Thyroid tumor detection segmentation and lymphatic metastasis state diagnosis method and system

By constructing the YOLOv8 detection model, HRNet segmentation model and Bayesian diagnostic classification model, the problems of low segmentation accuracy and high application cost in ultrasound images were solved, and high-precision thyroid tumor detection and neck lymph node metastasis status were achieved. Rapid diagnosis, suitable for clinical applications.

CN118967699APending Publication Date: 2024-11-15TONGJI HOSPITAL ATTACHED TO TONGJI MEDICAL COLLEGE HUAZHONG SCI TECH
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410938218.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

In existing ultrasound images, the segmentation accuracy of thyroid tumor detection is low, the application cost is high, and the rapid diagnosis of neck lymph node metastasis cannot be achieved.

Method used

By constructing the YOLOv8 detection model, HRNet segmentation model and Bayesian diagnostic classification model, combining data cleaning, privacy processing and annotation, the object detection, segmentation and diagnosis of ultrasound images can be realized, the implementation cost is reduced, and the detection segmentation accuracy is improved.

Benefits of technology

It realizes high-precision detection and segmentation of thyroid tumors and rapid diagnosis of neck lymph node metastasis status, reducing labor costs and is suitable for clinical applications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967699A_ABST
    Figure CN118967699A_ABST
Patent Text Reader

Abstract

The invention provides a thyroid tumor detection segmentation and lymphatic metastasis state diagnosis method and system, and the method comprises the steps: collecting thyroid tumor ultrasonic images, carrying out the data cleaning and privacy processing of the ultrasonic images, and selecting the ultrasonic images with the image quality meeting a preset requirement, and building a basic database; marking a thyroid tumor outer contour of the ultrasonic image in the basic database and a minimum enclosing rectangle corresponding to the outer contour; taking the marked ultrasonic image as input, performing target detection by constructing a YOLOv8 detection model, and segmenting a tumor edge through an HRNet segmentation model; the cervical lymph node metastasis state of the thyroid tumor is judged through semi-quantitative parameter measurement and a Bayesian diagnosis classification model; and based on the trained YOLOv8 detection model, the trained HRNet segmentation model and the trained Bayesian diagnosis classification model, carrying out detection segmentation on a to-be-detected ultrasonic image, and judging a cervical lymph node metastasis state. Through the scheme, the tumor detection segmentation precision can be improved, rapid diagnosis of the thyroid tumor neck lymph node metastasis state is realized, and the manpower and material resource cost is saved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of image recognition, and in particular relates to a method and system for thyroid tumor detection and segmentation and lymph node metastasis status diagnosis. Background Art

[0002] Ultrasound images are widely used in the clinical and scientific research of thyroid nodular diseases, and have become a necessary daily process in the diagnosis and treatment of thyroid nodular diseases in clinical practice. Ultrasound images can present and analyze the location, shape, and size of thyroid nodules, making them important in the early diagnosis of benign and malignant thyroid nodules, thyroid cancer screening, and the assessment of the status of cervical lymph nodes in thyroid cancer. Compared with CT and MRI, ultrasound has the advantages of being real-time, convenient, and affordable. It provides a non-invasive monitoring method, and evaluating the status of cervical lymph nodes in thyroid cancer helps to formulate the best surgical method and optimize the treatment plan, thereby further improving the patient's survival rate.

[0003] In clinical work, ultrasound doctors need to operate ultrasound probes and ultrasound equipment imaging parameters to obtain clear and visible thyroid ultrasound images in real time. At the same time, they also need to accurately identify the tumor area in the image based on clinical experience and measure the corresponding tumor area-related parameters and identify ultrasound features, such as: tumor size, internal echoes, echoes, boundaries, morphology, presence or absence of calcification, etc. In this process, whether it is real-time and accurate positioning of the thyroid tumor area or evaluating the tumor's cervical lymph node metastasis, it is affected by the ultrasound doctor's subjective experience, fatigue level, and psychological state, which may cause patients to miss the best treatment period.

[0004] Real-time detection and segmentation of tumors in ultrasound images and intelligent diagnosis are two key issues in clinical work. There are currently three main approaches to real-time detection and segmentation of tumors in ultrasound images: (1) traditional image filtering methods; (2) detection or segmentation methods based on deep learning; and (3) methods based on deep learning instance segmentation. These three methods can locate the tumor area to a certain extent, but there are the following problems that cannot be solved: (1) The data distribution of ultrasound images is affected by equipment parameters, operator and patient cooperation. Traditional image filtering methods cannot apply the same set of filtering parameters to all imaging scenarios, and their practical applications are limited; (2) Detection or segmentation methods based on deep learning are prone to false positive tumors, thereby increasing the workload of doctors; (3) The instance segmentation method based on deep learning can effectively reduce false positive tumors by performing segmentation detection through a feature map sharing mechanism based on the detection algorithm, but there are still some problems when applying this type of algorithm to ultrasound images. For example, the feature maps of instance segmentation are often small, the segmentation subnetwork is relatively simple, and it is impossible to segment the tumor contour in the ultrasound image very accurately; the instance segmentation algorithm usually couples the detection network and the segmentation network, and the network is relatively complex. When the detection and segmentation effect is not good, the entire network needs to be retrained and deployed, and the cost of implementation is high; it is impossible to achieve rapid diagnosis of the cervical lymph node metastasis status of thyroid tumors, etc.

[0005] Therefore, it is necessary to propose a method with high segmentation accuracy, low implementation cost and the ability to achieve thyroid tumor detection and segmentation and diagnosis of cervical lymph node metastasis status. Summary of the invention

[0006] In view of this, an embodiment of the present invention provides a thyroid tumor detection and segmentation and lymph node metastasis status diagnosis method and system, which is used to solve the problems of low tumor detection and segmentation accuracy in current ultrasound images, high application cost and inability to perform metastasis status diagnosis.

[0007] In a first aspect of an embodiment of the present invention, a method for detecting and segmenting a thyroid tumor and diagnosing lymph node metastasis status is provided, comprising:

[0008] Collect thyroid tumor ultrasound images, perform data cleaning and privacy processing on the ultrasound images, and select ultrasound images that meet the preset image quality requirements to establish a basic database;

[0009] Annotate the outer contour of the thyroid tumor in the ultrasound image of the basic database and the minimum circumscribed rectangle corresponding to the outer contour;

[0010] The annotated ultrasound image is used as input, the target detection is performed by building a YOLOv8 detection model, and the tumor edge is segmented by the HRNet segmentation model;

[0011] The status of cervical lymph node metastasis of thyroid tumors was determined by semiquantitative parameter measurement and Bayesian diagnostic classification model;

[0012] Based on the trained YOLOv8 detection model, HRNet segmentation model and Bayesian diagnosis classification model, the ultrasound images to be detected are detected and segmented, and the metastasis status of cervical lymph nodes is determined.

[0013] In a second aspect of an embodiment of the present invention, a thyroid tumor detection and segmentation and lymph node metastasis status diagnosis system is provided, comprising:

[0014] The data acquisition module is used to collect ultrasound images of thyroid tumors, perform data cleaning and privacy processing on the ultrasound images, and select ultrasound images whose image quality meets the preset requirements to establish a basic database;

[0015] A data annotation module, used to annotate the outer contour of the thyroid tumor in the ultrasound image in the basic database and the minimum circumscribed rectangle corresponding to the outer contour;

[0016] The model training module is used to take the annotated ultrasound image as input, perform target detection by building a YOLOv8 detection model, segment the tumor edge through the HRNet segmentation model, and determine the cervical lymph node metastasis status of thyroid tumors through semi-quantitative parameter measurement and Bayesian diagnostic classification model;

[0017] The detection and recognition module is used to detect and segment the ultrasound image to be detected based on the trained YOLOv8 detection model, HRNet segmentation model and Bayesian diagnosis and classification model, and to determine the metastasis status of the cervical lymph nodes.

[0018] In a third aspect of an embodiment of the present invention, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method described in the first aspect of the embodiment of the present invention when executing the computer program.

[0019] In a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method provided in the first aspect of the embodiment of the present invention are implemented.

[0020] In the embodiment of the present invention, based on the ultrasound image basic database, the YOLOv8 detection model, HRNet segmentation model and Bayesian diagnosis classification model are trained respectively to realize the target detection, segmentation and diagnosis of ultrasound images, which not only has high detection and segmentation accuracy, but also has low implementation cost, and can effectively save labor costs. At the same time, the coupling of the detection model, segmentation model and diagnosis model can be realized, the tumor area can be detected and segmented in real time, the semi-quantitative parameters of the tumor can be automatically measured, and the metastasis status of the cervical lymph nodes of thyroid tumors can be quickly diagnosed, which is convenient for clinical application. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 A schematic flow chart of a method for thyroid tumor detection and segmentation and lymph node metastasis status diagnosis provided by one embodiment of the present invention;

[0023] Figure 2 A schematic diagram of the structure of a thyroid tumor detection and segmentation and lymph node metastasis status diagnosis system provided by one embodiment of the present invention;

[0024] Figure 3 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] It should be understood that the term "including" and other similar expressions in the specification or claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions, such as a process, method, system, or device including a series of steps or units is not limited to the listed steps or units. In addition, "first" and "second" are used to distinguish different objects, not to describe a specific order.

[0027] See also Figure 1, a flowchart of a method for thyroid tumor detection and segmentation and lymph node metastasis status diagnosis provided by an embodiment of the present invention includes:

[0028] S101, collecting ultrasound images of thyroid tumors, performing data cleaning and privacy processing on the ultrasound images, and selecting ultrasound images whose image quality meets preset requirements to establish a basic database;

[0029] The thyroid ultrasound image is ultrasound image data recorded in the hospital case database, and the original ultrasound image needs to be cleaned and privacy processed (such as information anonymity).

[0030] The preset requirements are quality requirements that need to be met after professionals evaluate the image quality, such as average image quality, average imaging of the thyroid gland and surrounding tissues, but will not affect subsequent diagnosis.

[0031] Optionally, the ultrasound image quality is scored, and ultrasound images with image quality scores greater than a preset value are selected to construct a basic database.

[0032] For example, the scoring is based on the following criteria: 5 points for very good image quality, with the thyroid gland and surrounding tissues clearly visible; 4 points for good image quality, with relatively clear imaging of the thyroid gland and surrounding tissues; 3 points for average image quality, with average imaging of the thyroid gland and surrounding tissues, but not affecting subsequent diagnosis; 2 points for poor image quality, with blurred boundaries between the thyroid gland and surrounding tissues; 1 point for very poor image quality, with no subsequent diagnosis possible. After professionals have scored all images, they can select images with scores above 3 points.

[0033] S102, marking the outer contour of the thyroid tumor in the ultrasound image in the basic database and the minimum circumscribed rectangle corresponding to the outer contour;

[0034] The outer contour of the thyroid tumor in the ultrasound image is outlined. After all the outlines and proofreading are completed, they are exported as annotation files in JSON format. After the preliminary ROI is outlined, the circumscribed rectangular frame of the circumscribed contour is obtained. The outer contour of the ultrasound image and the corresponding circumscribed rectangular frame are used as training data and input into the detection model.

[0035] S103, taking the annotated ultrasound image as input, performing target detection by building a YOLOv8 detection model, and segmenting the tumor edge by using the HRNet segmentation model;

[0036] The YOLOv8 detection model includes an input module, a base network module, a feature fusion module, a single-stage detector and a two-stage detector;

[0037] The input module is used to perform data enhancement and preprocessing on the input ultrasound image to meet the requirements of the detection algorithm; the base network module is composed of hierarchical stacked convolutional layers, which is used to extract features from the ultrasound image and extract feature maps of each feature scale; the feature fusion module is used to perform feature fusion on multi-scale feature maps; the single-stage detector is used to realize the regression and category prediction of the multi-scale feature map detection box position; the two-stage detector is composed of ROIAlign combined with a convolutional layer, which is used to correct and filter false positives on the output results of the single-stage detector.

[0038] The image is cropped based on the outer contour of the thyroid tumor marked in the ultrasound image, and the cropped image is used as the input of the HRNet segmentation model. The feature map scale can maintain a high resolution during the entire network transmission process, which can effectively retain the characteristics of the tumor space in the ultrasound image.

[0039] S104. Determine the cervical lymph node metastasis status of thyroid tumors by semi-quantitative parameter measurement and Bayesian diagnostic classification model;

[0040] The Bayesian diagnostic classification model is a Bayesian ResNet34 network, which is used to diagnose cervical lymph node metastasis of thyroid tumors. After the ultrasonic image thyroid tumor is detected and segmented, the relevant parameters such as the long diameter, short diameter, area and texture of the tumor area are obtained through semi-quantitative parameter measurement, and a Bayesian diagnostic classification model is constructed to judge the status of the cervical lymph nodes of thyroid tumors in the ultrasonic image.

[0041] S105. Based on the trained YOLOv8 detection model, HRNet segmentation model and Bayesian diagnosis classification model, the ultrasound image to be detected is detected and segmented, and the metastasis status of the cervical lymph nodes is determined.

[0042] After training the YOLOv8 detection model, HRNet segmentation model and Bayesian diagnosis model based on the ultrasound image data in the basic database, the trained models are used to detect, segment and judge the newly acquired ultrasound images to be identified by the ultrasound instrument.

[0043] In the embodiment, by constructing a multi-level detection network and a segmentation network, the segmentation accuracy of the tumor contour can be improved, the lymph node metastasis status diagnosis network is combined with the detection network and the segmentation network, and the semi-quantitative parameters of the tumor are measured to achieve rapid diagnosis of cervical lymph node metastasis of thyroid tumors, thereby coupling the detection, segmentation and diagnosis of thyroid tumor ultrasound images, reducing implementation costs and making clinical applications more convenient.

[0044] In one embodiment, the basic database is updated in real time, and the detection accuracy of each model is evaluated; according to the model evaluation results, the models that do not meet the evaluation requirements in the YOLOv8 detection model, HRNet segmentation model and Bayesian diagnosis classification model are iteratively updated based on the basic database, thereby not only achieving.

[0045] The detection results, segmentation results and diagnosis results of each model are obtained regularly, and the model detection results, segmentation results and diagnosis results are evaluated. The evaluation is divided into qualitative evaluation and quantitative evaluation. The qualitative evaluation is evaluated according to the scores of professionals. The quantitative evaluation is specific to the task and purpose. The model output results are compared with the gold standard to obtain the scoring results. The commonly used quantitative evaluation method for the detection task is mAP (mean Average Precsion), the commonly used quantitative evaluation method for the segmentation task is the Dice coefficient, and the commonly used quantitative evaluation method for the diagnosis and classification task is AUC, specificity and sensitivity. According to the scoring results, when the score is less than a certain set value (such as three points and less than three points), the corresponding model is iteratively updated, or the mAP, AUC and Dice coefficient are less than a certain threshold. When it is iteratively updated. In the embodiment, it is only necessary to evaluate the effectiveness of each model separately and update a certain model separately, without affecting the other two, the efficiency of the algorithm iteration update can be improved, the accuracy of model detection, segmentation and diagnosis can be guaranteed, and the consumption of manpower and material resources in the model update can be reduced.

[0046] It should be understood that the serial numbers of the steps in the above embodiments do not imply a sequence of execution, and the execution sequence of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0047] Figure 2 A schematic diagram of the structure of a thyroid tumor detection and segmentation and lymph node metastasis status diagnosis system provided by an embodiment of the present invention, the system comprising:

[0048] The data acquisition module 210 is used to collect ultrasound images of thyroid tumors, perform data cleaning and privacy processing on the ultrasound images, and select ultrasound images whose image quality meets preset requirements to establish a basic database;

[0049] Wherein, the data acquisition module 210 includes:

[0050] The image selection unit is used to score the quality of ultrasound images and select ultrasound images with image quality scores greater than a preset value to construct a basic database.

[0051] The data annotation module 220 is used to annotate the outer contour of the thyroid tumor in the ultrasound image in the basic database and the minimum circumscribed rectangle corresponding to the outer contour;

[0052] The model training module 230 is used to take the annotated ultrasound image as input, perform target detection by constructing a YOLOv8 detection model, segment the tumor edge by using an HRNet segmentation model, and determine the cervical lymph node metastasis status of the thyroid tumor by using semi-quantitative parameter measurement and a Bayesian diagnostic classification model;

[0053] The YOLOv8 detection model includes an input module, a base network module, a feature fusion module, a single-stage detector and a two-stage detector;

[0054] The input module is used to perform data enhancement and preprocessing on the input ultrasound image to meet the requirements of the detection algorithm; the base network module is composed of hierarchical stacked convolutional layers, which is used to extract features from the ultrasound image and extract feature maps of each feature scale; the feature fusion module is used to perform feature fusion on multi-scale feature maps; the single-stage detector is used to realize the regression and category prediction of the multi-scale feature map detection box position; the two-stage detector is composed of ROIAlign combined with a convolutional layer, which is used to correct and filter false positives on the output results of the single-stage detector.

[0055] The detection and identification module 240 is used to detect and segment the ultrasound image to be detected based on the trained YOLOv8 detection model, HRNet segmentation model and Bayesian diagnosis classification model, and determine the metastasis status of the cervical lymph nodes.

[0056] Preferably, the detection and identification module 240 further includes:

[0057] The iterative update module is used to update the basic database in real time and evaluate the detection accuracy of each model. According to the model evaluation results, the YOLOv8 detection model, HRNet segmentation model and Bayesian diagnosis classification model that do not meet the evaluation requirements are iteratively updated based on the basic database.

[0058] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system and module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0059] Figure 3 1 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device is used for thyroid tumor detection and segmentation and cervical lymph node metastasis status diagnosis. Figure 3 As shown, the electronic device 3 of this embodiment includes: a memory 310, a processor 320 and a system bus 330, wherein the memory 310 includes an executable program 3101 stored thereon, and those skilled in the art can understand that Figure 3The electronic device structure shown in the figure does not constitute a limitation of the electronic device, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0060] Combine the following Figure 3 A detailed introduction to the various components of electronic equipment:

[0061] The memory 310 can be used to store software programs and modules, and the processor 320 executes various functional applications and data processing of the electronic device by running the software programs and modules stored in the memory 310. The memory 310 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device (such as cache data), etc. In addition, the memory 310 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0062] The memory 310 includes an executable program 3101 of the network request method, and the executable program 3101 can be divided into one or more modules / units, which are stored in the memory 310 and executed by the processor 320 to achieve detection, segmentation and diagnosis of ultrasound images, etc. The one or more modules / units can be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program 3101 in the electronic device 3. For example, the computer program 3101 can be divided into functional modules such as a data acquisition module, a data standard module, a model training module, and a detection and recognition module.

[0063] The processor 320 is the control center of the electronic device. It uses various interfaces and lines to connect various parts of the entire electronic device. By running or executing software programs and / or modules stored in the memory 310, and calling data stored in the memory 310, it performs various functions of the electronic device and processes data, thereby monitoring the overall status of the electronic device. Optionally, the processor 320 may include one or more processing units; preferably, the processor 320 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, application programs, etc., and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 320.

[0064] The system bus 330 is used to connect the various functional components inside the computer, and can transmit data information, address information, and control information. Its types can be, for example, PCI bus, ISA bus, CAN bus, etc. The instructions of the processor 320 are transmitted to the memory 310 through the bus, and the memory 310 feeds back data to the processor 320. The system bus 330 is responsible for the data and instruction exchange between the processor 320 and the memory 310. Of course, the system bus 330 can also be connected to other devices, such as network interfaces, display devices, etc.

[0065] In the embodiment of the present invention, the executable program executed by the processor 320 included in the electronic device includes:

[0066] Collect thyroid tumor ultrasound images, perform data cleaning and privacy processing on the ultrasound images, and select ultrasound images that meet the preset image quality requirements to establish a basic database;

[0067] Annotate the outer contour of the thyroid tumor in the ultrasound image of the basic database and the minimum circumscribed rectangle corresponding to the outer contour;

[0068] The annotated ultrasound image is used as input, and the target detection is performed by building a YOLOv8 detection model, and the tumor edge is segmented by the HRNet segmentation model;

[0069] The status of cervical lymph node metastasis of thyroid tumors was determined by semiquantitative parameter measurement and Bayesian diagnostic classification model;

[0070] Based on the trained YOLOv8 detection model, HRNet segmentation model and Bayesian diagnosis classification model, the ultrasound images to be detected are detected and segmented, and the metastasis status of cervical lymph nodes is determined.

[0071] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0072] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0073] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting and segmenting thyroid tumors and diagnosing lymph node metastasis status, characterized in that: include: Collect thyroid tumor ultrasound images, perform data cleaning and privacy processing on the ultrasound images, and select ultrasound images that meet the preset image quality requirements to establish a basic database; Annotate the outer contour of the thyroid tumor in the ultrasound image of the basic database and the minimum circumscribed rectangle corresponding to the outer contour; The annotated ultrasound image is used as input, and the target detection is performed by building a YOLOv8 detection model, and the tumor edge is segmented by the HRNet segmentation model; The status of cervical lymph node metastasis of thyroid tumors was determined by semiquantitative parameter measurement and Bayesian diagnostic classification model; Based on the trained YOLOv8 detection model, HRNet segmentation model and Bayesian diagnosis classification model, the ultrasound images to be detected are detected and segmented, and the metastasis status of cervical lymph nodes is determined.

2. The method according to claim 1, characterized in that The step of selecting ultrasound images whose image quality meets preset requirements to establish a basic database comprises: The ultrasound image quality is scored, and ultrasound images with image quality scores greater than a preset value are selected to construct a basic database.

3. The method according to claim 1, characterized in that The YOLOv8 detection model includes an input module, a base network module, a feature fusion module, a single-stage detector and a two-stage detector; The input module is used to perform data enhancement and preprocessing on the input ultrasound image to meet the requirements of the detection algorithm; the base network module is composed of stacked convolutional layers, which is used to extract features from the ultrasound image and extract feature maps of each feature scale; The feature fusion module is used to perform feature fusion on the multi-scale feature map; The single-stage detector is used to achieve the regression of the position of the multi-scale feature map detection box and the category prediction; the two-stage detector is composed of ROIAlign combined with a convolutional layer, which is used to correct the output results of the single-stage detector and filter false positives.

4. The method according to claim 1, characterized in that: The detecting and segmenting of the ultrasonic image to be detected based on the trained YOLOv8 detection model, HRNet segmentation model and Bayesian diagnosis classification model, and judging the metastasis status of the cervical lymph nodes also includes: Updating the basic database in real time and evaluating the detection accuracy of each model; According to the model evaluation results, the YOLOv8 detection model, HRNet segmentation model, and Bayesian diagnosis classification model that do not meet the evaluation requirements are iteratively updated based on the basic database.

5. A thyroid tumor detection and segmentation and lymph node metastasis status diagnosis system, characterized in that: include: The data acquisition module is used to collect ultrasound images of thyroid tumors, perform data cleaning and privacy processing on the ultrasound images, and select ultrasound images whose image quality meets the preset requirements to establish a basic database; A data annotation module, used to annotate the outer contour of the thyroid tumor in the ultrasound image in the basic database and the minimum circumscribed rectangle corresponding to the outer contour; The model training module is used to take the annotated ultrasound image as input, perform target detection by building a YOLOv8 detection model, segment the tumor edge through the HRNet segmentation model, and determine the cervical lymph node metastasis status of thyroid tumors through semi-quantitative parameter measurement and Bayesian diagnostic classification model; The detection and recognition module is used to detect and segment the ultrasound image to be detected based on the trained YOLOv8 detection model, HRNet segmentation model and Bayesian diagnosis and classification model, and to determine the metastasis status of the cervical lymph nodes.

6. The system according to claim 5, characterized in that The data acquisition module comprises: The image selection unit is used to score the quality of ultrasound images and select ultrasound images with image quality scores greater than a preset value to construct a basic database.

7. The system according to claim 5, characterized in that The YOLOv8 detection model includes an input module, a base network module, a feature fusion module, a single-stage detector and a two-stage detector; The input module is used to perform data enhancement and preprocessing on the input ultrasound image to meet the requirements of the detection algorithm; the base network module is composed of stacked convolutional layers, which is used to extract features from the ultrasound image and extract feature maps of each feature scale; The feature fusion module is used to perform feature fusion on the multi-scale feature map; The single-stage detector is used to achieve the regression of the position of the multi-scale feature map detection box and the category prediction; the two-stage detector is composed of ROIAlign combined with a convolutional layer, which is used to correct the output results of the single-stage detector and filter false positives.

8. The system according to claim 5, characterized in that The detection and identification module also includes: The iterative update module is used to update the basic database in real time and evaluate the detection accuracy of each model. According to the model evaluation results, the YOLOv8 detection model, HRNet segmentation model and Bayesian diagnosis classification model that do not meet the evaluation requirements are iteratively updated based on the basic database.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of a method for detecting and segmenting a thyroid tumor and diagnosing lymph node metastasis status as described in any one of claims 1 to 4 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed, the steps of a method for detecting and segmenting a thyroid tumor and diagnosing lymph node metastasis status as described in any one of claims 1 to 4 are implemented.

Citation Information

Cited By

  • Thyroid tumor B ultrasonic diagnosis method and system based on multi-model integration

    CN120612532A