Schistosomiasis B ultrasonic image fibrosis grading data dynamic analysis method

Abstract

A schistosomiasis B-mode ultrasound image fibrosis grading data dynamic analysis method relates to the technical field of image processing and medical diagnosis, and comprises the following steps: carrying out graying processing and quality screening operation on an original B-mode ultrasound image of a subject, and then carrying out texture feature extraction on the grayscale image to obtain a texture feature set; performing feature matching on the texture feature set to obtain a suspicious region of the grayscale image, performing fibrosis degree marking on the suspicious region of the grayscale image, and performing time feature extraction and spatial feature extraction on the grayscale image; the method comprises the steps of constructing a fibrosis prediction model, obtaining fibrosis prediction information of a subject according to the fibrosis prediction model and spatial features and time features of a grayscale image marked with the fibrosis degree, performing treatment measure retrieval according to the fibrosis prediction information of the subject, and obtaining corresponding treatment measures, so that the accuracy of fibrosis grading is remarkably improved; and the probability of misdiagnosis and missed diagnosis is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of image processing and medical diagnosis, in particular to a method for dynamically analyzing fibrosis classification data of schistosomiasis B-ultrasound images. Background Art

[0002] Schistosomiasis is a parasitic disease that seriously endangers human health. Liver fibrosis is one of the important pathological changes of schistosomiasis. The graded diagnosis of the degree of liver fibrosis in schistosomiasis through B-ultrasound images is of great significance for evaluating the condition and guiding treatment.

[0003] The traditional schistosomiasis B-ultrasound image fibrosis grading method mainly relies on the subjective experience of doctors, and has problems such as difficulty in ensuring diagnostic accuracy and consistency and low efficiency. In recent years, image analysis methods based on machine learning have gradually been applied to the diagnosis of schistosomiasis B-ultrasound images, but the existing methods are insufficient in processing the spatiotemporal characteristics of images and cannot fully mine the dynamic information in the B-ultrasound image sequence, resulting in the need to improve the accuracy of grading. Summary of the invention

[0004] In order to solve the above technical problems, the object of the present invention is to provide a method for dynamically analyzing schistosomiasis B-ultrasound image fibrosis classification data, comprising the following steps: Step s1: grayscale processing and quality screening operations are performed on the original B-ultrasound image of the subject, and then texture feature extraction is performed on the grayscale image to obtain a texture feature set; Step s2: performing feature matching on the texture feature set, obtaining suspicious areas of the grayscale image, marking the degree of fibrosis on the suspicious areas of the grayscale image, and performing temporal feature extraction and spatial feature extraction on the grayscale images marked with the degree of fibrosis corresponding to all original B-ultrasound images collected from the subject; Step s3: construct a fibrosis prediction model, obtain the fibrosis prediction information of the subject according to the spatial and temporal characteristics of the fibrosis prediction model and the grayscale image annotated with the degree of fibrosis, retrieve treatment measures according to the fibrosis prediction information of the subject, and obtain corresponding treatment measures.

[0005] Furthermore, the process of graying the original B-ultrasound image of the subject and performing quality screening operations includes: Collect original B-ultrasound images of multiple basic liver sections of the examinee, and obtain at least six basic liver sections for each examinee, which may include the subxiphoid sagittal section, the left subcostal transverse section, the right subcostal section, the right intercostal oblique section, etc. To improve the image diversity, multiple (such as 3) B-ultrasound images at different angles need to be taken for each section, and fine-tuned within a range of no more than 15° to the left and right of the long axis direction of the probe. After the acquisition is completed, it should be ensured that each examinee has at least 18 clear and usable B-ultrasound images. In special circumstances (such as abnormal liver morphology of the patient or too limited lesion area), other sections can be collected as appropriate, and the original B-ultrasound images are grayed to generate gray images. The gray values ​​of each pixel in the gray image are statistically analyzed to obtain a gray image histogram, in which the abscissa of the gray image histogram is the gray value and the ordinate is the frequency of occurrence of the gray value. Feature extraction is performed on the gray image histogram to obtain the gray mean and gray value distribution range; Preset the threshold interval corresponding to the grayscale mean and grayscale value distribution range, compare the grayscale mean and grayscale value distribution range corresponding to each grayscale image with the corresponding threshold interval, and eliminate the grayscale images whose corresponding grayscale mean and grayscale value distribution range are not within the corresponding threshold interval.

[0006] Furthermore, the texture feature extraction is performed on the grayscale image, and the process of obtaining the texture feature set includes: The grayscale value of each pixel in the grayscale image is quantized into discrete grayscale levels, and the fixed parameters of the grayscale co-occurrence matrix are preset. The grayscale co-occurrence matrix of the local area of ​​each pixel in the grayscale image is obtained by a sliding window, and the texture feature set of the local area of ​​each pixel in the grayscale image is obtained according to the grayscale co-occurrence matrix. The texture feature set includes contrast, energy, entropy, and inverse variance.

[0007] Furthermore, the process of performing feature matching on the texture feature set and obtaining the suspicious area of ​​the grayscale image includes: A texture feature set corresponding to a normal area is preset, and the texture feature set of the local area of ​​each pixel in the grayscale image is compared with the texture feature set corresponding to the normal area to obtain the texture feature similarity between the local area of ​​each pixel and the normal area, and the texture feature similarity between the local area of ​​each pixel and the normal area is compared with a preset texture feature similarity threshold. If the texture feature similarity between the local area of ​​the pixel and the normal area is less than the texture feature similarity threshold, the local area of ​​the pixel is marked as a suspicious area.

[0008] The calculation formula for obtaining the texture feature similarity between the local area of ​​each pixel and the normal area is: ; in, represents the texture feature similarity between the local area of ​​the i-th pixel and the normal area j, represents the contrast between the local area of ​​the i-th pixel and the normal area j, Represents the energy of the local area of ​​the i-th pixel, represents the energy of normal region j, represents the entropy of the local area of ​​the i-th pixel, represents the entropy of normal region j, represents the inverse variance of the local area of ​​the i-th pixel, represents the inverse variance of the normal region j, , , and Represents the weight factor.

[0009] Furthermore, the process of marking the degree of fibrosis in the suspicious area of ​​the grayscale image includes: A fibrosis recognition model is constructed based on deep learning, and a number of training images containing different fibrosis lesion areas are collected in advance, and the fibrosis degree of the training images containing different fibrosis lesion areas is annotated, and the judgment criteria of the fibrosis degree include: Level 0: uniform liver echo and clear blood vessel course; Level I: fish scale-like echoes are visible and the blood vessels are still clear; Level II: grid-like echoes and unclear blood vessel course; Level III: coarse grids or large light spots; the training images after fibrosis annotation are used as training data, and the fibrosis recognition model is trained using the training data to obtain a trained fibrosis recognition model; The grayscale image with the marked suspicious area is input into the fibrosis recognition model, and the fibrosis degree of the suspicious area in the grayscale image is marked according to the fibrosis recognition model.

[0010] Furthermore, the process of extracting time features and space features from grayscale images marked with fibrosis degrees corresponding to all original B-ultrasound images collected from the subject includes: The grayscale images with the degree of fibrosis marked corresponding to all the original B-ultrasound images collected from the subject are extracted, the acquisition time of the grayscale images is extracted, a grayscale image time series is generated according to the acquisition time of the grayscale images, a time convolutional neural network is constructed, the grayscale image time series is input into the time convolutional neural network, and the fibrosis degree change trend characteristics of the suspicious area are obtained according to the time convolutional neural network. At the same time, a spatial convolutional neural network is constructed to extract spatial features of the grayscale images with the degree of fibrosis marked, and obtain the spatial features of the grayscale images with the degree of fibrosis marked, wherein the spatial features include the coverage area and shape of the suspicious area.

[0011] Furthermore, a fibrosis prediction model is constructed. According to the fibrosis prediction model and the spatial characteristics and temporal characteristics of the grayscale image annotated with the degree of fibrosis, the process of obtaining the fibrosis prediction information of the subject includes: A fibrosis prediction model is constructed based on an RBF neural network, and a historical grayscale image time series of several subjects is collected in advance, and the time features and spatial features of the historical grayscale image time series of several subjects are extracted, and the time features and spatial features are used as a training set and a test set, and the training set is input into the fibrosis prediction model for training until the loss function training is stable, and the model parameters are saved, and the fibrosis prediction model is tested by the test set until it meets the preset requirements, and the fibrosis prediction model is output; The time characteristics, spatial characteristics and fibrosis degree corresponding to the grayscale image of the subject with the fibrosis degree marked are input into the fibrosis prediction model, and the fibrosis prediction information of the subject is output according to the fibrosis prediction model.

[0012] Building a deep learning-based fibrosis prediction model is a complex process that involves multiple steps such as model selection, training, validation, and testing. The following is a detailed supplementary description of this process: The present invention selects a convolutional neural network (CNN) suitable for time series analysis as a deep learning architecture. After determining the model architecture, the mean square error loss function is selected as the optimization target, and then the prepared training set is input into the selected deep learning model to start training. During the training process, the weights are continuously updated through the back propagation algorithm, so that the loss function gradually decreases until a stable state is reached. During this period, techniques such as Early Stopping are used to avoid overfitting. In addition to the basic training process, various parameters of the model are tuned through Grid Search, including learning rate, batch size, regularization coefficient, etc.

[0013] After the model training is completed and the parameters are adjusted, the final evaluation is performed through the test set to obtain the evaluation results of the model, which include classification indicators such as accuracy, recall, F1 score, etc. Based on the evaluation results on the test set, it is judged whether the model has reached the expected standards. If the requirements are met, the model parameters are saved and ready for deployment; if not ideal, it is necessary to return to a previous stage to re-examine issues such as data quality, model structure or training strategy.

[0014] Furthermore, the treatment measures are retrieved according to the fibrosis prediction information of the subject, and the process of obtaining the corresponding treatment measures includes: According to the fibrosis prediction information of the subject, the time change curve of the fibrosis degree and the regional change curve of the fibrosis lesion of the subject are obtained, and the time change curve of the fibrosis degree and the regional change curve of the fibrosis lesion of the subject are dynamically analyzed to obtain the change range of the fibrosis degree and the change range of the fibrosis lesion of the subject; A lesion treatment database is preset, and the lesion treatment database includes treatment measures corresponding to different ranges of change in fibrosis degrees and ranges of change in fibrosis lesion areas. Data is searched in the lesion treatment database according to the ranges of change in the fibrosis degree and ranges of change in fibrosis lesion areas of the subject, and the ranges of change in the fibrosis degree and ranges of change in fibrosis lesion areas corresponding to each treatment measure in the lesion database are obtained, and the similarity with the ranges of change in the fibrosis degree and ranges of change in fibrosis lesion areas of the subject is obtained, and treatment measures whose similarity meets the preset standards are obtained in the lesion database, and the treatment measures are marked as corresponding treatment measures.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. Through multi-dimensional feature extraction and fusion, not only the texture features of B-ultrasound images are extracted, but also the temporal and spatial features are further extracted. The texture features can reflect the information of liver tissue at the microscopic level, while the temporal features can reflect the development and change trend of fibrosis over time, and the spatial features show the distribution, area and shape of fibrotic lesions. By fusing and analyzing these multi-dimensional features, the actual situation of liver fibrosis caused by schistosomiasis can be grasped more comprehensively and accurately, thereby significantly improving the accuracy of fibrosis grading and reducing the probability of misdiagnosis and missed diagnosis.

[0016] 2. Accurate suspicious area positioning: When matching the texture feature set, the suspicious area in the grayscale image is accurately located by comparing the similarity with the texture feature set of the preset normal area. The feature matching method is more scientific and accurate than traditional naked eye observation or simple threshold judgment, and can timely detect early and tiny fibrosis lesions, providing a more reliable basis for subsequent diagnosis and treatment.

[0017] 3. Fibrosis degree labeling assisted by deep learning: Use deep learning to build a fibrosis recognition model, and use a large number of labeled training images containing different fibrosis lesions for training. The deep learning model has powerful feature learning and classification capabilities, and can automatically learn the characteristic patterns of different fibrosis degrees from images, so as to more accurately label the fibrosis degree of suspicious areas in grayscale images, avoiding the subjectivity and errors that may exist in manual labeling.

[0018] 4. Intelligent treatment measure retrieval: Based on the subject's fibrosis prediction information, the dynamic analysis of the fibrosis degree time change curve and the fibrosis lesion area change curve can quickly obtain the subject's fibrosis degree change range and lesion area change range. Then, data retrieval is performed in the preset lesion treatment database to automatically screen out treatment measures that meet the preset standards, providing doctors with the best treatment plan recommendations, avoiding the process of doctors manually reviewing a large amount of information and cases, and improving the efficiency of treatment decision-making.

[0019] 5. Model training based on big data: When building the fibrosis recognition model and the fibrosis prediction model, a large amount of historical data was used for training. These data include B-ultrasound images of different patients and corresponding fibrosis degree annotations, time characteristics, spatial characteristics and other information. Through big data learning, the model can learn a wider range of feature patterns and laws, improve the generalization ability and prediction accuracy of the model, and provide more reliable support for doctors' diagnosis and treatment decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a schematic diagram of a method for dynamically analyzing fibrosis grading data of schistosomiasis B-ultrasound images according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] The following is a clear and complete description of the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0022] like Figure 1 As shown, a method for dynamic analysis of schistosomiasis B-ultrasound image fibrosis classification data comprises the following steps: Step s1: grayscale processing and quality screening operations are performed on the original B-ultrasound image of the subject, and then texture feature extraction is performed on the grayscale image to obtain a texture feature set; Step s2: performing feature matching on the texture feature set, obtaining suspicious areas of the grayscale image, marking the degree of fibrosis on the suspicious areas of the grayscale image, and performing temporal feature extraction and spatial feature extraction on the grayscale images marked with the degree of fibrosis corresponding to all original B-ultrasound images collected from the subject; Step s3: construct a fibrosis prediction model, obtain the fibrosis prediction information of the subject according to the spatial and temporal characteristics of the fibrosis prediction model and the grayscale image annotated with the degree of fibrosis, retrieve treatment measures according to the fibrosis prediction information of the subject, and obtain corresponding treatment measures.

[0023] It should be further explained that, in the specific implementation process, the process of grayscale processing and quality screening operation on the original B-ultrasound image of the subject includes: Collect original B-ultrasound images of multiple basic liver sections of the examinee, and obtain at least six basic liver sections for each examinee, which may include the subxiphoid sagittal section, the left subcostal transverse section, the right subcostal section, the right intercostal oblique section, etc. To improve the image diversity, multiple (such as 3) B-ultrasound images at different angles need to be taken for each section, and fine-tuned within a range of no more than 15° to the left and right of the long axis direction of the probe. After the acquisition is completed, it should be ensured that each examinee has at least 18 clear and usable B-ultrasound images. In special circumstances (such as abnormal liver morphology of the patient or too limited lesion area), other sections can be collected as appropriate, and the original B-ultrasound images are grayed to generate gray images. The gray values ​​of each pixel in the gray image are statistically analyzed to obtain a gray image histogram, in which the abscissa of the gray image histogram is the gray value and the ordinate is the frequency of occurrence of the gray value. Feature extraction is performed on the gray image histogram to obtain the gray mean and gray value distribution range; Preset the threshold interval corresponding to the grayscale mean and grayscale value distribution range, compare the grayscale mean and grayscale value distribution range corresponding to each grayscale image with the corresponding threshold interval, and eliminate the grayscale images whose corresponding grayscale mean and grayscale value distribution range are not within the corresponding threshold interval.

[0024] It should be further explained that, in the specific implementation process, the process of extracting texture features from the grayscale image and obtaining the texture feature set includes: The grayscale value of each pixel in the grayscale image is quantized into discrete grayscale levels, and the fixed parameters of the grayscale co-occurrence matrix are preset. The grayscale co-occurrence matrix of the local area of ​​each pixel in the grayscale image is obtained by a sliding window, and the texture feature set of the local area of ​​each pixel in the grayscale image is obtained according to the grayscale co-occurrence matrix. The texture feature set includes contrast, energy, entropy, and inverse variance.

[0025] It should be further explained that, in the specific implementation process, the process of performing feature matching on the texture feature set and obtaining the suspicious area of ​​the grayscale image includes: A texture feature set corresponding to a normal area is preset, and the texture feature set of the local area of ​​each pixel in the grayscale image is compared with the texture feature set corresponding to the normal area to obtain the texture feature similarity between the local area of ​​each pixel and the normal area, and the texture feature similarity between the local area of ​​each pixel and the normal area is compared with a preset texture feature similarity threshold. If the texture feature similarity between the local area of ​​the pixel and the normal area is less than the texture feature similarity threshold, the local area of ​​the pixel is marked as a suspicious area.

[0026] The calculation formula for obtaining the texture feature similarity between the local area of ​​each pixel and the normal area is: ; in, represents the texture feature similarity between the local area of ​​the i-th pixel and the normal area j, represents the contrast between the local area of ​​the i-th pixel and the normal area j, Represents the energy of the local area of ​​the i-th pixel, represents the energy of normal region j, represents the entropy of the local area of ​​the i-th pixel, represents the entropy of normal region j, represents the inverse variance of the local area of ​​the i-th pixel, represents the inverse variance of the normal region j, , , and Represents the weight factor.

[0027] It should be further explained that, in the specific implementation process, the process of marking the degree of fibrosis in the suspicious area of ​​the grayscale image includes: A fibrosis recognition model is constructed based on deep learning, and a number of training images containing different fibrosis lesion areas are collected in advance, and the fibrosis degree of the training images containing different fibrosis lesion areas is annotated, and the judgment criteria of the fibrosis degree include: Level 0: uniform liver echo and clear blood vessel course; Level I: fish scale-like echoes are visible and the blood vessels are still clear; Level II: grid-like echoes and unclear blood vessel course; Level III: coarse grids or large light spots; the training images after fibrosis annotation are used as training data, and the fibrosis recognition model is trained using the training data to obtain a trained fibrosis recognition model; The grayscale image with the marked suspicious area is input into the fibrosis recognition model, and the fibrosis degree of the suspicious area in the grayscale image is marked according to the fibrosis recognition model.

[0028] It should be further explained that, in the specific implementation process, the process of extracting time features and space features from the grayscale images marked with the degree of fibrosis corresponding to all the original B-ultrasound images collected from the subject includes: The grayscale images with the degree of fibrosis marked corresponding to all the original B-ultrasound images collected from the subject are extracted, the acquisition time of the grayscale images is extracted, a grayscale image time series is generated according to the acquisition time of the grayscale images, a time convolutional neural network is constructed, the grayscale image time series is input into the time convolutional neural network, and the fibrosis degree change trend characteristics of the suspicious area are obtained according to the time convolutional neural network. At the same time, a spatial convolutional neural network is constructed to extract spatial features of the grayscale images with the degree of fibrosis marked, and obtain the spatial features of the grayscale images with the degree of fibrosis marked, wherein the spatial features include the coverage area and shape of the suspicious area.

[0029] It should be further explained that, in the specific implementation process, the process of constructing a fibrosis prediction model and obtaining the fibrosis prediction information of the subject according to the spatial characteristics and temporal characteristics of the fibrosis prediction model and the grayscale image annotated with the degree of fibrosis includes: A fibrosis prediction model is constructed based on an RBF neural network, and a historical grayscale image time series of several subjects is collected in advance, and the time features and spatial features of the historical grayscale image time series of several subjects are extracted, and the time features and spatial features are used as a training set and a test set, and the training set is input into the fibrosis prediction model for training until the loss function training is stable, and the model parameters are saved, and the fibrosis prediction model is tested by the test set until it meets the preset requirements, and the fibrosis prediction model is output; The time characteristics, spatial characteristics and fibrosis degree corresponding to the grayscale image of the subject with the fibrosis degree marked are input into the fibrosis prediction model, and the fibrosis prediction information of the subject is output according to the fibrosis prediction model.

[0030] Building a deep learning-based fibrosis prediction model is a complex process that involves multiple steps such as model selection, training, validation, and testing. The following is a detailed supplementary description of this process: The present invention selects a convolutional neural network (CNN) suitable for time series analysis as a deep learning architecture. After determining the model architecture, the mean square error loss function is selected as the optimization target, and then the prepared training set is input into the selected deep learning model to start training. During the training process, the weights are continuously updated through the back propagation algorithm, so that the loss function gradually decreases until a stable state is reached. During this period, techniques such as Early Stopping are used to avoid overfitting. In addition to the basic training process, various parameters of the model are tuned through Grid Search, including learning rate, batch size, regularization coefficient, etc.

[0031] After the model training is completed and the parameters are adjusted, the final evaluation is performed through the test set to obtain the evaluation results of the model, which include classification indicators such as accuracy, recall, F1 score, etc. Based on the evaluation results on the test set, it is judged whether the model has reached the expected standards. If the requirements are met, the model parameters are saved and ready for deployment; if not ideal, it is necessary to return to a previous stage to re-examine issues such as data quality, model structure or training strategy.

[0032] It should be further explained that, in the specific implementation process, the treatment measures are retrieved according to the fibrosis prediction information of the subject, and the process of obtaining the corresponding treatment measures includes: According to the fibrosis prediction information of the subject, the time change curve of the fibrosis degree and the regional change curve of the fibrosis lesion of the subject are obtained, and the time change curve of the fibrosis degree and the regional change curve of the fibrosis lesion of the subject are dynamically analyzed to obtain the change range of the fibrosis degree and the change range of the fibrosis lesion of the subject; A lesion treatment database is preset, and the lesion treatment database includes treatment measures corresponding to different ranges of change in fibrosis degrees and ranges of change in fibrosis lesion areas. Data is searched in the lesion treatment database according to the ranges of change in the fibrosis degree and ranges of change in fibrosis lesion areas of the subject, and the ranges of change in the fibrosis degree and ranges of change in fibrosis lesion areas corresponding to each treatment measure in the lesion database are obtained, and the similarity with the ranges of change in the fibrosis degree and ranges of change in fibrosis lesion areas of the subject is obtained, and treatment measures whose similarity meets the preset standards are obtained in the lesion database, and the treatment measures are marked as corresponding treatment measures.

[0033] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for dynamic analysis of schistosomiasis B-ultrasound image fibrosis classification data, characterized in that: The following steps are involved: Step s1: grayscale processing and quality screening operations are performed on the original B-ultrasound image of the subject, and then texture feature extraction is performed on the grayscale image to obtain a texture feature set; Step s2: performing feature matching on the texture feature set, obtaining suspicious areas of the grayscale image, marking the degree of fibrosis on the suspicious areas of the grayscale image, and performing temporal feature extraction and spatial feature extraction on the grayscale images marked with the degree of fibrosis corresponding to all original B-ultrasound images collected from the subject; Step s3: construct a fibrosis prediction model, obtain the fibrosis prediction information of the subject according to the spatial and temporal characteristics of the fibrosis prediction model and the grayscale image annotated with the degree of fibrosis, retrieve treatment measures according to the fibrosis prediction information of the subject, and obtain corresponding treatment measures.

2. The method for dynamic analysis of schistosomiasis B-ultrasound image fibrosis classification data according to claim 1, characterized in that: The process of graying the original B-ultrasound image of the subject and performing quality screening operations includes: Grayscale the original B-ultrasound image to generate a grayscale image, perform statistical analysis on the grayscale value of each pixel in the grayscale image to obtain the grayscale image histogram, perform feature extraction on the grayscale image histogram, and obtain the grayscale mean and grayscale value distribution range; A threshold interval corresponding to the grayscale mean and the grayscale value distribution range is preset, and grayscale images whose corresponding grayscale mean and grayscale value distribution range are not within the corresponding threshold interval are eliminated.

3. The method for dynamic analysis of schistosomiasis B-ultrasound image fibrosis classification data according to claim 2, characterized in that: The process of extracting texture features from grayscale images and obtaining texture feature sets includes: The grayscale value of each pixel in the grayscale image is quantized into discrete grayscale levels, and the fixed parameters of the grayscale co-occurrence matrix are preset. The grayscale co-occurrence matrix of the local area of ​​each pixel in the grayscale image is obtained by a sliding window, and the texture feature set of the local area of ​​each pixel in the grayscale image is obtained according to the grayscale co-occurrence matrix.

4. The method for dynamic analysis of schistosomiasis B-ultrasound image fibrosis classification data according to claim 3, characterized in that: The process of performing feature matching on the texture feature set and obtaining the suspicious area of ​​the grayscale image includes: A texture feature set corresponding to the normal area is preset, and the texture feature set of the local area of ​​each pixel in the grayscale image is compared with the texture feature set corresponding to the normal area to obtain the texture feature similarity between the local area of ​​each pixel and the normal area. If the texture feature similarity between the local area of ​​the pixel and the normal area is less than the preset texture feature similarity threshold, the local area of ​​the pixel is marked as a suspicious area.

5. The method for dynamic analysis of schistosomiasis B-ultrasound image fibrosis classification data according to claim 4, characterized in that: The process of marking the degree of fibrosis in suspicious areas of grayscale images includes: Constructing a fibrosis recognition model based on deep learning, pre-collecting a number of training images containing different fibrosis lesion areas, and annotating the fibrosis degree of the training images containing different fibrosis lesion areas, using the training images after fibrosis annotation as training data, and using the training data to train the fibrosis recognition model to obtain a trained fibrosis recognition model; The grayscale image with the marked suspicious area is input into the fibrosis recognition model, and the fibrosis degree of the suspicious area in the grayscale image is marked according to the fibrosis recognition model.

6. A method for dynamic analysis of schistosomiasis B-ultrasound image fibrosis classification data according to claim 5, characterized in that: The process of temporal feature extraction and spatial feature extraction for grayscale images includes: The grayscale images with the completed fibrosis degree marking corresponding to all the original B-ultrasound images collected from the subject are extracted, the acquisition time of the grayscale images is extracted, a grayscale image time series is generated according to the acquisition time of the grayscale images, a temporal convolutional neural network is constructed, the grayscale image time series is input into the temporal convolutional neural network, and the fibrosis degree change trend characteristics of the suspicious area are obtained according to the temporal convolutional neural network. At the same time, a spatial convolutional neural network is constructed to extract spatial features of the grayscale images with the completed fibrosis degree marking, and obtain the spatial features of the grayscale images with the completed fibrosis degree marking.

7. A method for dynamic analysis of schistosomiasis B-ultrasound image fibrosis classification data according to claim 6, characterized in that: The process of obtaining the subject's predicted fibrosis information includes: Building a fibrosis prediction model based on an RBF neural network, pre-collecting a number of historical grayscale image time series of subjects, extracting time features and spatial features of the historical grayscale image time series of the subjects, using the time features and spatial features as training data, and outputting a trained fibrosis prediction model; The time characteristics, spatial characteristics and fibrosis degree corresponding to the grayscale image of the subject with the fibrosis degree marked are input into the fibrosis prediction model, and the fibrosis prediction information of the subject is output according to the fibrosis prediction model.

8. The method for dynamic analysis of schistosomiasis B-ultrasound image fibrosis classification data according to claim 7, characterized in that: The process of retrieving treatment measures based on the subject's predicted fibrosis information and obtaining the corresponding treatment measures includes: According to the fibrosis prediction information of the subject, the time change curve of the fibrosis degree and the regional change curve of the fibrosis lesion of the subject are obtained, and the time change curve of the fibrosis degree and the regional change curve of the fibrosis lesion of the subject are dynamically analyzed to obtain the change range of the fibrosis degree and the change range of the fibrosis lesion of the subject; A lesion treatment database is preset, and the lesion treatment database includes treatment measures corresponding to different ranges of change in fibrosis degrees and ranges of change in fibrosis lesion areas. Data is retrieved in the lesion treatment database according to the range of change in the fibrosis degree and range of change in the fibrosis lesion area of ​​the subject, and treatment measures whose similarity meets the preset standards are obtained in the lesion database, and the treatment measures are marked as corresponding treatment measures.

Images