Cervical vertebra MRI image degenerative region positioning detection method based on deep learning

Abstract

The invention relates to a cervical vertebra MRI (Magnetic Resonance Imaging) image degenerative region positioning detection method based on deep learning. The method comprises the steps that firstly, an original cervical vertebra MRI image is acquired and preprocessed; then, a degenerative region positioning model is constructed based on a TOOD network, and the degenerative region positioning model comprises a feature extraction network, a multi-scale self-attention mechanism feature fusion network and a detection head network; then, training the degenerative region positioning model based on the preprocessed original cervical vertebra MRI image; and finally, inputting a cervical vertebra MRI image to be predicted into the trained degenerative region positioning model, and outputting a degenerative region positioning detection result. The deep learning model is used for detecting the cervical intervertebral disc, degeneration grading detection can be directly carried out on the intervertebral disc according to the T2 weighted image of the MRI, and the positioning detection efficiency is effectively improved.

Description

Technical Field

[0001] The present application relates to the technical field of medical image analysis, and in particular to a method for localizing and detecting degenerative areas in cervical spine MRI images based on deep learning. Background Art

[0002] Neck pain, as a highly prevalent musculoskeletal disease and the fourth leading cause of disability, has become a serious public health problem worldwide, placing a huge burden on patients, healthcare systems, and the economic structure of countries. Despite the high prevalence and huge burden on society, neck pain has received relatively less research attention than low back pain. Similar to low back pain, a widely recognized cause of neck pain is the degeneration of the cervical intervertebral disc. The intervertebral disc is composed of the outer annulus fibrosus (AF), the inner gelatinous nucleus pulposus (NP), and the cartilaginous endplates. It is located between two adjacent vertebrae and acts as a shock absorber for the spine. The most important physiological changes of disc degeneration begin in the nucleus pulposus, usually manifested as a loss of disc height with reduced water content, accompanied by a decrease in the yield strength of the annulus fibrosus. These degenerative changes may alter biomechanical transmission and sensitize pain-sensing nerve fibers in the annulus fibrosus and nucleus pulposus, leading to disc herniation, nerve compression, and discogenic pain.

[0003] T2-weighted magnetic resonance imaging (MRI) is the most commonly used imaging modality for diagnosing cervical degenerative disc disease (CDDD) because of its advantages in detecting disc morphology and nucleus pulposus water content. Currently, the most widely used MRI classification system for disc degeneration is based on disc structure and signal intensity and the distinction between the nucleus pulposus and annulus fibrosus. Although this system provides a clear definition for each grade of disc degeneration, it is still a time-consuming and laborious task in clinical practice and is highly dependent on the expertise of radiologists and surgeons. Accurate and rapid automatic classification of cervical disc degeneration on MRI remains a challenge.

[0004] At present, the clinical judgment of cervical disc degeneration mainly relies on the manual reading of films by clinicians and radiologists, which is time-consuming and susceptible to problems with the professional level of doctors and subjective images. In addition, the existing disc detection system mainly focuses on segmenting and locating the cervical spine, which is the basic step of the disc intelligent auxiliary diagnosis system, and the steps are complex and redundant.

[0005] Therefore, in the face of rising incidence rates of cervical spondylosis, severe complications and a shortage of medical resources, there is an urgent need for a method in the relevant technologies that can accurately locate and detect areas of cervical disc degenerative lesions. Summary of the invention

[0006] Based on this, it is necessary to provide a deep learning-based method for localizing and detecting degenerative areas in cervical MRI images to address the above technical problems.

[0007] In a first aspect, the present application provides a method for localizing and detecting degenerative regions in cervical MRI images based on deep learning. The method comprises:

[0008] Obtain original cervical spine MRI images and perform preprocessing;

[0009] A degenerative region localization model is constructed based on the TOOD network, wherein the degenerative region localization model includes a feature extraction network, a multi-scale self-attention mechanism feature fusion network, and a detection head network;

[0010] Training the degenerative area localization model based on the original cervical spine MRI image after preprocessing;

[0011] The cervical MRI image to be predicted is input into the trained degenerative area positioning model, and the degenerative area positioning detection result is output.

[0012] Optionally, in one embodiment of the present application, the preprocessing includes:

[0013] The original cervical MRI image is labeled with cervical disc degeneration classification according to the Pfirrmann classification method.

[0014] Optionally, in one embodiment of the present application, constructing a degenerative region positioning model based on a TOOD network includes:

[0015] ResNet50 is used as the backbone network for extracting image features, feature pyramid network is used as the multi-scale self-attention mechanism feature fusion network, and classification head network and regression head network are used as detection head network.

[0016] Optionally, in one embodiment of the present application, the training of the degenerative area localization model based on the original cervical vertebra MRI image after preprocessing includes:

[0017] The error is back-propagated based on the network loss function, and the model parameters are updated by iterative operation. The network loss function includes a target positioning loss function and a target classification loss function.

[0018] Optionally, in one embodiment of the present application, the formula of the target positioning loss function is:

[0019]

[0020] Among them, b i represents the positioning box predicted by the model, Represents the real box of the target, and uses the GIoU loss function to calculate the positioning loss of a single predicted target box. Used for joint optimization of the participating localization loss, Npos Indicates the number of anchors of positive samples.

[0021] Optionally, in one embodiment of the present application, the formula of the target classification loss function is:

[0022]

[0023] Among them, s i represents the classification score of the i-th positive sample, s j represents the classification score of the jth negative sample, γ is the focusing parameter, BCE() is the binary cross entropy function, Used for joint optimization of the participating localization loss, N pos Indicates the number of anchors of positive samples, N neg Indicates the number of anchors for negative samples.

[0024] In a second aspect, the present application also provides a device for localizing and detecting degenerative areas in cervical MRI images based on deep learning. The device comprises:

[0025] A data acquisition module is used to acquire original cervical spine MRI images and perform preprocessing;

[0026] A degenerative region localization model construction module, used to construct a degenerative region localization model based on the TOOD network, wherein the degenerative region localization model includes a feature extraction network, a multi-scale self-attention mechanism feature fusion network, and a detection head network;

[0027] A degenerative region positioning model training module, used for training the degenerative region positioning model based on the original cervical spine MRI image after preprocessing;

[0028] The degenerative region positioning detection module is used to input the cervical MRI image to be predicted into the trained degenerative region positioning model and output the degenerative region positioning detection result.

[0029] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the steps of the methods described in the above embodiments.

[0030] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method described in each of the above embodiments are implemented.

[0031] The above-mentioned method for localizing and detecting degenerative regions of cervical MRI images based on deep learning, first, obtains the original cervical MRI image and performs preprocessing; then, constructs a degenerative region localization model based on the TOOD network, and the degenerative region localization model includes a feature extraction network, a multi-scale self-attention mechanism feature fusion network and a detection head network; then, the degenerative region localization model is trained based on the original cervical MRI image after preprocessing; finally, the cervical MRI image to be predicted is input into the trained degenerative region localization model, and the degenerative region localization detection result is output. That is to say, based on the improvement of the TOOD network, a multi-scale self-attention mechanism module is added to the framework, which enhances the global feature map modeling, thereby improving the target detection accuracy of the network. In addition, the introduction of neural memory ordinary differential equations into the head network enhances the nonlinear representation ability, guides the training of the network, and enhances its robustness. Using the deep learning model to detect the cervical intervertebral disc, the degeneration grade of the intervertebral disc can be directly detected according to the T2-weighted image of MRI, which effectively improves the efficiency of localization detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a diagram of the application environment of a method for localizing and detecting degenerative areas in cervical MRI images based on deep learning in one embodiment;

[0033] Figure 2 It is a flowchart of a method for localizing and detecting degenerative areas in cervical MRI images based on deep learning in one embodiment;

[0034] Figure 3 is a schematic diagram of a sagittal slice of a vertebral body in one embodiment;

[0035] Figure 4 A schematic diagram of annotating sagittal data of MRI cervical spine in one embodiment;

[0036] Figure 5 A schematic diagram of the structure of a degenerative region positioning model in one embodiment;

[0037] Figure 6 A schematic diagram of a degenerative region positioning model training process in one embodiment;

[0038] Figure 7 is a schematic diagram of a degenerative region positioning model detection process in one embodiment;

[0039] Figure 8 is a schematic diagram of a degenerative region positioning detection result in one embodiment;

[0040] Fig. 9 A schematic diagram of a degenerative region positioning detection report in one embodiment;

[0041] Fig.10 It is a structural block diagram of a degenerative area positioning detection device for cervical MRI images based on deep learning in one embodiment;

[0042] Fig.11 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0044] The method for localizing and detecting the degenerative area of ​​cervical MRI images based on deep learning provided in the embodiments of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the terminal 102 can be but is not limited to various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.

[0045] In one embodiment, Figure 2 As shown in the figure, a method for localizing and detecting degenerative areas in cervical MRI images based on deep learning is provided. Figure 1 The server in the example is used to illustrate the following steps:

[0046] S201: Acquire original cervical spine MRI images and perform preprocessing.

[0047] In the embodiment of the present application, first, the original three-dimensional cervical spine MRI image is obtained, and its sagittal image is extracted and preprocessed to obtain cervical intervertebral disc classification training sample data. Specifically, the original medical image is obtained by a nuclear magnetic resonance imaging device, such as Figure 3 As shown in FIG. 1 , a T2-weighted sagittal slice is extracted from the original three-dimensional cervical MRI image data format, including the intervertebral disc from cervical vertebra C23 to cervical vertebra C67; the image is manually sorted and annotated by a doctor, as shown in FIG. Figure 4As shown, the intervertebral disc target frame positioning coordinates and the classification label of the intervertebral disc degeneration grade are obtained, wherein the target frame is calibrated by a rectangular frame, it is determined that the intervertebral disc is included in the minimum rectangular frame, and the coordinates of the upper left corner and the lower right corner of the rectangular frame are saved.

[0048] Specifically, in one embodiment of the present application, the preprocessing includes:

[0049] The original cervical MRI image is labeled with cervical disc degeneration classification according to the Pfirrmann classification method.

[0050] In one embodiment of the present application, the original cervical MRI image is graded and classified for cervical disc degeneration according to the Pfirrmann grading method, and the degeneration grade can be divided into 5 levels: Level 1: usually asymptomatic, the intervertebral disc is still healthy; Level 2: there may be slight discomfort or no symptoms; Level 3: intermittent pain or discomfort may occur; Level 4: frequent pain, which may be accompanied by certain functional impairments; Level 5: persistent severe pain, significant functional impairment, and may require medical intervention.

[0051] S203: constructing a degenerative region positioning model based on the TOOD network, wherein the degenerative region positioning model includes a feature extraction network, a multi-scale self-attention mechanism feature fusion network, and a detection head network.

[0052] In an embodiment of the present application, a degenerative area positioning model is constructed based on the TOOD network, which includes a feature extraction network, a multi-scale self-attention mechanism feature fusion network and a detection head network. The brightness height features of the intervertebral disc are mainly extracted through the backbone feature extraction network structure, and a multi-scale feature map is obtained through the multi-scale self-attention mechanism feature fusion network. The extracted feature map is divided into two feature branches through the detection head network. The first branch is the positioning of the intervertebral disc target frame, and the second branch is the grading and classification of intervertebral disc degeneration. The prediction results of the above branches are screened and integrated through the fusion module according to the non-maximum suppression of the target frame information to complete the model construction.

[0053] Specifically, in one embodiment of the present application, the constructing of a degenerative region positioning model based on a TOOD network includes:

[0054] ResNet50 is used as the backbone network for extracting image features, feature pyramid network is used as the multi-scale self-attention mechanism feature fusion network, and classification head network and regression head network are used as detection head network.

[0055] In one embodiment of the present application, the feature extraction network and the multi-scale self-attention mechanism feature fusion network together constitute the feature extraction main network. Specifically, Figure 5As shown in the figure, ResNet50 is used as the backbone network for extracting image features to extract the initial feature map. The feature pyramid network is used as the multi-scale self-attention mechanism feature fusion network, and the low-level high-resolution features are fused with the high-level low-resolution features through lateral connections to help the network better understand the semantic information at different levels in the image. Starting from the high-level feature map, it is gradually upsampled downward. The low-level feature map is adjusted to have a number of channels through 1×1 convolution, and then added to the upsampled high-level feature map to generate a multi-scale feature map. The classification head network and the regression head network are used as the detection head network. The multi-scale feature map is continuously convolved to obtain the classification feature map and the regression feature map. The classification and positioning feature maps are input into the feature aligner for feature alignment to generate the task-aligned feature map. Two independent head networks are used for classification and regression respectively, and then the probability distribution of the category is predicted at each position on the classification feature map, and the bounding box regression parameters are predicted at each position on the regression feature map, which are usually four values: left, top, right, and bottom offsets.

[0056] S205: Training the degenerative region localization model based on the original cervical spine MRI image after preprocessing.

[0057] In the embodiment of the present application, the original cervical spine MRI image after preprocessing is input into the degenerative region localization model as a training set, and the initial degenerative region localization model based on the deep learning model is trained to generate a degenerative region localization model. Specifically, Figure 6 As shown, first, the cervical disc degeneration training sample data is used as the input image, input into the initial degenerative region positioning model, and the model operation is performed to obtain the model output result. Then the output result is compared with the cervical disc degeneration labeled image to obtain the model detection error.

[0058] Specifically, in one embodiment of the present application, the training of the degenerative region localization model based on the original cervical spine MRI image after preprocessing includes:

[0059] The error is back-propagated based on the network loss function, and the model parameters are updated by iterative operation. The network loss function includes a target positioning loss function and a target classification loss function.

[0060] In one embodiment of the present application, the network loss function is trained once per training cycle, the error is back-propagated using the network loss function, and the model parameters are updated iteratively. After one hundred training cycles, the model training is completed when the model training meets the convergence condition. The network loss function includes the target positioning loss function L reg And the target classification loss function L cls , network loss function L TAL =L reg +L cls, anchor box alignment metric t = s α ×μ β , where s and μ are the classification score and IOU value, respectively, and α and β are adjustable hyperparameters.

[0061] Specifically, in one embodiment of the present application, the formula of the target positioning loss function is:

[0062]

[0063] Among them, b i represents the positioning box predicted by the model, Represents the real box of the target, and uses the GIoU loss function to calculate the positioning loss of a single predicted target box. Used for joint optimization of the participating localization loss, N pos Indicates the number of anchors of positive samples.

[0064] In one embodiment of the present application, the formula of the target classification loss function is:

[0065]

[0066] Among them, s i represents the classification score of the i-th positive sample, s j represents the classification score of the jth negative sample, γ is the focusing parameter, BCE() is the binary cross entropy function, Used for joint optimization of the participating localization loss, N pos Indicates the number of anchors of positive samples, N neg Indicates the number of anchors for negative samples.

[0067] S207: Input the cervical vertebra MRI image to be predicted into the trained degenerative region positioning model, and output the degenerative region positioning detection result.

[0068] In the embodiment of the present application, after the model training is completed, the cervical MRI image to be predicted is input into the degenerative region positioning model to perform positioning detection of the degenerative region. Specifically, Figure 7As shown in the figure, the cervical MRI image is input into the cervical disc degeneration area localization model, and the deep learning model will perform continuous convolution operations on the image. The feature extraction module consists of a series of 1×1 and 3×3 convolutional layers of residual models, which are used to extract fine features of cervical MRI images. The feature extraction module can automatically extract features of different fields of view. Specifically, the initial convolution and pooling are performed first to preliminarily adjust the size and depth of the input image, and then the image features are gradually extracted through multiple residual blocks. Each residual block passes through 1×1, 3×3 and 1×1 convolutional layers and contains shortcut connections to help the network learn deeper features. After that, the feature map is reduced to a vector of fixed length for subsequent processing. Finally, the final feature extraction output is performed through the fully connected layer. The ResNet50 deep network is used to extract the initial feature map, starting from the high-level feature map and gradually upsampling. The low-level feature map is adjusted to the number of channels through 1×1 convolution and added to the upsampled high-level feature map. Finally, a multi-scale feature map is generated. Then, the feature map is predicted by the detection head network. The multi-scale feature map is continuously convolved to obtain the classification feature map and the regression feature map. The classification and positioning feature maps are input into the feature aligner for feature alignment to generate the task-aligned feature map. Two independent head networks are used for classification and regression respectively. Then, the probability distribution of the category is predicted at each position on the classification feature map, and the bounding box regression parameters are predicted at each position on the regression feature map, which are usually four values: left, top, right, and bottom offsets. The intervertebral disc location, intervertebral disc degeneration type, and confidence are output, such as Figure 8 As shown, a visual detection report containing intervertebral disc degeneration information is generated based on the output results, such as Fig. 9 shown.

[0069] In the above-mentioned method for localizing and detecting degenerative regions of cervical MRI images based on deep learning, first, the original cervical MRI images are obtained and preprocessed; then, a degenerative region localization model is constructed based on the TOOD network, and the degenerative region localization model includes a feature extraction network, a multi-scale self-attention mechanism feature fusion network and a detection head network; then, the degenerative region localization model is trained based on the original cervical MRI images after preprocessing; finally, the cervical MRI images to be predicted are input into the trained degenerative region localization model, and the degenerative region localization detection results are output. That is to say, based on the improvement of the TOOD network, a multi-scale self-attention mechanism module is added to the framework, which enhances the global feature map modeling, thereby improving the target detection accuracy of the network. In addition, the introduction of neural memory ordinary differential equations into the head network enhances the nonlinear representation ability, guides the training of the network, and enhances its robustness. Using the deep learning model to detect the cervical intervertebral disc can directly detect the degeneration grade of the intervertebral disc according to the T2-weighted image of MRI, effectively improving the efficiency of localization detection.

[0070] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0071] Based on the same inventive concept, the embodiment of the present application also provides a deep learning-based cervical vertebra MRI image degenerative area positioning detection device for implementing the deep learning-based cervical vertebra MRI image degenerative area positioning detection method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiments of one or more deep learning-based cervical vertebra MRI image degenerative area positioning detection devices provided below can refer to the limitations of the deep learning-based cervical vertebra MRI image degenerative area positioning detection method above, and will not be repeated here.

[0072] In one embodiment, Fig.10 As shown, a cervical spine MRI image degenerative region positioning detection device 1000 based on deep learning is provided, comprising: a data acquisition module 1001, a degenerative region positioning model construction module 1003, a degenerative region positioning model training module 1005 and a degenerative region positioning detection module 1007, wherein:

[0073] The data acquisition module 1001 is used to acquire the original cervical spine MRI image and perform preprocessing.

[0074] The degenerative region localization model construction module 1003 is used to construct a degenerative region localization model based on the TOOD network, and the degenerative region localization model includes a feature extraction network, a multi-scale self-attention mechanism feature fusion network and a detection head network.

[0075] The degenerative region localization model training module 1005 is used to train the degenerative region localization model based on the original cervical spine MRI image after preprocessing.

[0076] The degenerative region positioning detection module 1007 is used to input the cervical vertebra MRI image to be predicted into the trained degenerative region positioning model and output the degenerative region positioning detection result.

[0077] In one embodiment of the present application, the data acquisition module is further used to:

[0078] The original cervical MRI image is labeled with cervical disc degeneration classification according to the Pfirrmann classification method.

[0079] In one embodiment of the present application, the degenerative region positioning model building module is also used to:

[0080] ResNet50 is used as the backbone network for extracting image features, feature pyramid network is used as the multi-scale self-attention mechanism feature fusion network, and classification head network and regression head network are used as detection head network.

[0081] In one embodiment of the present application, the degenerative region positioning model training module is further used to:

[0082] The error is back-propagated based on the network loss function, and the model parameters are updated by iterative operation. The network loss function includes a target positioning loss function and a target classification loss function.

[0083] In one embodiment of the present application, the formula of the target positioning loss function is:

[0084]

[0085] Among them, b i represents the positioning box predicted by the model, Represents the real box of the target, and uses the GIoU loss function to calculate the positioning loss of a single predicted target box. Used for joint optimization of the participating localization loss, N pos Indicates the number of anchors of positive samples.

[0086] In one embodiment of the present application, the formula of the target classification loss function is:

[0087]

[0088] Among them, s i represents the classification score of the i-th positive sample, s j represents the classification score of the jth negative sample, γ is the focusing parameter, BCE() is the binary cross entropy function, Used for joint optimization of the participating localization loss, N pos Indicates the number of anchors of positive samples, N neg Indicates the number of anchors for negative samples.

[0089] Each module in the above-mentioned deep learning-based cervical MRI image degenerative area positioning detection device can be implemented in whole or in part by software, hardware, and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0090] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.11 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for localizing and detecting degenerative areas of cervical MRI images based on deep learning is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0091] Those skilled in the art will understand that Fig.11 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0092] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0093] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0094] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0095] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0096] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0097] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0098] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for localizing and detecting degenerative areas in cervical MRI images based on deep learning, characterized in that: The method comprises: Obtain original cervical spine MRI images and perform preprocessing; A degenerative region localization model is constructed based on the TOOD network, wherein the degenerative region localization model includes a feature extraction network, a multi-scale self-attention mechanism feature fusion network, and a detection head network; Training the degenerative area localization model based on the original cervical spine MRI image after preprocessing; The cervical MRI image to be predicted is input into the trained degenerative area positioning model, and the degenerative area positioning detection result is output.

2. The method for localizing and detecting degenerative areas in cervical MRI images based on deep learning according to claim 1, characterized in that: The pre-processing comprises: The original cervical MRI image is labeled with cervical disc degeneration classification according to the Pfirrmann classification method.

3. The method for localizing and detecting degenerative areas in cervical MRI images based on deep learning according to claim 1, characterized in that: The degenerative area positioning model constructed based on the TOOD network includes: ResNet50 is used as the backbone network for extracting image features, feature pyramid network is used as the multi-scale self-attention mechanism feature fusion network, and classification head network and regression head network are used as detection head network.

4. The method for localizing and detecting degenerative areas in cervical MRI images based on deep learning according to claim 1, characterized in that: The training of the degenerative area localization model based on the original cervical MRI image after preprocessing includes: The error is back-propagated based on the network loss function, and the model parameters are updated by iterative operation. The network loss function includes a target positioning loss function and a target classification loss function.

5. The method for localizing and detecting degenerative areas in cervical MRI images based on deep learning according to claim 4, characterized in that: The formula of the target positioning loss function is: Among them, b i represents the positioning box predicted by the model, Represents the real box of the target, and uses the GIoU loss function to calculate the positioning loss of a single predicted target box. Used for joint optimization of the participating localization loss, N pos Indicates the number of anchors of positive samples.

6. The method for localizing and detecting degenerative areas in cervical MRI images based on deep learning according to claim 4, characterized in that: The formula of the target classification loss function is: Among them, s i represents the classification score of the i-th positive sample, s j represents the classification score of the jth negative sample, γ is the focusing parameter, BCE() is the binary cross entropy function, Used for joint optimization of the participating localization loss, N pos Indicates the number of anchors of positive samples, N neg Indicates the number of anchors of negative samples.

7. A deep learning-based degenerative area positioning detection device for cervical MRI images, characterized in that: The device comprises: A data acquisition module is used to acquire original cervical spine MRI images and perform preprocessing; A degenerative region localization model construction module, used to construct a degenerative region localization model based on the TOOD network, wherein the degenerative region localization model includes a feature extraction network, a multi-scale self-attention mechanism feature fusion network, and a detection head network; A degenerative region positioning model training module, used for training the degenerative region positioning model based on the original cervical spine MRI image after preprocessing; The degenerative region positioning detection module is used to input the cervical MRI image to be predicted into the trained degenerative region positioning model and output the degenerative region positioning detection result.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images