresponding to the CT image to be processed; Obtaining a magnetic resonance image of the designated examination site by magnetic resonance imaging; Extracting peripheral nerve features from the magnetic resonance image to obtain peripheral nerve feature information of the designated inspection site; Based on the peripheral nerve characteristic information of the designated inspection site, the peripheral nerves in the CT image to be processed are marked to obtain a marked CT image; Inputting the marked CT image and the CT scanning parameters into a pre-trained peripheral nerve automatic generation model to obtain a result image of the peripheral nerve course of the designated inspection site, wherein the peripheral nerve automatic generation model is a neural network model; Before inputting the marked CT image and the CT scanning parameters into the pre-trained peripheral nerve automatic generation model, the method further includes: Acquire a plurality of first CT images of different examination parts obtained by CT scanning, and acquire a first magnetic resonance image corresponding to each first CT image obtained by magnetic resonance imaging; Acquire CT scanning parameters corresponding to each first CT image; Extracting peripheral nerve features from the plurality of first magnetic resonance images respectively to obtain peripheral nerve feature information of each first magnetic resonance image; Based on the peripheral nerve feature information of each first magnetic resonance image, the peripheral nerves in the corresponding first CT image are marked respectively to obtain a corresponding first marked CT image; Acquire a first image with clear peripheral nerve courses obtained by other examination methods at the different examination sites; According to the clear peripheral nerve courses in each first image, the courses of the peripheral nerves are respectively marked on the first marked CT images to obtain a control CT image with clear courses of the peripheral nerves; Constructing a data set by using the CT scanning parameters, the first labeled CT images and the control CT images corresponding to each first CT image; The peripheral nerve automatic generation model is trained based on the data set to obtain a peripheral nerve automatic generation model that meets preset requirements.

Images