Ultrasonic imaging medical record learning system

Abstract

The invention provides an ultrasonic imaging medical record learning system comprising a registration module used for registering a patient user, a doctor user and a student user; the medical record uploading module is used for uploading medical records to the database; the case analysis module is used for analyzing cases generated by the database according to the medical records by the doctor user and generating a case analysis report; the case learning module is used for learning the case and the case analysis report by the student user; the simulation exercise module is used for generating case simulation exercises according to the case and the case analysis report; and the answer statistics module is used for recording the answer performance of the case simulation exercises analyzed by the student user. In the invention, the student user can refer to the medical record of the patient user, the case generated by the related medical record of the learning database and the case analysis report made by the doctor user for the patient user, and the efficiency of medical record learning analysis is obviously improved.

Description

Technical Field

[0001] The present invention relates to the field of medical artificial intelligence technology, and in particular to an ultrasonic imaging medical case record learning system. Background Art

[0002] At present, the existing medical imaging systems mostly use manual analysis and lack systematic case analysis and learning and training functions. Doctors often rely on static images and manual records during remote consultations, which affects communication efficiency. For example, US Patent US201715720143 provides a virtual case simulation practice function, but it relies on a static case library and cannot dynamically integrate real-time patient data, causing the teaching content to lag behind clinical practice. At the same time, medical students and young doctors lack high-quality training resources and interactive learning opportunities, making it difficult to improve their diagnostic capabilities. The training modules in some existing systems are limited to the explanation of theoretical knowledge and are often unable to be combined with real cases, resulting in poor learning results.

[0003] In summary, how to improve the efficiency of medical record learning and analysis is a technical problem that technical personnel in this field currently need to solve. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide an ultrasound imaging medical record learning system, which can significantly improve the efficiency of medical record learning and analysis. The specific scheme is as follows: An ultrasonic imaging medical record learning system, characterized by comprising a registration module, a medical record upload module, a case analysis module, a case learning module, a simulation exercise module, a question answering statistics module and a database, wherein: The registration module is used to register patient users, doctor users and student users; The medical record uploading module is used to upload the medical records to the database; The case analysis module is used for the doctor user to analyze the cases generated by the database according to the medical records and generate a case analysis report; The case study module is used for the student user to study the case and the case analysis report; The simulation exercise module is used to generate case simulation exercises based on the medical records and the case analysis report; The answer statistics module is used to record the answer performance of the student user in analyzing the case simulation practice questions.

[0005] Preferably, the registration module includes a registration unit and an identity verification unit.

[0006] Preferably, the medical record uploading module further includes an ultrasound image data collection unit, an ultrasound image data preprocessing unit, and an ultrasound image preclassification unit, wherein: The ultrasonic image data collection unit is used to collect the ultrasonic image information to be processed from the ultrasonic image shooting device, and the ultrasonic image information to be processed includes ultrasonic images, device parameters, and imaging modes; The ultrasonic image data preprocessing unit is used to perform graphic processing on the ultrasonic image information to be processed according to a preset image algorithm to obtain preprocessed ultrasonic image information, wherein the preset image algorithm includes performing privacy information removal, size adjustment, grayscale distribution, contrast adjustment, noise elimination, and image edge detection on the ultrasonic image information to be processed; The ultrasound image pre-classification unit is used to classify the pre-processed ultrasound image information according to a preset classification algorithm, wherein the preset classification algorithm includes allocating the pre-processed ultrasound image information with similar characteristics to the same training domain according to the device type, imaging parameters, anatomical site and quality level of the pre-processed ultrasound image information.

[0007] Preferably, the preset image algorithm includes the grayscale distribution adjustment, the contrast adjustment, the noise elimination and the image edge detection, wherein: The grayscale distribution adjustment and the contrast adjustment are implemented by an adaptive grayscale normalization algorithm, specifically a combination of dynamic range compression and contrast enhancement, and the calculation formula is as follows: ; Among them, the , are the local mean and standard deviation of the image in the ultrasonic image information to be processed, is the contrast gain factor, is the brightness offset, The image information in the ultrasonic image information to be processed is located at position The pixel value of The image of the ultrasonic image information to be processed after being processed by the adaptive grayscale normalization algorithm is at position The pixel value of The noise elimination adopts an anisotropic denoising model, combined with non-local mean and gradient constraints, and the calculation formula is as follows: ; in, When is the total variation model, When it is an isotropic smoothing model, is the image information in the ultrasonic image information to be processed, for The gradient information of The ultrasonic image information to be processed is The image information at the point The ultrasonic image information to be processed is The image information at the point, and To adjust the coefficients of the data fidelity term and the regularization term weights, w(p,q) is a weight function calculated based on pixel block similarity. The noise image information in the image information in the ultrasonic image information to be processed, The denoised image information is the denoised image information after the ultrasonic image information to be processed is processed by the anisotropic denoising model; The image edge detection is a multi-scale edge detection, which integrates the Canny operator and the wavelet transform. The calculation formula is as follows: ; Among them, the For the current scale s The Gaussian kernel of is the threshold function, The current scale s The weight coefficient is The current scale s The gradient amplitude under The current scale s The corresponding gradient threshold, The image information in the ultrasonic image information to be processed is located at position The pixel value of The image information in the ultrasonic image information to be processed is converted by the multi-scale edge detection at the position The pixel value of .

[0008] Preferably, the preset classification algorithm includes attention-enhanced convolution, multi-task classification objective function loss, and feature domain similarity measurement, wherein: The attention-enhanced convolution embeds the channel attention mechanism (SE-block) in CNN: ; As stated, For global pooling, and is the weight of the fully connected layer, The preprocessed ultrasound image information is The Fourier convolution of the image pixel value at the position, is the exclusive OR operator, is the output feature of the preprocessed ultrasound image information; The classification tasks in the multi-task classification objective function are established according to one or more of the device type, the imaging parameters, the anatomical part and the quality level, and the multi-task classification objective function is as follows: ; When K=n corresponds to n classification tasks, n is the number of classification tasks of the classification task and n is greater than or equal to 1, the is the task weight coefficient, is the true label of category c in the kth task, In the kth task, the probability of category c is is the number of categories of the kth task, All trainable parameters in the model; The feature domain similarity metric is based on the similarity metric of the domain assignment criterion of Wasserstein distance, and the calculation formula is as follows: ; Said is a feature extractor used to measure the distribution differences of different image domains. and stated are different feature domains corresponding to the preprocessed ultrasound image information, is the marginal distribution, for Based on edge distribution Strategy.

[0009] Preferably, the case analysis module further includes a cyclic adversarial learning unit and a resolution enhancement unit, wherein: The cyclic adversarial learning unit is used to perform domain conversion training on the pre-processed ultrasound image information, and improve the clarity and detail expression of the pre-processed ultrasound image information through generator and discriminator adversarial learning; The resolution enhancement unit is used to reconstruct the resolution of the pre-processed ultrasound image information, improve the spatial resolution and detail contrast of the pre-processed ultrasound image information, and generate a medical record of high-resolution ultrasound image information for the doctor user.

[0010] Preferably, the ultrasound imaging medical case record learning system further includes a consultation module, which is used for the patient user to consult with the doctor user to generate the medical record.

[0011] Preferably, the ultrasound imaging medical record learning system further includes an online registration module for the patient user to register with the doctor user.

[0012] Preferably, the ultrasound imaging medical case record learning system also includes a live broadcast module for the doctor user to explain the case to the student user.

[0013] Preferably, the ultrasound imaging medical record learning system also includes a payment module.

[0014] The present invention provides a medical case study learning system based on ultrasound image data preprocessing, multi-task classification and adversarial learning, which is used to improve the analysis ability of medical students on clinical cases. Student users can refer to the medical records of patient users, learn the cases generated by the relevant medical records in the database and the case analysis reports made by doctor users for the patient users, which significantly improves the efficiency of medical case study analysis. Among them, through the case analysis module and the case learning module, student users can directly study the medical records of real patients and the case analysis reports generated by doctors, and combine them with dynamically updated case data to achieve a deep integration of theory and practice, significantly shortening the clinical experience accumulation cycle of medical students. The simulation exercise module generates exercises based on real cases, and records learning performance through the answer statistics module, providing instant feedback to help students improve their diagnostic capabilities in a targeted manner; Adaptive grayscale normalization algorithm and dynamic normalization strategy can align the grayscale distribution of images from different devices, solving the problem of inconsistent image quality caused by device differences. The anisotropic denoising model combines non-local mean and gradient constraints to effectively suppress noise while retaining the edge structure of ultrasound images, thereby improving image clarity. Multi-scale edge detection combines Canny operator and wavelet transform to enhance image detail features, optimize ultrasound image processing quality, and assist doctors in more accurate analysis of lesions. The multi-task classification algorithm automatically classifies ultrasound images in multiple dimensions, such as device type and anatomical part, to reduce manual labeling costs; the recurrent adversarial learning unit solves the problem of cross-device and cross-part data enhancement through domain transfer technology, thus enhancing data classification and training efficiency; Therefore, this system solves the problems of low efficiency in medical image analysis, lack of teaching resources, and uneven data quality through innovative algorithms and modular design, and realizes the integrated collaboration of medical record learning, image processing, and teaching interaction. It has significant clinical application value and teaching promotion potential. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the implementation scheme of the present invention or the technical scheme in the prior art, the drawings required for use in the implementation scheme or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only the implementation scheme of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0016] Figure 1A schematic diagram of the structure of an ultrasound imaging medical record learning system provided by an embodiment of the present invention; Figure 2 A schematic diagram of a registration verification process of a registration module in a specific ultrasound imaging medical record learning system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] The embodiment of the present invention provides an ultrasound imaging medical record learning system, such as Figure 1 As shown, it includes: a registration module 1, a medical record uploading module 2, a case analysis module 3, a case learning module 4, a simulation exercise module 5, a question answering statistics module 6 and a database 7.

[0019] Registration module, used to register patient users, doctor users and student users.

[0020] The registration module includes a registration unit and an identity verification unit.

[0021] It is understandable that all systems will have administrators, who have all permissions. Figure 2 As shown, this system supports PC and mobile terminals. The registration unit is registered through a mobile phone number. Of course, it can also be registered through other methods such as email. In my country, since mobile phone numbers have real-name authentication features, this implementation plan is explained using mobile phone number registration.

[0022] The authentication unit is managed by the administrator. Figure 2 In the , the administrator sends a verification code through the mobile phone number to complete the administrator verification. After entering the administrator interface, the mobile phone number is granted the doctor identity. Of course, the mobile phone number can also be granted the student identity and patient identity, which will not be repeated here. After the user registers or logs in with the mobile phone number, he is determined to be a doctor or other identity according to the identity granted by the administrator.

[0023] When using the case study module, the identity granted to ordinary users is student users, so Figure 2 The identity of a middle school student is illustrated with an ordinary mobile phone number.

[0024] Medical record upload module, used to upload medical records to the database.

[0025] The medical record upload module includes automatic uploading of medical records generated by patient users using the consultation and diagnosis module, and also includes manual uploading of corresponding medical records by patient users, doctor users or student users.

[0026] The uploaded content includes medical imaging files such as B-ultrasound photos, CT photos, and medical treatment videos, as well as text descriptions of symptoms, signs, and their occurrence time, nature or degree, and location, as well as temporary files of patient users, such as registration records, consultation records, and payment records. The database aggregates the medical records uploaded by the same patient user.

[0027] In order to further improve the user effect of the implementation scheme of the present invention, the medical record uploading module also includes an ultrasound image data collection unit, an ultrasound image data preprocessing unit, and an ultrasound image preclassification unit, wherein: The ultrasound image data collection unit is used to collect the ultrasound image information to be processed from the ultrasound image shooting equipment. The ultrasound image information to be processed includes ultrasound images, equipment parameters, and imaging modes. The ultrasound image data collection unit collects image data from ultrasound image shooting equipment (including ultrasound equipment of different brands and models and non-ultrasound equipment such as mobile terminals) to ensure the diversity and breadth of the data, and builds high-quality training and test data sets through data format conversion and standardized storage, providing a stable data foundation for subsequent enhancements.

[0028] It can be understood that the ultrasound image data collection unit supports collecting the ultrasound image information to be processed from one ultrasound image shooting device, and also supports collecting the ultrasound image information to be processed from multiple ultrasound image shooting devices at the same time.

[0029] The ultrasound image data preprocessing unit is used to perform graphic processing on the ultrasound image information to be processed according to a preset image algorithm to obtain preprocessed ultrasound image information. The preset image algorithm includes removing privacy information (such as patient information, hospitalization information), resizing, grayscale distribution, contrast adjustment, noise elimination, and image edge detection for the ultrasound image information to be processed; the ultrasound image data preprocessing unit performs normalization processing on the image, including operations such as unified size, grayscale distribution, contrast adjustment, and denoising to improve the basic image quality. At the same time, it adopts edge detection and structure enhancement algorithms to optimize image details and provide high-quality ultrasound image input for doctor user labeling and intelligent labeling.

[0030] The preset image algorithm includes the grayscale distribution adjustment, the contrast adjustment, the noise elimination and the image edge detection, wherein: The above grayscale distribution adjustment and the above contrast adjustment are implemented through an adaptive grayscale normalization algorithm, which specifically adopts a combination of dynamic range compression and contrast enhancement. The calculation formula is as follows: ; Among them, the above , is the local mean and standard deviation of the image in the above-mentioned ultrasonic image information to be processed. In practical applications, the local mean of the image can be adjusted according to the size of the ultrasonic image information to be processed. For example, the range of the local mean of the image includes part to all of the image information in the ultrasonic image information to be processed. is the contrast gain factor, the above is the brightness offset, the above The image information in the above-mentioned ultrasonic image information to be processed is at position The pixel value of The image of the above-mentioned ultrasonic image information to be processed after being processed by the above-mentioned adaptive grayscale normalization algorithm is at position The pixel value of The present invention also introduces a dynamic normalization strategy when processing grayscale distribution adjustment, specifically grayscale correction based on histogram matching: ; Represents the cumulative distribution function, which realizes the grayscale distribution alignment of images from different devices. The ultrasound image information to be processed is The image pixel value at the location, The image to be processed is the image corrected by the dynamic normalization strategy. The image pixel value at the location.

[0031] The above noise elimination adopts an anisotropic denoising model, combined with non-local mean and gradient constraints, and the calculation formula is as follows: ; When p = 1, the model is suitable for preserving the edge structure of ultrasound images; when p = 2, the model is suitable for isotropic smoothing to suppress speckle noise. is the image information in the ultrasound image information to be processed, for The gradient information of The above-mentioned ultrasound image information to be processed is The image information at the point The above-mentioned ultrasound image information to be processed is The image information at the point, and To adjust the coefficients of the data fidelity term and the regularization term weights, w(p,q) is the weight function calculated based on the pixel block similarity. is the noise image information in the image information in the above-mentioned ultrasonic image information to be processed, The denoised image information is the denoised image information after the above-mentioned ultrasonic image information to be processed is processed by the above-mentioned anisotropic denoising model; The above image edge detection is a multi-scale edge detection, which combines the Canny operator and wavelet transform. The calculation formula is as follows: ; Among them, the above For the current scale s The Gaussian kernel of is the threshold function, the above For the current scale s The weight coefficient is For the current scale s The gradient amplitude under For the current scale s The corresponding gradient threshold, mentioned above The image information in the above-mentioned ultrasonic image information to be processed is at position The pixel value of The image information in the above-mentioned ultrasonic image information to be processed is converted by the above-mentioned multi-scale edge detection at the position The pixel value of .

[0032] In order to improve the application effect of image edge detection, the embodiment disclosed in the present invention also includes introducing a structure enhancement operator to calculate edge-preserving morphological enhancement: ; is the Laplace kernel, is the edge mask, To control the strength, To input the original image, Enhances the image for output.

[0033] The ultrasound image pre-classification unit is used to classify the pre-processed ultrasound image information according to a preset classification algorithm. The preset classification algorithm includes allocating the pre-processed ultrasound image information with similar characteristics to the same training domain according to the device type, imaging parameters, anatomical site and quality level of the pre-processed ultrasound image information.

[0034] The ultrasound image pre-classification unit automatically classifies ultrasound images and uses the CNN+attention mechanism to identify the device type, imaging parameters, anatomical location, and quality level of the image. Through image classification, images with similar features can be assigned to the same training domain, ensuring that subsequent adversarial learning is more accurate and avoiding enhancement distortion problems caused by different imaging conditions.

[0035] The above preset classification algorithms include attention-enhanced convolution, multi-task classification objective function loss, and feature domain similarity measurement, among which, The above attention-enhanced convolution embeds the channel attention mechanism (SE-block) in CNN: ; Above, For global pooling, the above and the above is the weight of the fully connected layer, To pre-process the ultrasound image information Fourier convolution of the image pixel value at the position, above is the same or equal operator, the above is the output feature of the above preprocessed ultrasound image information; The classification tasks in the multi-task classification objective function are established according to one or more of the device type, the imaging parameters, the anatomical part and the quality level. The multi-task classification objective function is as follows: ; When K=n corresponds to n classification tasks, the above n is the number of classification tasks of the above classification task and n is greater than or equal to 1, the above is the task weight coefficient, the above is the true label of category c in the kth task. In the kth task, the probability of category c, as mentioned above is the number of categories of the kth task, the above All trainable parameters in the model; The above feature domain similarity measure is based on the similarity measure of the domain assignment criterion of Wasserstein distance, and the calculation formula is as follows: ; Above is a feature extractor used to measure the distribution differences of different image domains. and the above The above-mentioned preprocessed ultrasound image information corresponds to different feature domains. is the marginal distribution, the above and Based on edge distribution Strategy.

[0036] The case analysis module is used by doctor users to analyze cases generated by the database based on medical records and generate case analysis reports.

[0037] The database aggregates the medical records uploaded by the same patient user to generate a case. The doctor user analyzes the case and generates a case analysis report based on the patient user's current medical history and past medical history in the database. Different doctor users can make case analysis reports for the same patient and upload them to the database.

[0038] The case analysis module also includes a cyclic adversarial learning unit and a resolution enhancement unit, wherein: The cyclic adversarial learning unit is used to perform domain conversion training on pre-processed ultrasound image information. Through adversarial learning between the generator and the discriminator, the clarity and detail of the pre-processed ultrasound image information are improved. The cyclic adversarial learning unit uses the classification-based CycleGAN domain transfer technology to perform independent domain conversion training for different categories of ultrasound images to solve the problem that traditional CycleGAN has no clear target domain when enhancing data on different imaging devices and different parts. Through the adversarial learning mechanism of the generator and the discriminator, unpaired enhancement of high- and low-quality images is achieved, significantly improving the clarity and detail of ultrasound images.

[0039] The resolution enhancement unit is used to reconstruct the resolution of the pre-processed ultrasound image information, improve the spatial resolution and detail contrast of the pre-processed ultrasound image information, and generate medical records of high-resolution ultrasound image information for doctor users. The resolution enhancement unit further uses super-resolution reconstruction algorithms (such as ESRGAN and SwinIR) to improve the spatial resolution and detail contrast of the image, providing better data input for subsequent annotation and intelligent annotation for doctor users.

[0040] It can be understood that this embodiment supports doctor user labeling and system intelligent labeling, that is, it supports doctor users to analyze cases generated by the database based on medical records of high-resolution ultrasound imaging information, thereby generating case analysis reports, and also supports intelligent AI analysis of cases generated by the database based on medical records of high-resolution ultrasound imaging information, thereby generating case analysis reports.

[0041] The case study module is used for student users to study cases and case analysis reports.

[0042] Student users can refer to the medical records of patient users, study the cases generated by the relevant medical records in the database, and the case analysis reports made by doctor users for the patient users. If during the study period, the patient user uses the consultation module to conduct a new consultation, the newly uploaded medical records and the updated cases of the newly uploaded medical records in the database will be displayed according to the time nodes for comparison with the current study.

[0043] When a student user studies a patient user's case through the case study module, the corresponding uploaded medical records, cases generated by the database, and case analysis reports made by the doctor user are displayed in sequence according to the time nodes when the patient user uploaded different medical records.

[0044] When student users learn a specific case through the case learning module, the medical records uploaded by different patient users, the cases generated by the database, and the case analysis reports made by the doctor users are displayed separately. For each patient user, the corresponding uploaded medical records, the cases generated by the database, and the case analysis reports made by the doctor users are displayed in sequence based on the time node when the patient user uploaded different medical records.

[0045] The simulation exercise module is used to generate case simulation exercises based on medical records and case analysis reports, and the case simulation exercises are analyzed by student users.

[0046] After the student user has learned using the case study module, the simulation practice module generates case simulation exercises for the student to learn based on the medical records and case analysis reports. Of course, in actual applications, the simulation practice module can be assigned by the doctor user, or the simulation practice module can generate case simulation exercises using the database through big data technology. Of course, the doctor user or the student user who did not participate in the simulation exercise can watch the analysis process of the case simulation exercises by the student user who participated in the simulation exercise.

[0047] The answer statistics module is used to record the student user's answer performance in analyzing case simulation exercises. The answer performance and the doctor user's annotation score for the student user will be uploaded to the database.

[0048] In order to improve the use effect of student users, a specific implementation scheme disclosed by the present invention also includes a live broadcast module, in which a doctor user explains the case to the student user. Of course, when the patient user consults the doctor user through the consultation module, the live broadcast can be broadcast to the student user. The live broadcast must consider the privacy of the patient and ask for the patient's consent. The live broadcast module also includes a privacy processing unit, which is used to eliminate the patient's personal information that is not necessary for teaching during the live broadcast.

[0049] In specific implementations, registration, consultation, learning, watching live broadcasts and interaction all involve the steps of placing orders and paying fees, so the implementation scheme disclosed in the present invention also includes a payment module.

[0050] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0051] The above is a detailed introduction to an ultrasound imaging medical record learning system provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above implementation plan is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. An ultrasound imaging medical record learning system, characterized in that: It includes registration module, medical record upload module, case analysis module, case study module, simulation exercise module, answer statistics module and database, among which, The registration module is used to register patient users, doctor users and student users, and includes a registration unit and an identity verification unit; The medical record uploading module is used to upload the medical records to the database; The case analysis module is used for the doctor user to analyze the cases generated by the database according to the medical records and generate a case analysis report; The case study module is used for the student user to study the case and the case analysis report; The simulation exercise module is used to generate case simulation exercises based on the medical records and the case analysis report; The answer statistics module is used to record the answer performance of the student user in analyzing the case simulation exercise questions; The medical record uploading module also includes an ultrasound image data collection unit, an ultrasound image data preprocessing unit, and an ultrasound image preclassification unit, wherein: The ultrasonic image data collection unit is used to collect the ultrasonic image information to be processed from the ultrasonic image shooting device, and the ultrasonic image information to be processed includes ultrasonic images, device parameters, and imaging modes; The ultrasonic image data preprocessing unit is used to perform graphic processing on the ultrasonic image information to be processed according to a preset image algorithm to obtain preprocessed ultrasonic image information, wherein the preset image algorithm includes performing privacy information removal, size adjustment, grayscale distribution adjustment, contrast adjustment, noise elimination, and image edge detection on the ultrasonic image information to be processed; The ultrasound image pre-classification unit is used to classify the pre-processed ultrasound image information according to a preset classification algorithm, wherein the preset classification algorithm includes allocating the pre-processed ultrasound image information with similar characteristics to the same training domain according to the device type, imaging parameters, anatomical site and quality level of the pre-processed ultrasound image information.

2. The ultrasound imaging medical record learning system according to claim 1, characterized in that: The grayscale distribution adjustment and the contrast adjustment are implemented by an adaptive grayscale normalization algorithm, specifically a combination of dynamic range compression and contrast enhancement, and the calculation formula is as follows: ； Among them, the , are the local mean and standard deviation of the image in the ultrasonic image information to be processed, is the contrast gain factor, is the brightness offset, The image information in the ultrasonic image information to be processed is located at position The pixel value of The image of the ultrasonic image information to be processed after being processed by the adaptive grayscale normalization algorithm is at position The pixel value of The noise elimination adopts an anisotropic denoising model, combined with non-local mean and gradient constraints, and the calculation formula is as follows: ； in, When is the total variation model, When it is an isotropic smoothing model, Waiting for Processing image information in ultrasonic image information, the for The gradient information of The ultrasonic image information to be processed is The image information at the point The ultrasonic image information to be processed is The image information at the point, and To adjust the coefficients of the data fidelity term and the regularization term weights, w(p,q) is a weight function calculated based on pixel block similarity. is the noise image information in the image information in the ultrasonic image information to be processed, The denoised image information is the denoised image information after the ultrasonic image information to be processed is processed by the anisotropic denoising model; The image edge detection is a multi-scale edge detection, which integrates the Canny operator and the wavelet transform. The calculation formula is as follows: ； Among them, the For the current scale s The Gaussian kernel of is the threshold function, The current scale s The weight coefficient is The current scale s The gradient amplitude under The current scale s The corresponding gradient threshold, The image information in the ultrasonic image information to be processed is located at position The pixel value of The image information in the ultrasonic image information to be processed is converted by the multi-scale edge detection at the position The pixel value of .

3. The ultrasound imaging medical record learning system according to claim 2, characterized in that: The preset classification algorithm includes attention-enhanced convolution, multi-task classification objective function loss, and feature domain similarity measurement, wherein: The attention-enhanced convolution embeds the channel attention mechanism (SE-block) in CNN: ； As stated, For global pooling, and stated is the weight of the fully connected layer, The preprocessed ultrasound image information is The Fourier convolution of the image pixel value at the position, is the exclusive OR operator, is the output feature of the preprocessed ultrasound image information; The classification tasks in the multi-task classification objective function are established according to one or more of the device type, the imaging parameters, the anatomical part and the quality level, and the multi-task classification objective function is as follows: ； When K=n corresponds to n classification tasks, n is the number of classification tasks of the classification task and n is greater than or equal to 1, the is the task weight coefficient, is the true label of category c in the kth task, In the kth task, the probability of category c is is the number of categories of the kth task, All trainable parameters in the model; The feature domain similarity metric is based on the similarity metric of the domain assignment criterion of Wasserstein distance, and the calculation formula is as follows: ； Said is a feature extractor used to measure the distribution differences of different image domains. and stated are different feature domains corresponding to the preprocessed ultrasound image information, is the marginal distribution, and Based on edge distribution Strategy.

4. The ultrasound imaging medical record learning system according to claim 3, characterized in that: The case analysis module also includes a cyclic adversarial learning unit and a resolution enhancement unit, wherein: The cyclic adversarial learning unit is used to perform domain conversion training on the pre-processed ultrasound image information, and improve the clarity and detail expression of the pre-processed ultrasound image information through generator and discriminator adversarial learning; The resolution enhancement unit is used to reconstruct the resolution of the pre-processed ultrasound image information, improve the spatial resolution and detail contrast of the pre-processed ultrasound image information, and generate a medical record of high-resolution ultrasound image information for the doctor user.

5. The ultrasound imaging medical record learning system according to claim 4, characterized in that: It also includes a consultation module, which is used for the patient user to consult with the doctor user and generate the medical record.

6. The ultrasound imaging medical record learning system according to claim 5, characterized in that: It also includes an online registration module for the patient user to register with the doctor user.

7. The ultrasound imaging medical record learning system according to claim 6, characterized in that: It also includes a live broadcast module, which is used by the doctor user to explain the case to the student user.

8. The ultrasound imaging medical record learning system according to claim 7, characterized in that: A payment module is also included.

Images