A carpal tunnel syndrome prediction method and device based on b-mode ultrasound video

Abstract

The application discloses a carpal tunnel syndrome prediction method and device based on B-mode ultrasonic video, and the method comprises the following steps: acquiring a target carpal tunnel B-mode ultrasonic video; inputting the target ultrasonic video into a trained video segmentation model, determining a median nerve region of a carpal tunnel of each video frame in the target carpal tunnel B-mode ultrasonic video through the video segmentation model, and determining a target prediction parameter based on the median nerve region of each video frame; inputting the target prediction parameter into a trained carpal tunnel syndrome classification model, and determining a prediction result of the target carpal tunnel B-mode ultrasonic video through the carpal tunnel syndrome classification model. The method realizes efficient, accurate and interpretable carpal tunnel syndrome prediction through automatic segmentation and classification models, thereby improving prediction efficiency and consistency, reducing the work burden of doctors, and optimizing the utilization of medical resources.

Description

Technical Field

[0001] This application relates to the field of medical imaging technology, and in particular to a method and device for predicting carpal tunnel syndrome based on B-mode ultrasound video. Background Technology

[0002] Carpal tunnel syndrome is one of the most common peripheral neuropathy disorders, characterized by compression of the median nerve in the wrist. The traditional method for predicting carpal tunnel syndrome is electrophysiological prediction testing, which, although widely considered the gold standard, suffers from practical problems such as high cost, time-consuming procedure, and invasiveness.

[0003] In recent years, with the development of ultrasound technology, especially the innovation of high-frequency ultrasound transducers, the application of ultrasound technology in the prediction of carpal tunnel syndrome has gradually increased. Ultrasound prediction has advantages such as being non-invasive, simple to operate, and radiation-free. It can provide a basis for prediction by measuring parameters such as the cross-sectional area and swelling rate of the median nerve.

[0004] However, existing ultrasound prediction methods typically require physicians to manually locate the frames with the largest and smallest cross-sectional areas of the median nerve in the ultrasound video, and to manually delineate and measure the Region of Interest (ROI). This method is not only time-consuming and labor-intensive, but the prediction results are also highly dependent on the physician's experience, resulting in significant subjectivity and uncertainty. Furthermore, there are currently no clear standards to guide the specific operational procedures and evaluation criteria for ultrasound prediction of carpal tunnel syndrome, leading to poor standardization and consistency in the prediction process and results. While some automated prediction technologies based on deep learning exist, these technologies are usually based on static image classification, requiring physicians to manually select specific frames, and lack utilization of global information from the ultrasound video. This may result in insufficient prediction accuracy, and the interpretability of these automated prediction systems is also inadequate. Summary of the Invention

[0005] The technical problem this application aims to solve is to provide a method and device for predicting carpal tunnel syndrome based on B-mode ultrasound video, addressing the shortcomings of existing technologies, in order to solve the problem of how to provide an efficient and automated method for predicting carpal tunnel syndrome, thereby reducing the subjectivity and uncertainty of prediction.

[0006] To address the aforementioned technical problems, a first aspect of this application provides a method for predicting carpal tunnel syndrome based on B-mode ultrasound video, the method comprising:

[0007] Acquire B-mode ultrasound video of the target carpal tunnel;

[0008] The target ultrasound video is input into a trained video segmentation model, and the carpal tunnel median nerve region in each video frame of the target carpal tunnel B-mode ultrasound video is determined by the video segmentation model. The target prediction parameters are determined based on the carpal tunnel median nerve region in each video frame.

[0009] The target prediction parameters are input into a trained carpal tunnel syndrome classification model, and the prediction result of the target carpal tunnel B-mode ultrasound video is determined by the carpal tunnel syndrome classification model.

[0010] In one possible implementation of this embodiment, the process of acquiring the B-mode ultrasound video of the target carpal tunnel specifically includes:

[0011] Acquire B-mode ultrasound video of the carpal tunnel;

[0012] The carpal tunnel B-mode ultrasound video is preprocessed to obtain the target carpal tunnel B-mode ultrasound video. The preprocessing operation includes desensitization operation and / or pre-cropping operation.

[0013] In one possible implementation of this embodiment, the pre-cropping operation includes:

[0014] The location range of the median nerve region is determined based on all video frames in the carpal tunnel B-mode ultrasound video. Each frame of the carpal tunnel B-mode ultrasound video is then cropped based on the location range, wherein the location range includes the coordinates of the upper left, upper right, lower left, and lower right corners of the video frame.

[0015] In one possible implementation of this embodiment, the process of determining the target prediction parameters based on the median carpal tunnel nerve region of each video frame specifically includes:

[0016] A first prediction parameter is determined based on the median nerve region of the carpal tunnel in each video frame. The first prediction parameter includes: the circumference, cross-sectional area, anteroposterior diameter, and flattening of the median nerve.

[0017] Target prediction parameters are determined based on the first prediction parameters. The target prediction parameters include: maximum to minimum perimeter ratio, swelling rate, maximum flattening rate, maximum to minimum anteroposterior diameter ratio, and maximum cross-sectional area.

[0018] In one possible implementation of this embodiment, the trained video segmentation model includes an encoder module, a decoder module, and a target perceptron module. The encoder module includes several Transformer blocks, each Transformer block including several encoding layers and merging layers. The encoding layers include self-attention layers and hybrid feedforward neural network layers. The decoder module includes upsampling layers and multilayer perceptron layers.

[0019] In one possible implementation of this embodiment, the process of determining the median carpal tunnel nerve region in each video frame of the target carpal tunnel B-mode ultrasound video using the video segmentation model specifically includes:

[0020] The initial image of each video frame in the B-mode ultrasound video of the target carpal tunnel is obtained, and the initial images are overlapped and cropped to obtain several image blocks with a preset resolution.

[0021] The image blocks with preset resolutions are converted into first feature vectors corresponding to the image blocks through convolution operations.

[0022] The first feature vector is input into the encoder module for processing to obtain the second feature vector, which is a multi-scale image feature vector.

[0023] The third feature vector is obtained by performing upsampling and concatenation operations on the second feature vector based on the upsampling layer of the decoder module.

[0024] The target perceptron module performs fusion processing on the third feature vector to obtain the median nerve region of the carpal tunnel in each video frame of the B-mode ultrasound video of the target carpal tunnel.

[0025] In one possible implementation of this embodiment, the process of obtaining the third feature vector by performing upsampling and concatenation operations on the second feature vector based on the upsampling module specifically includes:

[0026] The multilayer perceptron layer of the decoder module performs feature fusion processing on the second feature vector to obtain a low-dimensional feature vector;

[0027] The third feature vector is obtained by performing linear interpolation on the low-dimensional feature vector.

[0028] A second aspect of this application provides an automatic prediction device for carpal tunnel syndrome based on B-mode ultrasound video, wherein the automatic prediction device for carpal tunnel syndrome based on B-mode ultrasound video specifically includes:

[0029] The video acquisition module is used to acquire B-mode ultrasound video of the target carpal tunnel;

[0030] The parameter determination module is used to input the target ultrasound video into a trained video segmentation model, determine the median carpal tunnel nerve region of each video frame in the target carpal tunnel B-mode ultrasound video through the video segmentation model, and determine the target prediction parameters based on the median carpal tunnel nerve region of each video frame.

[0031] The prediction module is used to input the target prediction parameters into a trained carpal tunnel syndrome classification model, and to determine the prediction result of the target carpal tunnel B-mode ultrasound video through the carpal tunnel syndrome classification model.

[0032] A third aspect of this application provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps in the carpal tunnel syndrome prediction method based on B-mode ultrasound video as described above.

[0033] A fourth aspect of this application provides a terminal device, comprising: a processor, a memory, and a communication bus; wherein the memory stores a computer-readable program executable by the processor;

[0034] The communication bus enables communication between the processor and the memory;

[0035] When the processor executes the computer-readable program, it implements the steps in the carpal tunnel syndrome prediction method based on B-mode ultrasound video as described in any of the preceding claims.

[0036] Beneficial effects:

[0037] Compared with the prior art, this application provides a method and device for predicting carpal tunnel syndrome based on B-mode ultrasound video. The method acquires B-mode ultrasound video of the target carpal tunnel through a video acquisition module, providing a non-invasive, low-cost, and easily accessible source of carpal tunnel syndrome prediction data, thereby solving the problems of high invasiveness and high cost of traditional electrophysiological examinations.

[0038] Secondly, the method inputs the target ultrasound video into a trained video segmentation model to automatically segment the median nerve region of the carpal tunnel in each video frame, thereby achieving automatic and accurate segmentation of the median nerve region, improving the consistency and efficiency of prediction, and solving the problems of strong subjectivity and time-consuming and laborious manual delineation of ROI in traditional methods.

[0039] Furthermore, the method automatically determines target prediction parameters, including circumference, cross-sectional area, anteroposterior diameter, and flattening ratio, based on the median nerve region of the carpal tunnel in each video frame. This provides a quantitative and standardized prediction parameter, thereby improving the scientific nature and objectivity of the prediction and solving the problems of high subjectivity and insufficient accuracy of manually measured parameters in the prior art.

[0040] Furthermore, the method automatically generates prediction results by inputting the target prediction parameters into a trained carpal tunnel syndrome classification model, thereby achieving automated and accurate prediction of carpal tunnel syndrome and improving the consistency and repeatability of predictions.

[0041] Furthermore, the method also uses an encoder module and a decoder module to perform multi-scale feature extraction and upsampling concatenation to ensure the accuracy of feature extraction, thereby solving the problems of information loss and insufficient accuracy in the feature extraction process. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 A flowchart of a carpal tunnel syndrome prediction method based on B-mode ultrasound video provided in this application.

[0044] Figure 2 This application provides a flowchart of a method for predicting carpal tunnel syndrome based on B-mode ultrasound video.

[0045] Figure 3 This is a structural diagram of the video segmentation model for a carpal tunnel syndrome prediction method based on B-mode ultrasound video provided in this application.

[0046] Figure 4 This is a structural principle diagram of a carpal tunnel syndrome prediction method based on B-mode ultrasound video provided in this application.

[0047] Figure 5 A schematic diagram of the terminal device provided in this application. Detailed Implementation

[0048] This application provides a method and device for predicting carpal tunnel syndrome based on B-mode ultrasound video. To make the purpose, technical solution, and effects of this application clearer and more explicit, the following detailed description is provided with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining this application and are not intended to limit this application.

[0049] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this application means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0050] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0051] It should be understood that the sequence number and size of each step in this embodiment do not imply the order of execution. The execution order of each process is determined by its function and internal logic, and should not constitute any limitation on the implementation process of this application embodiment.

[0052] Significant progress has been made in the prediction of carpal tunnel syndrome. Traditional electrophysiological prediction methods, while widely considered the gold standard, suffer from practical problems such as high cost, time-consuming procedures, and invasiveness. These issues greatly limit the widespread adoption and application of electrophysiological prediction.

[0053] In recent years, with the development of ultrasound technology, especially the innovation of high-frequency ultrasound transducers, ultrasound technology has been widely used in the prediction of carpal tunnel syndrome. Ultrasound technology has advantages such as being non-invasive, simple to operate, and radiation-free. It can provide predictive evidence by measuring parameters such as the cross-sectional area and swelling rate of the median nerve. However, although ultrasound prediction has to some extent solved the problems of invasiveness and high cost associated with traditional electrophysiological examinations, existing ultrasound prediction techniques still have some shortcomings.

[0054] Existing ultrasound prediction methods typically require physicians to manually locate the frames with the largest and smallest cross-sectional areas of the median nerve in the ultrasound video, and to manually delineate and measure the Region of Interest (ROI). This method is not only time-consuming and labor-intensive, but the prediction results are also highly dependent on the physician's experience, resulting in significant subjectivity and uncertainty. Furthermore, there are currently no clear standards to guide the specific operational procedures and evaluation criteria for ultrasound prediction of carpal tunnel syndrome, leading to poor standardization and consistency in the prediction process and results.

[0055] To overcome these problems, several deep learning-based automated prediction techniques have been proposed. These techniques typically rely on image classification methods, automatically extracting image features for classification and prediction through model training. However, these automated techniques also have limitations. First, they often require doctors to manually select specific frames, increasing the complexity and subjectivity of the operation. Second, existing automated prediction methods often lack utilization of global information from ultrasound videos, relying solely on static images for prediction, which may lead to insufficient prediction accuracy. Furthermore, the interpretability of these systems is also problematic, making them difficult for doctors to trust and accept.

[0056] Therefore, although existing technologies have made some progress in predicting carpal tunnel syndrome, several problems remain to be addressed. These problems include strong subjectivity in prediction, lack of standardization, insufficient interpretability of automated prediction technologies, and inadequate utilization of global information. Technical solutions to these problems need to improve prediction accuracy and efficiency while ensuring the objectivity and standardization of the prediction process, providing physicians with more reliable auxiliary prediction tools.

[0057] To address the aforementioned issues, this application discloses a method for predicting carpal tunnel syndrome based on B-mode ultrasound video. The method includes acquiring a target carpal tunnel B-mode ultrasound video; inputting the target ultrasound video into a trained video segmentation model to determine the median nerve region of the carpal tunnel in each video frame of the target carpal tunnel B-mode ultrasound video; determining target prediction parameters based on the median nerve region of each video frame; inputting the target prediction parameters into a trained carpal tunnel syndrome classification model; and determining the prediction result of the target carpal tunnel B-mode ultrasound video through the carpal tunnel syndrome classification model. This method, through automated segmentation and classification models, achieves non-invasive, efficient, accurate, and interpretable prediction of carpal tunnel syndrome, thereby improving prediction efficiency and consistency, reducing the workload of doctors, and optimizing the utilization of medical resources.

[0058] The application content will be further explained below with reference to the accompanying drawings and the description of the embodiments.

[0059] This embodiment provides a method for predicting carpal tunnel syndrome based on B-mode ultrasound video, such as... Figure 1 and Figure 2 As shown, the method includes:

[0060] S100, Acquire B-mode ultrasound video of the target carpal tunnel.

[0061] Specifically, the process of acquiring B-mode ultrasound video of the target carpal tunnel includes:

[0062] S110. Obtain B-mode ultrasound video of the carpal tunnel.

[0063] Specifically, the B-mode ultrasound video is a sequence of two-dimensional images generated by an ultrasound probe emitting ultrasound waves and receiving echo signals. These two-dimensional images are used to display the anatomical structures within the carpal tunnel, including the median nerve. The acquisition of the target carpal tunnel B-mode ultrasound video can be achieved by performing an ultrasound scan of the patient's wrist using a B-mode ultrasound device.

[0064] S120. Perform preprocessing operations on the carpal tunnel B-mode ultrasound video to obtain the target carpal tunnel B-mode ultrasound video. The preprocessing operations include desensitization operations and / or pre-cropping operations.

[0065] Specifically, the desensitization process refers to the process of desensitizing ultrasound videos before data processing to protect patient privacy. This desensitization includes removing any personal patient information that may be present in the video, such as name and medical record number. The desensitized video retains only essential medical imaging data, ensuring patient privacy and security.

[0066] The pre-cutting operation includes:

[0067] Based on all video frames in the carpal tunnel B-mode ultrasound video, the location range of the median nerve region is determined. Each frame of the carpal tunnel B-mode ultrasound video is then cropped based on this location range, where the location range includes the coordinates of the upper left, upper right, lower left, and lower right corners of the video frame. Cropping each video frame using this coordinate range preserves the possible location of the median nerve, thereby reducing the amount of video data and lowering the computational complexity of subsequent processing.

[0068] S200. Input the target ultrasound video into the trained video segmentation model, determine the median carpal tunnel nerve region of each video frame in the target carpal tunnel B-mode ultrasound video through the video segmentation model, and determine the target prediction parameters based on the median carpal tunnel nerve region of each video frame.

[0069] Specifically, such as Figure 3 As shown, the trained video segmentation model includes an encoder module, a decoder module, and a target perceptron module.

[0070] The encoder module includes several Transformer blocks, each containing several encoding layers and merging layers. The encoding layers include self-attention layers and hybrid feedforward neural network layers. The encoder module extracts multi-scale features from the input image through multi-layer stacking. Specifically, the self-attention layer calculates the correlation between each position in the input feature sequence and other positions, capturing global information. The hybrid feedforward neural network layer performs non-linear transformations on the features at each position, enhancing feature representation capabilities.

[0071] The decoder module comprises an upsampling layer and a multi-layer perceptron layer, used to upsample and fuse the multi-scale features extracted by the encoder module. The upsampling layer uses bilinear interpolation to upsample the low-resolution feature map to a high resolution, while the MLP layer further processes the upsampled features to generate the final segmentation result.

[0072] The target perceptron module fuses the features output by the decoder module to obtain the accurate segmentation result of the median nerve in each frame of the image.

[0073] Before the trained video segmentation model is further trained, the image data used for training can undergo data augmentation. Data augmentation refers to the process of transforming and expanding the training image data before model training to generate diverse datasets. This process can improve the model's generalization ability and avoid overfitting. In the field of medical image processing, data augmentation can simulate different imaging conditions and variations, improving the model's adaptability to real-world clinical scenarios.

[0074] The data augmentation operations specifically include:

[0075] Translation: The image is translated horizontally and vertically. Translation can simulate different ultrasound probe positions, enhancing the model's ability to recognize the median nerve in different locations. By randomly translating the image horizontally and vertically by a certain number of pixels (e.g., ±10 pixels), it is ensured that the median nerve region remains within the image range.

[0076] Horizontal and vertical flipping: Flipping images horizontally and vertically. Flipping operations can increase data diversity and avoid model bias towards specific orientations. By randomly selecting a certain percentage of images for horizontal and vertical flipping (e.g., 50%), the consistency of the flipped images with the original images in terms of anatomical structure is ensured.

[0077] Rotation: Rotating the image by a certain angle. Rotation can enhance the model's adaptability to different scanning angles. By randomly rotating the image by a certain angle (e.g., ±15 degrees) and then appropriately cropping or padding the rotated image, the image size remains consistent.

[0078] Scaling: The image is scaled. Scaling can simulate different probe distances, enhancing the model's ability to recognize median nerves of different sizes. This is achieved by randomly selecting a certain proportion of the image for scaling (e.g., 0.8x to 1.2x), and then appropriately cropping or padding the scaled image to maintain consistent image size.

[0079] Stretch: Stretch the image. Stretching can enhance the model's adaptability to different deformations, especially the deformation of the median nerve under different compression conditions. Maintain consistent image size by randomly stretching the image horizontally and vertically (e.g., ±10%) and then appropriately cropping or padding the stretched image.

[0080] Random cropping: Images are randomly cropped to obtain different local regions. Random cropping can simulate different visual ranges, enhancing the model's ability to handle partially missing information. By randomly selecting a certain proportion of the image for cropping (e.g., cropping 90% of the original image), it is ensured that the cropped image still contains the central neural region.

[0081] By employing the aforementioned data augmentation operations, diverse training datasets can be generated, significantly improving the model's generalization ability and robustness. Specific effects include:

[0082] Enhance the model's adaptability to different imaging conditions: Through translation, flipping, rotation, scaling and stretching operations, the model can adapt to different ultrasound probe positions, orientations and distances, improving the ability to identify the median nerve.

[0083] Reduce overfitting: By increasing the diversity of data, avoid the model from overfitting the training data and improve the prediction accuracy on unseen data.

[0084] Simulating clinical scenario variations: Data augmentation can simulate imaging variations under different operating conditions in actual clinical practice, improving the model's performance in real-world application scenarios.

[0085] Furthermore, the image data used for training is annotated frame-by-frame and pixel-by-pixel by experienced ultrasound physicians to obtain training image data and its corresponding labels. This training image data and corresponding labels are then fed into the video segmentation model for training. Under the supervision of the labels, the video segmentation model can learn the mapping relationship between each pixel in each image and its corresponding label; that is, each pixel corresponds to either the median nerve or the background. After training, the video segmentation model can automatically delineate the median region of interest (ROI) of the input image data. By calculating the ROI, the median nerve data of the input image data can be obtained.

[0086] Specifically, the process of determining the median carpal tunnel nerve region in each video frame of the target carpal tunnel B-mode ultrasound video using the video segmentation model includes:

[0087] S210. Obtain the initial image of each video frame in the B-mode ultrasound video of the target carpal tunnel, and overlap and crop the initial images to obtain several image blocks with a preset resolution.

[0088] S211. The plurality of image blocks with preset resolutions are converted into first feature vectors corresponding to the image blocks by convolution operation;

[0089] S212. The first feature vector is input into the encoder module for processing to obtain a second feature vector, wherein the second feature vector is a multi-scale image feature vector.

[0090] Specifically, in step S210, the B-mode ultrasound video of the target carpal tunnel is first extracted frame by frame to obtain an initial image for each frame. The initial image has a fixed resolution, for example, H×W (height × width). Then, the size of the cropping window is set to a fixed value (e.g., 4×4 pixels), and cropping is performed with a certain step size. The step size is usually smaller than the size of the cropping window to ensure a certain overlap between image blocks. A typical step size setting can be half the size of the cropping window (e.g., 2 pixels). This overlapping cropping method ensures that information is not lost at the boundaries of the image blocks, helping to maintain the local continuity of the image, thereby capturing more detailed information and improving the effect of subsequent feature extraction. Through the above overlapping cropping operation, each initial frame image is converted into several image blocks with a preset resolution. These image blocks will be used for subsequent convolution operations and feature extraction.

[0091] In step S211, a convolution operation is first applied to each image patch with a preset resolution obtained in step S210. The convolution operation uses a fixed-size convolution kernel (such as 3×3 or 5×5) and is performed on the image patch using a sliding window. The convolution operation can extract local feature information in the image patch, such as edges, textures, and shapes. It should be noted that the convolution result of each image patch is usually a multi-channel feature map. The feature map is unfolded into a one-dimensional vector to obtain the corresponding first feature vector.

[0092] In step S212, these feature vectors are processed by the encoder module to obtain multi-scale image feature vectors. Specifically:

[0093] The encoder module adopts a Transformer block-based architecture. Each Transformer block contains several self-attention layers and hybrid feedforward neural network layers (Mix-FFN).

[0094] The self-attention layer is used to calculate the correlation between each position in the input feature vector and other positions, capture global dependencies, and help the model understand the contextual information between image patches; the hybrid feedforward neural network layer is used to perform nonlinear transformation on the feature vector of each position to enhance the expressive power of the model.

[0095] The input first feature vector is processed by multiple Transformer blocks of the encoder, with each block performing self-attention calculation and feedforward neural network transformation. The encoder module extracts multi-scale image feature vectors by downsampling layer by layer. For example, the first feature vector is input into the first Transformer block to obtain a processed feature vector; the processed feature vector is then input into the next Transformer block, and the above operation is repeated until all Transformer blocks are traversed; the output of each Transformer block is a feature vector at a different scale, which captures information about the image at different scales.

[0096] The final feature vector output by the encoder module is the second feature vector. This second feature vector contains multi-scale image information and can describe the details and global structure of the central neural region.

[0097] S213. Based on the upsampling layer of the decoder module, the second feature vector is upsampled and concatenated to obtain the third feature vector;

[0098] Specifically, the process of obtaining the third feature vector by performing upsampling and concatenation operations on the second feature vector based on the upsampling module includes:

[0099] S2131. The multilayer perceptron layer of the decoder module performs feature fusion processing on the second feature vector to obtain a low-dimensional feature vector;

[0100] S2132. Perform linear interpolation on the low-dimensional feature vector to obtain the third feature vector.

[0101] Specifically, in step S213, the upsampling operation refers to the process of restoring a low-resolution feature map to a high-resolution version. Through upsampling, the information in the feature map can be refined, making it closer to the resolution of the original image. The stitching operation refers to merging multiple feature maps along the channel dimension. By stitching, feature information at different scales can be combined, improving the model's ability to capture image details.

[0102] In step S2131, feature fusion refers to integrating multiple feature maps, extracting key information, and removing redundant features to obtain a more representative low-dimensional feature vector. The multilayer perceptron (MLP) layer of the decoder module is used for feature fusion processing. Specifically, the stitched high-resolution feature map is input into the MLP layer. The MLP layer contains multiple fully connected layers. Each fully connected layer performs a weighted summation of the input features and transforms them using a non-linear activation function. Through the combination of multiple fully connected layers and activation functions, the MLP layer can fully fuse the feature maps and extract the most representative low-dimensional feature vector. The low-dimensional feature vector refers to the feature representation after feature fusion processing, which has fewer dimensions but contains richer semantic information.

[0103] In step S2132, the linear interpolation is an interpolation method that generates new data points by performing linear interpolation calculations between existing data points. Linear interpolation can improve resolution while maintaining data smoothness. The process of performing linear interpolation on the low-dimensional feature vector specifically includes:

[0104] For each low-dimensional feature vector, its corresponding position in the original image is found, and a new feature value is calculated using a linear combination of adjacent data points. Through multiple linear interpolation operations, the low-dimensional feature vector is progressively upscaled to a higher resolution, resulting in a third feature vector. This third feature vector refers to the high-resolution feature representation after linear interpolation, containing rich detail and precise semantic information. It is important to note that the third feature vector after linear interpolation has high resolution and accuracy, providing high-quality feature support for subsequent segmentation tasks.

[0105] S214. The target perceptron module performs fusion processing on the third feature vector to obtain the median nerve region of the carpal tunnel in each video frame of the B-mode ultrasound video of the target carpal tunnel.

[0106] Specifically, the process of determining the target prediction parameters based on the median nerve region of the carpal tunnel in each video frame includes:

[0107] S220. Determine a first prediction parameter based on the median nerve region of the carpal tunnel in each video frame. The first prediction parameter includes: the circumference, cross-sectional area, anteroposterior diameter, and flattening of the median nerve.

[0108] S221. Determine target prediction parameters based on the first prediction parameters. The target prediction parameters include: maximum to minimum perimeter ratio, swelling rate, maximum flattening rate, maximum to minimum anteroposterior diameter ratio, and maximum cross-sectional area.

[0109] After obtaining the median neural region of each video frame output by the video segmentation model, these regions need to be further analyzed to determine preliminary prediction parameters. The process of determining the first prediction parameters specifically includes:

[0110] Determine the perimeter of the median nerve: Perform boundary detection on the segmented median nerve region in each frame and calculate its perimeter. The perimeter of the median nerve refers to the total length of its outer contour, which can be obtained using a precise boundary detection algorithm.

[0111] Determine the cross-sectional area of ​​the median nerve: The cross-sectional area is obtained by calculating the total number of pixels within the median nerve region. This cross-sectional area refers to the projected area of ​​the median nerve in the image frame and is used to reflect the degree of nerve compression.

[0112] Determining the anteroposterior diameter of the median nerve: The anteroposterior diameter is obtained by measuring the longest and shortest diameters of the median nerve region. This anteroposterior diameter reflects the degree of deformation of the median nerve.

[0113] The flatness ratio of the median nerve is determined, which is the ratio of the long diameter to the short diameter of the median nerve region, used to determine the degree of flatness of the median nerve. The flatness ratio is calculated as the ratio of the longest diameter to the shortest diameter.

[0114] After determining the first prediction parameters for each frame, the target prediction parameters used for the final prediction can be calculated by comprehensively analyzing the first prediction parameters. The process of determining the target prediction parameters specifically includes:

[0115] The maximum-minimum circumference ratio is determined by calculating the ratio of the maximum to the minimum circumference of the median nerve in all frames of the ultrasound video. This ratio reflects the variation of the median nerve throughout the scanning process and helps to determine its compression status.

[0116] The swelling rate is determined as the percentage increase in the cross-sectional area of ​​the median nerve in a given frame relative to its cross-sectional area in the uncompressed state. The degree of nerve swelling can be assessed by comparing the cross-sectional areas in the normal and compressed states.

[0117] The maximum flattening ratio is determined by calculating the maximum flattening ratio of the median nerve across all frames in the ultrasound video. This maximum flattening ratio reflects the degree of deformation of the median nerve under the most severe compression.

[0118] The ratio of the maximum to minimum anteroposterior diameter is determined by calculating the ratio of the maximum to minimum anteroposterior diameter of the median nerve across all frames in the ultrasound video. This ratio describes the dynamic changes of the median nerve during the scanning process and helps determine its compression status.

[0119] The maximum cross-sectional area is determined by taking the maximum cross-sectional area of ​​the median nerve across all frames in the ultrasound video. This maximum cross-sectional area typically appears in the frame where the median nerve is most severely compressed and is a key parameter reflecting the degree of nerve swelling.

[0120] S300. Input the target prediction parameters into the trained carpal tunnel syndrome classification model, and determine the prediction result of the target carpal tunnel B-mode ultrasound video through the carpal tunnel syndrome classification model.

[0121] Based on the above steps, first prediction parameters, including perimeter, cross-sectional area, anteroposterior diameter, and flattening ratio, have been extracted from the median neural segmentation region of each frame. Furthermore, target prediction parameters for prediction, such as the maximum-to-minimum perimeter ratio, swelling ratio, maximum flattening ratio, maximum-to-minimum anteroposterior diameter ratio, and maximum cross-sectional area, are calculated based on these first prediction parameters.

[0122] This application uses Support Vector Machine (SVM) as the classification model for carpal tunnel syndrome. SVM is a supervised learning model suitable for classification tasks with high-dimensional data and has strong generalization ability.

[0123] In step S300, the target prediction parameters extracted from the new ultrasound video are first input into the trained SVM classification model.

[0124] Subsequently, the SVM classification model processes the input prediction parameters and calculates the classification result based on the mapping relationship learned during training:

[0125] Map the input target prediction parameters to a high-dimensional space;

[0126] Based on the classification boundary of the hyperplane, determine whether the input sample belongs to the normal category or the carpal tunnel syndrome category.

[0127] The output of the classification model is either normal or complete carpal tunnel syndrome.

[0128] In addition, before training the video segmentation model and the carpal tunnel syndrome classification model, the model also needs to undergo:

[0129] For model hyperparameter settings: Select preset hyperparameters, such as learning rate, batch size, and number of training epochs;

[0130] For the loss function setting of the model: use the cross-entropy loss function or the Di ce loss function to measure the difference between the model segmentation result and the label;

[0131] For model optimization algorithm settings: Use the Adam optimizer or other advanced optimization algorithms to gradually adjust the model parameters and minimize the loss function;

[0132] The training process for the video segmentation model and the carpal tunnel syndrome classification model also includes:

[0133] Initialization: Randomly initialize model parameters;

[0134] Forward propagation: The input data is propagated forward through the model to calculate the output result;

[0135] Loss calculation: Calculate the loss function value based on the model output and label data;

[0136] Backpropagation: The backpropagation algorithm is used to calculate the gradient of the loss function with respect to the model parameters.

[0137] Parameter update: Update model parameters using optimization algorithms;

[0138] Repeat the above steps until the training loss converges or the preset number of training rounds is reached.

[0139] The optimization and evaluation process of the video segmentation model and the carpal tunnel syndrome classification model includes:

[0140] During the training process of the model, the model performance can be optimized using the following methods:

[0141] Use L2 regularization or Dropout to prevent model overfitting; employ a learning rate decay strategy to gradually reduce the learning rate and ensure the model is more stable in the later stages of training; regularly save model parameters to prevent model loss due to training interruption and to select the optimal model parameters for subsequent applications.

[0142] During the evaluation of the model, the model's effectiveness can be assessed using the following methods:

[0143] The model performance is evaluated periodically on the validation set, and metrics such as segmentation accuracy, Dice coefficient, and IoU are calculated to ensure that the model performs well on unseen data. Cross-validation is used to further verify the model's generalization ability and stability.

[0144] Specifically, the training process of the carpal tunnel syndrome classification model includes:

[0145] During training data preparation:

[0146] First, the data source was determined. The training dataset included median nerve prediction parameters obtained through ultrasound video segmentation and parameter extraction. These median nerve prediction parameters covered various patient scenarios, including healthy individuals and patients with carpal tunnel syndrome.

[0147] Next, the labeling data was determined. This labeling data was assigned to each sample by experienced physicians based on electrophysiological measurements, clinical manifestations, and medical records. The labeling results were divided into two categories: normal and complete carpal tunnel syndrome.

[0148] During model training:

[0149] First, feature standardization is performed. In order to improve the model training effect, the input target prediction parameters are first standardized so that their mean is zero and their variance is one.

[0150] Next, SVM training is performed by inputting the standardized prediction parameters and corresponding label data into the SVM classifier. The SVM classifier finds an optimal hyperplane to classify the data into two categories: normal and carpal tunnel syndrome, thereby maximizing the classification margin.

[0151] The hyperplane selection is performed again. SVM searches for a hyperplane in high-dimensional space that maximizes the distance between sample points on both sides of the classification boundary, thereby distinguishing normal samples from carpal tunnel syndrome samples.

[0152] The kernel function is then selected. In a specific embodiment of this application, the radial basis function (RBF) is selected as the kernel function of the SVM to handle nonlinear classification problems.

[0153] Finally, parameter optimization is performed. The hyperparameters of SVM (such as the penalty parameter C and the kernel function parameter γ) are optimized using cross-validation to ensure that the model achieves optimal performance.

[0154] During the evaluation of training results:

[0155] First, a validation set evaluation is performed. The trained SVM model is evaluated using an independent validation set, and metrics such as classification accuracy, precision, recall, and F1 score are calculated.

[0156] Cross-validation is then performed to further verify the model's generalization ability and ensure that the model can maintain high classification performance on unseen data.

[0157] In summary, this embodiment provides a method for predicting carpal tunnel syndrome based on B-mode ultrasound video. By acquiring B-mode ultrasound video of the carpal tunnel and performing preprocessing operations on the video, data integrity and privacy protection are ensured. Preprocessing guarantees data privacy and processing efficiency, helping to reduce the computational complexity of subsequent processing.

[0158] Furthermore, by inputting the target ultrasound video into a trained video segmentation model and processing it frame by frame to determine the median nerve region, the encoder and decoder modules, especially the Transformer block-based design, efficiently extract multi-scale features and automatically and accurately segment the median nerve region, reducing the subjectivity and uncertainty of manual operation. In addition, various data augmentation operations, such as translation, flipping, rotation, scaling, and stretching, are used to generate diverse training datasets, improving the model's generalization ability and robustness, and enhancing its adaptability to different imaging conditions and clinical scenarios. A Support Vector Machine (SVM) classification model is employed, and through supervised learning and optimized training, automatic and accurate classification of carpal tunnel syndrome is achieved, reducing reliance on physician subjective judgment and ensuring the objectivity and standardization of the prediction process.

[0159] In other words, this application significantly reduces prediction time by employing data augmentation techniques, multi-scale feature extraction, convolutional operations, and support vector machine classification models, leveraging automated processing. Compared to traditional methods, this improves prediction efficiency, enabling faster prediction services for more patients. Furthermore, automated segmentation and quantification of prediction parameters reduce the subjectivity of manual operations, providing a scientific and objective basis for predictions and improving the accuracy and consistency of prediction results. In addition, the use of clinically meaningful target prediction parameters and interpretable SVM classification models ensures the interpretability and standardization of the prediction process, contributing to improved consistency and repeatability. The automated segmentation and classification steps significantly reduce the workload of physicians in ultrasound video analysis, lower operational complexity, and improve clinical efficiency.

[0160] Based on the above-mentioned method for predicting carpal tunnel syndrome using B-mode ultrasound video, this embodiment provides a device for predicting carpal tunnel syndrome based on B-mode ultrasound video, such as... Figure 4 As shown, the device includes:

[0161] Video acquisition module 100 is used to acquire B-mode ultrasound video of the target carpal tunnel;

[0162] The parameter determination module 200 is used to input the target ultrasound video into a trained video segmentation model, determine the carpal tunnel median nerve region of each video frame in the target carpal tunnel B-mode ultrasound video through the video segmentation model, and determine the target prediction parameters based on the carpal tunnel median nerve region of each video frame.

[0163] The prediction module 300 is used to input the target prediction parameters into a trained carpal tunnel syndrome classification model, and to determine the prediction result of the target carpal tunnel B-mode ultrasound video through the carpal tunnel syndrome classification model.

[0164] Based on the above-mentioned method for predicting carpal tunnel syndrome using B-mode ultrasound video, this application also provides a terminal device, such as... Figure 5 As shown, it includes at least one processor 20; a display screen 21; and a memory 22, and may also include a communication interface 23 and a bus 24. The processor 20, display screen 21, memory 22, and communication interface 23 can communicate with each other via the bus 24. The display screen 21 is configured to display a preset user guide interface in the initial setup mode. The communication interface 23 can transmit information. The processor 20 can call logical instructions in the memory 22 to execute the methods described in the above embodiments.

[0165] Furthermore, the logical instructions in the aforementioned memory 22 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0166] The memory 22, as a computer-readable storage medium, can be configured to store software programs, computer-executable programs, such as program instructions or modules corresponding to the methods in the embodiments of this disclosure. The processor 20 executes functional applications and data processing by running the software programs, instructions, or modules stored in the memory 22, thereby implementing the methods in the above embodiments.

[0167] The memory 22 may include a program storage area and a data storage area. The program storage area may store the operating system and application programs required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 22 may include high-speed random access memory (RAM) and non-volatile memory. Examples include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, as well as transient storage media.

[0168] Furthermore, the specific process of loading and executing multiple instruction processors in the aforementioned storage medium and terminal device has been described in detail in the above method, and will not be repeated here.

[0169] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for predicting carpal tunnel syndrome based on B-mode ultrasound video, characterized in that, The method includes: Acquire B-mode ultrasound video of the target carpal tunnel; The target carpal tunnel B-mode ultrasound video is input into a trained video segmentation model. The carpal tunnel median nerve region in each video frame of the target carpal tunnel B-mode ultrasound video is determined by the video segmentation model. Target prediction parameters are determined based on the carpal tunnel median nerve region in each video frame. The target prediction parameters are input into a trained carpal tunnel syndrome classification model, and the prediction result of the target carpal tunnel B-mode ultrasound video is determined by the carpal tunnel syndrome classification model. The process of determining the target prediction parameters based on the median nerve region of the carpal tunnel in each video frame specifically includes: Based on the median nerve region of the carpal tunnel in each video frame, a first prediction parameter is determined. The first prediction parameter includes: the circumference, cross-sectional area, anteroposterior diameter, and flatness ratio of the median nerve. The flatness ratio refers to the ratio of the long diameter to the short diameter of the median nerve region, which is used to reflect the degree of flatness of the median nerve. Target prediction parameters are determined based on the first prediction parameters. The target prediction parameters include: maximum to minimum perimeter ratio, swelling rate, maximum flattening rate, maximum to minimum anteroposterior diameter ratio, and maximum cross-sectional area. The swelling rate refers to the percentage increase in the cross-sectional area of ​​the median nerve in a certain frame relative to its cross-sectional area in the uncompressed state.

2. The method for predicting carpal tunnel syndrome based on B-mode ultrasound video according to claim 1, characterized in that, The process of acquiring B-mode ultrasound video of the target carpal tunnel specifically includes: Acquire B-mode ultrasound video of the carpal tunnel; The carpal tunnel B-mode ultrasound video is preprocessed to obtain the target carpal tunnel B-mode ultrasound video. The preprocessing operation includes desensitization operation and / or pre-cropping operation.

3. The method for predicting carpal tunnel syndrome based on B-mode ultrasound video according to claim 2, characterized in that, The pre-cutting operation includes: The location range of the median nerve region is determined based on all video frames in the carpal tunnel B-mode ultrasound video. Each frame of the carpal tunnel B-mode ultrasound video is then cropped based on the location range, wherein the location range includes the coordinates of the upper left, upper right, lower left, and lower right corners of the video frame.

4. The method for predicting carpal tunnel syndrome based on B-mode ultrasound video according to claim 1, characterized in that, The trained video segmentation model includes an encoder module, a decoder module, and a target perceptron module. The encoder module includes several Transformer blocks, each containing several encoding layers and merging layers. The encoding layers include self-attention layers and hybrid feedforward neural network layers. The decoder module includes upsampling layers and multilayer perceptron layers.

5. The method for predicting carpal tunnel syndrome based on B-mode ultrasound video according to claim 4, characterized in that, The process of determining the median carpal tunnel nerve region in each video frame of the target carpal tunnel B-mode ultrasound video using the video segmentation model specifically includes: The initial image of each video frame in the B-mode ultrasound video of the target carpal tunnel is obtained, and the initial images are overlapped and cropped to obtain several image blocks with a preset resolution. The image blocks with preset resolutions are converted into first feature vectors corresponding to the image blocks through convolution operations. The first feature vector is input into the encoder module for processing to obtain the second feature vector, which is a multi-scale image feature vector. The third feature vector is obtained by performing upsampling and concatenation operations on the second feature vector based on the upsampling layer of the decoder module. The target perceptron module performs fusion processing on the third feature vector to obtain the median nerve region of the carpal tunnel in each video frame of the B-mode ultrasound video of the target carpal tunnel.

6. The method for predicting carpal tunnel syndrome based on B-mode ultrasound video according to claim 5, characterized in that, The process of obtaining the third feature vector by performing upsampling and concatenation operations on the second feature vector based on the upsampling module specifically includes: The multilayer perceptron layer of the decoder module performs feature fusion processing on the second feature vector to obtain a low-dimensional feature vector; The third feature vector is obtained by performing linear interpolation on the low-dimensional feature vector.

7. An automatic prediction device for carpal tunnel syndrome based on B-mode ultrasound video, characterized in that, The device specifically includes: The video acquisition module is used to acquire B-mode ultrasound video of the target carpal tunnel; The parameter determination module is used to input the target carpal tunnel B-mode ultrasound video into a trained video segmentation model, determine the carpal tunnel median nerve region of each video frame in the target carpal tunnel B-mode ultrasound video through the video segmentation model, and determine the target prediction parameters based on the carpal tunnel median nerve region of each video frame. The process of determining the target prediction parameters based on the median nerve region of the carpal tunnel in each video frame specifically includes: Based on the median nerve region of the carpal tunnel in each video frame, a first prediction parameter is determined. The first prediction parameter includes: the circumference, cross-sectional area, anteroposterior diameter, and flatness ratio of the median nerve. The flatness ratio refers to the ratio of the long diameter to the short diameter of the median nerve region, which is used to reflect the degree of flatness of the median nerve. Target prediction parameters are determined based on the first prediction parameters. The target prediction parameters include: maximum to minimum perimeter ratio, swelling rate, maximum flattening rate, maximum to minimum anteroposterior diameter ratio, and maximum cross-sectional area. The swelling rate refers to the percentage increase in the cross-sectional area of ​​the median nerve in a certain frame relative to its cross-sectional area in the uncompressed state. The prediction module is used to input the target prediction parameters into a trained carpal tunnel syndrome classification model, and to determine the prediction result of the target carpal tunnel B-mode ultrasound video through the carpal tunnel syndrome classification model.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps in the carpal tunnel syndrome prediction method based on B-mode ultrasound video as described in any one of claims 1-6.

9. A terminal device, characterized in that, include: Processor and memory; The memory stores a computer-readable program that can be executed by the processor; When the processor executes the computer-readable program, it implements the steps in the carpal tunnel syndrome prediction method based on B-mode ultrasound video as described in any one of claims 1-6.

Images