Thyroid nodule benign and malignant judgment equipment and use method thereof

Abstract

The invention provides thyroid nodule benign and malignant judgment equipment and a use method thereof. The equipment comprises an image analysis and segmentation module, an image feature extraction module, a model calling module and a benign and malignant judgment result output module. According to the method, potential features of the nodule region and the perijunction region are fully mined, rich and representative nodule features and perijunction features are obtained, the actual situation of the thyroid nodule can be reflected more comprehensively, and the obtained benign probability and malignant probability of the thyroid nodule are more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of medical devices, and more particularly, to a device for judging the benignity and malignancy of thyroid nodules and a method for using the same. Background Art

[0002] In the field of clinical diagnosis of thyroid nodules, accurately judging the benignity and malignancy of nodules has always been a crucial topic. The traditional method for judging the benignity and malignancy of thyroid nodules mainly relies on the Thyroid Imaging Reporting and Data System (TI-RADS). This method mainly evaluates based on ultrasonic semantic features such as the composition, echo, shape, margin, and echogenic foci of the nodules. However, this evaluation method is greatly affected by the subjective judgment of doctors, resulting in significant differences in the accuracy and consistency of the diagnostic results.

[0003] With the development of medical technology, artificial intelligence (AI) and radiomics technology have gradually been applied to the diagnosis of thyroid nodules. Radiomics can extract a large number of features that are imperceptible to the naked eye from traditional images, providing a more objective quantitative basis for clinical decision-making. However, previous studies have had many limitations. On the one hand, the sample sizes of some studies are small, making it difficult to construct a model with wide applicability and stability. On the other hand, in the study of thyroid nodule radiomics, most studies only focus on the internal region of the nodules, ignoring the important information that the perinodular region may provide.

[0004] In fact, the tumor microenvironment plays a key role in the occurrence, development, and treatment response of tumors. As an important part of the tumor microenvironment, the perinodular region of the nodules may contain information valuable for judging the benignity and malignancy of thyroid nodules. Currently, in the research of other tumors, such as breast cancer and lung cancer, it has been confirmed that the perinodular radiomics features have diagnostic and predictive capabilities in disease assessment. However, in the research of thyroid nodules, there are few relevant reports based on perinodular images. Therefore, developing a device for judging the benignity and malignancy of thyroid nodules that can effectively utilize the perinodular image information of nodules and combine clinical data and internal image features of nodules has important clinical significance and application value, and is expected to fill the gaps in the existing technology and improve the accuracy and reliability of judging the benignity and malignancy of thyroid nodules. Summary of the Invention

[0005] To solve the technical problems existing in the above background art, the present invention provides a device for judging the benignity and malignancy of thyroid nodules, a method for using the same, an electronic device, a computer storage medium, and a computer program product.

[0006] The present invention provides a device for judging the benignity and malignancy of thyroid nodules, which includes an image analysis and segmentation module, an image feature extraction module, a model calling module, and a benignity and malignancy judgment result output module; The image analysis and segmentation module is used to identify the nodule region and the perinodular region from the received two-dimensional gray-scale ultrasound image of the target patient, and segment the nodule region image and the perinodular region image; The image feature extraction module is used to perform convolutional feature extraction on the nodule region image and the perinodular region image, and obtain nodule features and perinodular features respectively; The model calling module is used to call a benign and malignant composite judgment model to perform in-depth processing on the nodule features and the perinodular features, and obtain the benign probability and / or malignant probability of the nodule of the target patient; wherein, the benign and malignant composite judgment model is constructed based on a deep learning algorithm; The benign and malignant judgment result output module is used to output the benign probability and / or the malignant probability.

[0007] Optionally, the image analysis and segmentation module identifying the nodule region and the perinodular region from the two-dimensional gray-scale ultrasound image includes: Using ITK-SNAP software to identify the nodule region from the two-dimensional gray-scale ultrasound image, and taking the identified nodule region as a reference, expanding outwards by several unit lengths respectively to obtain the expanded perinodular region.

[0008] Optionally, the specified number of times of expanding outwards is determined by the following method, including: Extracting the structural features of the nodule region, and using a classification model to perform preliminary analysis on the structural features to obtain the preliminary malignant probability of the nodule region; wherein, the structural features include the size information and shape information of the nodule region; Based on the preliminary malignant probability, matching to obtain the specified number of times of the unit length expanded outwards.

[0009] Optionally, the benign and malignant composite judgment model performing in-depth processing on the nodule features and the perinodular features to obtain the benign probability and / or malignant probability of the nodule of the target patient includes: The benign and malignant composite judgment model includes a first judgment model, a second judgment model, a third judgment model and an integration model; The first judgment model performs in-depth processing on the nodule features to obtain a first probability, the second judgment model performs in-depth processing on the perinodular features to obtain a second probability, and the third judgment model performs in-depth processing on the nodule features and the perinodular features to obtain a third probability; The integration model performs fusion processing on the first probability, the second probability, and the third probability to obtain the benign probability and / or malignant probability of the nodule of the target patient.

[0010] Optionally, the first judgment model, the second judgment model, and the third judgment model in the benign and malignant composite judgment model are pre-trained; Among them, the training data used for pre-training the first judgment model includes nodule feature data and first marking data; the training data used for pre-training the second judgment model includes perinodular feature data and second marking data; the training data used for pre-training the third judgment model includes nodule feature data, perinodular feature data, and third marking data.

[0011] Optionally, the first judgment model, the second judgment model, and the third judgment model in the benign and malignant composite judgment model are pre-trained; Among them, the training data used for pre-training the first judgment model includes nodule feature data and first three-dimensional nodule data; the training data used for pre-training the second judgment model includes perinodular feature data and second three-dimensional nodule data; the training data used for pre-training the third judgment model includes nodule feature data, perinodular feature data, and third three-dimensional nodule data.

[0012] The present invention also provides a usage method of a device for judging the benignity and malignancy of a thyroid nodule, and the method includes: Input the two-dimensional gray-scale ultrasound image of the thyroid of the target patient into the device for judging the benignity and malignancy of a thyroid nodule described in any one of the foregoing; The device for judging the benignity and malignancy of a thyroid nodule outputs the benign probability and / or the malignant probability.

[0013] The present invention also provides an electronic device, which is applied to a device for judging the benignity and malignancy of a thyroid nodule described in any one of the foregoing, and includes a memory storing executable program codes; a processor coupled to the memory.

[0014] The present invention also provides a computer storage medium, which is applied to a device for judging the benignity and malignancy of a thyroid nodule described in any one of the foregoing, and a computer program is stored on the storage medium.

[0015] The present invention also provides a computer program product, which is applied to a device for judging the benignity and malignancy of a thyroid nodule described in any one of the foregoing, and the computer program product includes a computer program stored in a computer storage medium.

[0016] The beneficial effects of the present invention are at least as follows: The present invention fully excavates the potential features of the nodule area and the perinodular area, obtains rich and representative nodule features and perinodular features, can more comprehensively reflect the actual situation of the thyroid nodule, and makes the obtained benign probability and malignant probability of the thyroid nodule more accurate. Description of the Drawings

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0018] Figure 1 It is a schematic structural diagram of a device for judging the benignity and malignancy of thyroid nodules disclosed in an embodiment of the present invention.

[0019] Figure 2 It is a schematic diagram of identifying the nodule region and the perinodular region in a two-dimensional gray-scale ultrasound image disclosed in an embodiment of the present invention.

[0020] Figure 3 It is a schematic structural diagram of a benign and malignant composite judgment model disclosed in an embodiment of the present invention. Detailed implementation manners

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0022] For the above technical problems, as Figure 1 shown, an embodiment of the present invention discloses a device for judging the benignity and malignancy of thyroid nodules. The device includes an image analysis and segmentation module, an image feature extraction module, a model calling module, and a benign and malignant judgment result output module; The image analysis and segmentation module is used to identify the nodule region and the perinodular region from the received two-dimensional gray-scale ultrasound image of the target patient's thyroid gland, and segment the nodule region image and the perinodular region image; The image feature extraction module is used to perform convolutional feature extraction on the nodule region image and the perinodular region image to obtain nodule features and perinodular features respectively; The model calling module is used to call a benign and malignant composite judgment model to perform in-depth processing on the nodule features and the perinodular features to obtain the benign probability and / or malignant probability of the nodules of the target patient; wherein, the benign and malignant composite judgment model is constructed based on a deep learning algorithm; The benign and malignant judgment result output module is used to output the benign probability and / or the malignant probability.

[0023] The above-mentioned device of the present invention first accurately identifies the nodule region and the perinodular region from the two-dimensional gray-scale ultrasound image of the thyroid gland, and then adopts the convolutional feature extraction technology to fully explore the potential features of the nodule region and the perinodular region, obtaining rich and representative nodule features and perinodular features, which can more comprehensively reflect the actual situation of the thyroid nodule. Then, the benign and malignant composite judgment model constructed based on the deep learning algorithm has powerful learning and analysis capabilities, and can deeply process the extracted nodule features and perinodular features, that is, comprehensively consider the information inside and around the nodule. Compared with the model that only relies on the internal features of the nodule, it greatly improves the accuracy and reliability of the benign and malignant judgment of the thyroid nodule. Finally, the benign probability and / or the malignant probability are output, providing clear and definite diagnostic references for doctors, helping doctors formulate treatment plans more efficiently and accurately, reducing problems such as over-treatment or under-treatment caused by inaccurate diagnosis, and thus improving the clinical diagnosis and treatment level of thyroid nodules.

[0024] Optionally, the image analysis and segmentation module identifies the nodule region and the perinodular region from the two-dimensional gray-scale ultrasound image, including: Using the ITK-SNAP software to identify the nodule region from the two-dimensional gray-scale ultrasound image, and taking the identified nodule region as a reference, expanding outward by several unit lengths respectively to obtain the expanded perinodular region.

[0025] In this embodiment, ITK-SNAP is a software tool for medical image segmentation and analysis, which is widely used in medical research and clinical practice. It is mainly used to process medical images such as MRI (magnetic resonance imaging), CT (computed tomography), etc., and is particularly suitable for the automatic and semi-automatic segmentation of tissues and organs. The software provides a model-based automatic segmentation function, which uses the existing labeled data for training and can automatically identify and segment new images. Refer to Figure 2 as shown Figure 2 The left image in is the two-dimensional gray-scale ultrasound image. Using the ITK-SNAP software, the nodule region can be identified from the left image, that is, the red region in the middle image.

[0026] After the nodule region is identified, it is expanded outward based on the nodule region. The unit distance of each expansion is fixed, and this expansion is based on the actual boundary contour of the nodule region. After expanding to the specified number of times, the appropriate perinodular region is obtained, that is Figure 2 the colored region outside the red region in the right region, and each color represents one expansion.

[0027] Optionally, the specified number of times of the unit length of the outward expansion is determined by the following method, including: Extract the structural features of the nodule region, and use a classification model to preliminarily analyze the structural features to obtain the preliminary malignancy probability of the nodule region; wherein, the structural features include the size information and shape information of the nodule region. Match and obtain the specified number of times of the unit length expanded outward based on the preliminary malignancy probability.

[0028] In this embodiment, the present invention uses a simpler classification model to preliminarily analyze the structural features of the nodule region to obtain the preliminary malignancy probability, and then adjusts the above-mentioned specified number of times according to the size of the preliminary malignancy probability, so as to dynamically change the size of the perinodular region. Among them, when the preliminary malignancy probability that the nodule belongs to malignancy is lower, the perinodular region is set smaller, that is, there is no need to refer to too many features in the perinodular region at this time; while when the preliminary malignancy probability is higher, the perinodular region is set larger at this time, so that more auxiliary information related to the benign and malignant degree of the nodule is included, which helps to more deeply and accurately analyze the benign and malignant nature of the nodule region with a high malignancy probability.

[0029] The structural features mainly include the size information and shape information of the nodule region. Among them, the size information includes specific measurement indexes such as the length, width, height, diameter, perimeter, area of the nodule, etc., and these indexes can reflect the size of the nodule. For example, through the ultrasonic image, it can be measured how many millimeters the length of the nodule is and how many millimeters the width is, etc. The shape information involves the shape of the nodule, such as being circular, oval, irregular shape, and whether the edge of the nodule is smooth and whether there is a lobulated shape, etc. These structural features are the basic descriptions of the nodule and provide basic data for subsequent analysis.

[0030] The classification model can be constructed based on machine learning algorithms (such as decision trees, support vector machines, neural networks, etc.), and it outputs a preliminary malignancy probability. This probability is an estimated value, which represents the likelihood that the nodule is malignant according to the structural features such as the size and shape of the nodule. For example, through the analysis of the classification model, it may be concluded that the probability that the nodule is malignant is 30%, which means that according to the size and shape of the nodule, there is a 30% possibility that the nodule is malignant.

[0031] The mapping relationship between the specified number of times and the preliminary malignancy probability can be preset. For example, when the preliminary malignancy probability is between 0 - 20%, the specified number of times is 1 time; when the preliminary malignancy probability is between 20% - 50%, the specified number of times is 3 times; when the preliminary malignancy probability exceeds 50%, the specified number of times is 5 times, etc.

[0032] Optionally, the benign and malignant composite judgment model deeply processes the nodule features and the perinodular features to obtain the benign probability and / or malignant probability of the nodule of the target patient, including: The benign and malignant composite judgment model includes a first judgment model, a second judgment model, a third judgment model, and an integration model; The first judgment model deeply processes the nodule features to obtain a first probability, the second judgment model deeply processes the perinodular features to obtain a second probability, and the third judgment model deeply processes the nodule features and the perinodular features to obtain a third probability; The integration model performs a fusion process on the first probability, the second probability, and the third probability to obtain the benign probability and / or malignant probability of the nodule of the target patient.

[0033] In this embodiment, as Figure 3 shown, the benign and malignant composite judgment model of the present invention consists of four parts, namely a first judgment model, a second judgment model, a third judgment model, and an integration model. This structural design aims to process nodule features from different perspectives through multiple sub-models and finally fuse the results of these sub-models through the integration model to obtain a more accurate judgment result of the benign and malignant of thyroid nodules. Specifically: The first judgment model is a model specifically for deeply processing nodule features. It can adopt deep learning algorithms such as deep neural network (DNN), convolutional neural network (CNN), or recurrent neural network (RNN), etc. This model receives the nodule features extracted from the image feature extraction module, and these features may include geometric features of the nodule (such as size, shape, perimeter, area, etc.), intensity features (such as the distribution of gray values, etc.), and texture features (such as features described by gray-level co-occurrence matrix, gray-level run length matrix, etc.). After deep processing, the first judgment model outputs a probability value, that is, the first probability, which represents the possibility that the nodule is malignant judged only based on the nodule features.

[0034] The input of the second judgment model is the perinodular features. The perinodular features are the features of the area around the thyroid nodule, including information such as the texture of the perinodular tissue, blood flow distribution, density, and the boundary condition with the surrounding tissue. These features can reflect the interaction relationship between the nodule and the surrounding tissue. Similar to the first judgment model, the second judgment model uses deep learning algorithms for deep processing. It can learn the hidden information in the perinodular features that may be related to the benign and malignant of the thyroid nodule and output the "second probability", which represents the possibility that the thyroid nodule is malignant judged only based on the perinodular features.

[0035] The third judgment model processes both the nodule features and the perinodular features simultaneously, that is, by comprehensively considering the information of the nodule itself and around the nodule, to evaluate the situation of the thyroid nodule from a more comprehensive perspective. This combination of inputs can capture more factors that may affect the benign and malignant of the nodule, including both the attributes of the nodule itself and the influencing factors of the interaction between the nodule and the surrounding tissue.

[0036] The integrated model can use the weighted average method or the Bayesian fusion method to fuse the above three probabilities into a benign probability and / or a malignant probability. Among them, when using the weighted average method, the weighted value of the third probability can be set to be the highest, the first probability is the second, and the weighted value of the second probability is the smallest. When using the Bayesian fusion method, the three probabilities are regarded as prior or posterior probabilities, calculated according to the Bayesian formula, and combined with the uncertainty and correlation of different probabilities to comprehensively obtain a more accurate probability estimate.

[0037] Optionally, the first judgment model, the second judgment model, and the third judgment model in the benign and malignant composite judgment model are pre-trained; Among them, the training data used for pre-training the first judgment model includes nodule feature data and first labeled data; the training data used for pre-training the second judgment model includes perinodular feature data and second labeled data; the training data used for pre-training the third judgment model includes nodule feature data, perinodular feature data, and third labeled data.

[0038] In this embodiment, in view of the different analysis methods of the above three judgment models, their corresponding training data are also different. Among them, the above labeled data refers to the benign probability or malignant probability corresponding to each feature data set manually or automatically by a machine.

[0039] Optionally, the first judgment model, the second judgment model, and the third judgment model in the benign and malignant composite judgment model are pre-trained; Among them, the training data used for pre-training the first judgment model includes nodule feature data and first nodule three-dimensional data; the training data used for pre-training the second judgment model includes perinodular feature data and second nodule three-dimensional data; the training data used for pre-training the third judgment model includes nodule feature data, perinodular feature data, and third nodule three-dimensional data.

[0040] In this embodiment, different from the previous embodiment, the labeled data in this embodiment is replaced by nodule three-dimensional data. The nodule three-dimensional data refers to the features extracted from the three-dimensional image features of the corresponding thyroid nodule. The three-dimensional image refers to, for example, an MRI image, or a series of two-dimensional images with spatial positioning characteristics are collected by scanning the thyroid from multiple angles with an ultrasound probe, and then these two-dimensional images are combined and reconstructed into a three-dimensional image by using computer software through feature positioning feature recognition; or, the human body is scanned by X-ray to obtain multiple two-dimensional tomographic images, and then reconstructed into a three-dimensional image by a computer. The present invention does not make specific limitations on this.

[0041] When each model in the benign and malignant composite judgment model is being trained, the three-dimensional nodule data can be analyzed first. For example, using the classification model described above, the probabilities that the corresponding nodules are benign or malignant can be obtained, and thus the labeled data can be indirectly obtained. Then, based on the feature data and the labeled data, new training data is constructed and used to train the corresponding model.

[0042] An embodiment of the present invention also provides a usage method of a device for judging the benignity and malignancy of thyroid nodules. The method includes: Input the two-dimensional gray-scale ultrasound image of the thyroid of the target patient into the device for judging the benignity and malignancy of thyroid nodules described in any one of the foregoing. The device for judging the benignity and malignancy of thyroid nodules outputs the benign probability and / or the malignant probability.

[0043] An embodiment of the present invention also provides an electronic device, which is applied to a device for judging the benignity and malignancy of thyroid nodules as described in any one of the foregoing, including a memory storing executable program codes; and a processor coupled to the memory.

[0044] An embodiment of the present invention also provides a computer storage medium, which is applied to a device for judging the benignity and malignancy of thyroid nodules as described in any one of the foregoing, and a computer program is stored on the storage medium.

[0045] An embodiment of the present invention also provides a computer program product, which is applied to a device for judging the benignity and malignancy of thyroid nodules as described in any one of the foregoing, and the computer program product includes a computer program stored in a computer storage medium.

[0046] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (apparatus), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, and the combination of processes and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate for implementing in the process Figure 1 one process or multiple processes and / or blocks Figure 1 a device for the function specified in one block or multiple blocks.

[0047] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device implements in the process Figure 1 one process or multiple processes and / or blocksFigure 1 The functions specified in one or more boxes.

[0048] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide for implementing the steps of the functions specified in one or more processes and / or boxes Figure 1 One process or more processes and / or boxes Figure 1 The steps of the functions specified in one or more boxes.

[0049] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claimed rights.

Claims

1. A device for judging whether a thyroid nodule is benign or malignant, characterized in that: The device includes an image analysis and segmentation module, an image feature extraction module, a model calling module, and a benign and malignant judgment result output module; The image analysis and segmentation module is used to identify the nodule area and the peri-nodular area from the received two-dimensional grayscale ultrasound image of the thyroid gland of the target patient, and segment the image of the nodule area and the peri-nodular area to obtain the image; The image feature extraction module is used to perform convolution feature extraction on the nodule region image and the peri-nodule region image to obtain nodule features and peri-nodule features respectively; The model calling module is used to call the benign and malignant composite judgment model to perform deep processing on the nodule characteristics and the peri-nodular characteristics to obtain the benign probability and / or malignant probability of the nodule of the target patient; wherein the benign and malignant composite judgment model is constructed based on a deep learning algorithm; The benign and malignant judgment result output module is used to output the benign probability and / or the malignant probability.

2. The device for judging whether a thyroid nodule is benign or malignant according to claim 1, characterized in that: The image analysis and segmentation module identifies the nodule area and the peri-nodular area from the two-dimensional grayscale ultrasound image, including: The nodule region is identified from the two-dimensional grayscale ultrasound image using ITK-SNAP software, and the identified nodule region is used as a reference to expand outward by a number of unit lengths to obtain the expanded perinodular region.

3. The device for judging whether a thyroid nodule is benign or malignant according to claim 2, characterized in that: The specified number of times the unit length is expanded outward is determined by the following methods, including: Extracting structural features of the nodule region, and using a classification model to preliminarily analyze the structural features to obtain a preliminary malignancy probability of the nodule region; wherein the structural features include size information and shape information of the nodule region; The specified number of outward expansion unit lengths is obtained based on the preliminary malignancy probability matching.

4. The device for judging whether a thyroid nodule is benign or malignant according to claim 1, characterized in that: The benign and malignant composite judgment model performs in-depth processing on the nodule features and the peri-nodular features to obtain the benign probability and / or malignant probability of the nodule of the target patient, including: The benign and malignant composite judgment model includes a first judgment model, a second judgment model, a third judgment model and an integrated model; The first judgment model performs deep processing on the nodule feature to obtain a first probability, the second judgment model performs deep processing on the peri-nodule feature to obtain a second probability, and the third judgment model performs deep processing on the nodule feature and the peri-nodule feature to obtain a third probability; The integrated model fuses the first probability, the second probability, and the third probability to obtain the benign probability and / or malignant probability of the nodule of the target patient.

5. The device for judging whether a thyroid nodule is benign or malignant according to claim 4, characterized in that: The first judgment model, the second judgment model, and the third judgment model in the benign and malignant compound judgment model are pre-trained; Among them, the training data used for pre-training of the first judgment model includes nodule feature data and first label data; the training data used for pre-training of the second judgment model includes nodule feature data and second label data; the training data used for pre-training of the third judgment model includes nodule feature data, nodule feature data and third label data.

6. The device for judging whether a thyroid nodule is benign or malignant according to claim 4, characterized in that: The first judgment model, the second judgment model, and the third judgment model in the benign and malignant compound judgment model are pre-trained; Among them, the training data used for pre-training of the first judgment model includes nodule feature data and first nodule three-dimensional data; the training data used for pre-training of the second judgment model includes circumnodule feature data and second nodule three-dimensional data; the training data used for pre-training of the third judgment model includes nodule feature data, circumnodule feature data and third nodule three-dimensional data.

7. A method for using a device for judging whether a thyroid nodule is benign or malignant, characterized in that: The method comprises: Inputting a two-dimensional grayscale ultrasound image of the thyroid gland of a target patient into the device for judging benign or malignant thyroid nodules according to any one of claims 1 to 6; The device for judging benign or malignant thyroid nodules outputs the benign probability and / or the malignant probability.

8. An electronic device, applied to the device for judging whether a thyroid nodule is benign or malignant according to any one of claims 1 to 6, characterized in that: The invention comprises a memory storing executable program code; and a processor coupled to the memory.

9. A computer storage medium, applied to a device for judging benign or malignant thyroid nodules according to any one of claims 1 to 6, characterized in that: The storage medium stores a computer program.

10. A computer program product, applied to the device for judging benign or malignant thyroid nodules according to any one of claims 1 to 6, characterized in that: The computer program product includes a computer program stored in a computer storage medium.

Images