Ultrasonic automatic report generation system based on artificial intelligence

Abstract

The invention discloses an automatic ultrasonic report generation system based on artificial intelligence, which relates to the technical field of ultrasound and comprises an image data acquisition module, an ultrasonic image data processing module, an image data depth analysis module, an automatic ultrasonic report generation module and a user interaction feedback module. According to the invention, through the ultrasonic report automatic generation module, automatically identified and extracted medical features can be conveniently converted into a readable artificial intelligence ultrasonic report, so that a clinician can quickly and accurately check and use the medical features, and the lesion risk of a patient can be further predicted according to the current disease of the patient; meanwhile, based on the current disease and lesion risk, suggestions for further examination and treatment are provided for the patient, a current ultrasonic examination report is compared and analyzed with historical case data, a diagnosis reference is provided through a comparison result, a doctor is assisted to make a decision, the workload of the doctor is reduced, and the diagnosis efficiency is improved. And the medical service quality and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the field of ultrasound technology, and in particular to an automatic ultrasound report generation system based on artificial intelligence. Background Art

[0002] Ultrasound examination is a method of examining the human body using ultrasonic technology, including B-ultrasound, color ultrasound, and elastic imaging ultrasound. Ultrasound is a mechanical wave with a frequency exceeding 20,000 Hz, which is beyond the range of human hearing. The working principle of ultrasound examination is to radiate sound waves into tissues and judge tumors or other abnormalities based on the strength and depth of the echo. By detecting reflected waves, different tissue characteristics are identified and the information is displayed in the form of images, curves, or tiles to diagnose internal diseases of the human body. In the medical field, ultrasound examination is a commonly used diagnostic method. As the complexity and specialization of ultrasound image interpretation tasks continue to increase, users have higher and higher requirements for the accuracy and efficiency of ultrasound image reports. There is also an increasing demand for using artificial intelligence image recognition to generate ultrasound image reports as an auxiliary tool for physicians' diagnosis and treatment. The existing Chinese patent, CN201811350730.9, is a system for automatically generating intelligent analysis reports based on big data. It proposes a system for aggregating data resources and knowledge resources from different data sources, and intelligently pushing related data and related knowledge based on themes; a knowledge graph system is constructed based on knowledge logic and attribute relationships, and automatically associates knowledge related to subject terms based on subject terms to achieve intelligent retrieval and question-answering; a model engine system is used to associate analysis models involved in application fields corresponding to subject terms to achieve in-depth analysis of indicator data; an XML expert opinion library automatic push system is used to automatically associate knowledge points corresponding to subject terms, automatically push problem-solving knowledge points, and automatically generate intelligent reports. Compared with traditional intelligent reports, intelligent analysis reports automatically generated based on big data are objective and authoritative, and can greatly improve the work efficiency of researchers; Traditional ultrasound imaging analysis and report generation mainly rely on the personal experience of ultrasound physicians for analysis. Manual analysis and reporting methods rely on personal experience, and the report content is highly subjective, and the accuracy of the report results cannot be guaranteed. At the same time, when analyzing and processing ultrasound images, the image information cannot be effectively processed, analyzed, and deeply interpreted, resulting in the subsequent report content being unable to fully reflect the patient's disease information, and unable to provide patients with further examinations and effective treatment recommendations, affecting the efficiency of patients' actual diagnosis and treatment. Summary of the Invention

[0003] The present invention provides an artificial intelligence-based ultrasound automatic report generation system, which can effectively solve the problem proposed in the above background technology that traditional ultrasound imaging analysis and report generation mainly rely on the personal experience of ultrasound physicians for analysis, the manual analysis and reporting methods rely on personal experience, the report content is highly subjective, and the accuracy of the report results cannot be guaranteed. At the same time, when analyzing and processing ultrasound images, the image information cannot be effectively processed, analyzed and deeply interpreted, resulting in the subsequent report content being unable to fully reflect the patient's disease information, and unable to provide patients with further examinations and effective treatment suggestions, affecting the efficiency of the patient's actual diagnosis and treatment.

[0004] To achieve the above objectives, the present invention provides the following technical solutions: an artificial intelligence-based automatic ultrasound report generation system that realizes the process from ultrasound image data processing to automatic generation of accurate and detailed ultrasound medical reports, assisting doctors in making precise decisions; It includes image data acquisition module, ultrasound image data processing module, image data depth analysis module, ultrasound report automatic generation module, and user interaction feedback module; The image data acquisition module is responsible for collecting and acquiring ultrasound images, and the ultrasound image data processing module pre-processes the acquired ultrasound images; The image data in-depth analysis module identifies disease information in ultrasound images, the ultrasound report automatic generation module converts the analysis results into artificial intelligence ultrasound reports that meet clinical standards, and the user interaction feedback module establishes a two-way collaborative closed loop between doctors and the system; The ultrasound image data processing module denoises and enhances the original image, and adjusts the denoising intensity according to the local contrast of the image through an adaptive denoising factor, thereby ensuring the preservation of image details; The calculation formula is as follows:

[0005] in, Calculated by the following formula, it reflects the change in local contrast of the image:

[0006] Denoising Factor According to the local contrast Dynamic calculation, low-contrast areas rely more on denoising, while high-contrast areas enhance image details; in, It represents the local contrast of the image around the pixel (x,y), which can be obtained by calculating the local standard deviation or grayscale gradient.

[0007] According to the technical scheme, the image data acquisition and obtaining module is a front-end data entrance of the system, is responsible for connecting various ultrasonic imaging devices of the hospital, collects and obtains original ultrasonic image information from the ultrasonic examination device, and receives and obtains device examination parameters and related patient information.

[0008] According to the technical scheme, the ultrasonic image data processing module pre-processes the collected original ultrasonic image, including image denoising, image enhancement and image segmentation. By processing the original ultrasonic image information, in the process of ultrasonic image processing, the ultrasonic image also needs to be calibrated to form a unified standard for the size, color and gray scale of each ultrasonic image after calibration.

[0009] According to the technical scheme, the ultrasonic image data processing module manually marks the feature significant medical information on the ultrasonic image by a manual marking method in the process of processing the ultrasonic image, and records the manually marked information. Image segmentation is realized by extracting key features through a deep learning model. In order to improve the segmentation accuracy, the features extracted by different convolution layers and image spatial information are fused, so that the model can better cope with complex ultrasonic image structures. The calculation formula is as follows:

[0010] In the ultrasonic image, part of the image may lose information due to noise or other reasons. A generative adversarial network (GAN) is used to repair the missing part, and the context information of the image is used to infer the content of the missing area.

[0011] The generative network G generates a complete image according to the known enhanced image and the missing area information, and generates a complete image. In order to ensure the effect of image processing, a quality evaluation mechanism is introduced to evaluate the quality of each image in real time, and the subsequent image processing steps are adjusted according to the quality score Q. If the image quality is lower than the set threshold, the system will automatically trigger the reprocessing process .

[0012] According to the technical scheme, the image data deep analysis module uses a deep learning model to identify and extract key features of the pre-processed ultrasonic image, and detects and quantifies the medical feature information fed back in the image information by identifying and extracting key features in the ultrasonic image. The key features include the location of nodules and cysts in the image, as well as the size of nodules and the area of ​​cysts. It is also necessary to identify and locate the distribution of various organs in the human body in the ultrasound image.

[0013] According to the above technical solution, the image data in-depth analysis module also supports a specific pattern image recognition function, which is used to assist in the diagnosis of complex diseases by identifying specific features in ultrasound images, and quickly and accurately identify the growth trend and type of tumors.

[0014] According to the above technical solution, the automatic ultrasound report generation module converts the automatically identified and extracted medical features into natural and coherent medical text, realizing the conversion of complex analysis results into readable artificial intelligence ultrasound reports, allowing clinicians to view and use them quickly and accurately. It can also combine the experience and clinical knowledge accumulated over a long period of time to identify the nature of the lesions, generate more specific diagnostic suggestions, and achieve in-depth analysis of the lesions.

[0015] According to the above technical solution, the automatic ultrasound report generation module can further predict the patient's lesion risk based on the patient's current symptoms during the process of generating the intelligent ultrasound examination report, and provide the patient with recommendations for further examination and treatment based on the current symptoms and lesion risk; The current ultrasound examination report can also be compared and analyzed with historical case data, and the comparison results can provide diagnostic references to assist doctors in making decisions.

[0016] According to the above technical solution, the user interaction feedback module includes a user interaction unit and a user feedback unit; The user interaction unit is an information interaction interface between patients, doctors and the system, which facilitates doctors and patients to have a deeper understanding and feedback of the content information of the report, enables patients to understand the content of the report information more accurately, and at the same time clarify the lesion prediction and subsequent treatment recommendations in the report information, and also supports doctors to modify the result information of the automatic report.

[0017] According to the above technical solution, the user feedback unit collects user feedback information and converts the user feedback information into a basis for system improvement. By building a convenient user feedback channel, patients can put forward improvement suggestions for any aspect of the report, and doctors can put forward more professional or detailed interpretation requirements, and encourage users to feedback on usage experience and error information.

[0018] Compared with the prior art, the present invention has the following beneficial effects: 1. The image data acquisition obtaining module facilitates the collection of original ultrasound image information from the ultrasound examination device, and after obtaining the image data and patient information, the obtained information is classified and arranged for subsequent processing. When classifying and arranging the ultrasound image data information, the ultrasound image data information is associated with a unified database, and the ultrasound image information is stored in accordance with the time sequence arrangement order and the detected part of the patient's body, so as to facilitate subsequent quick search and application, and improve the efficiency of subsequent image analysis and application.

[0019] 2. The ultrasound image data processing module facilitates the processing of the collected original ultrasound image, optimizes the image quality and highlights the medical features, facilitates the subsequent image analysis and intelligent report generation, and manually marks the feature significant medical information on the ultrasound image by manual labeling method, so as to facilitate the use of special manual labeling to help subsequent fast, significant and accurate identification of feature significant medical information. The image data deep analysis module facilitates the identification and extraction of key features of the preprocessed ultrasound image, detects and quantifies the medical feature information feedback in the image information by identifying and extracting the key features in the ultrasound image, and uses the specific features in the ultrasound image to assist in the diagnosis of complex diseases, and quickly and accurately identifies the growth trend and type of tumor.

[0020] 3. The ultrasound report automatic generation module facilitates the conversion of automatically identified and extracted medical features into readable artificial intelligence ultrasound reports, so that the clinician can quickly and accurately view and use, and combine the long-term accumulated experience and clinical knowledge to identify the nature of the lesion, generate more specific diagnosis suggestions, and the ultrasound report automatic generation module can further predict the risk of the patient's lesion according to the patient's current symptoms, and provide further examination and treatment suggestions for the patient based on the current symptoms and lesion risk, and compare the current ultrasound examination report with the historical case data to provide diagnosis reference through the comparison result, assist the doctor to make decision, reduce the doctor's work burden, and improve the quality and efficiency of medical service.

[0021] 4. The user interaction feedback module facilitates the doctor and patient to understand and feedback the content information of the report more deeply, so that the patient can understand the report information content more accurately, and also supports the doctor to modify the automatic report result information, and facilitates further discussion and communication between the patient and the doctor. And the user's feedback information is collected and converted into the basis for system improvement, a user problem feedback database is established, common error information is stored in the database, the status of each feedback request is recorded and tracked, and it is ensured that all feedback information can be processed in time and targeted, so as to facilitate the continuous optimization of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0023] In the attached figure: Figure 1 It is an architectural diagram of the report generation system of the present invention. DETAILED DESCRIPTION

[0024] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0025] Example: Figure 1 As shown, the present invention provides a technical solution, an artificial intelligence-based automatic ultrasound report generation system, which realizes the process from ultrasound image data processing to automatic generation of accurate and detailed ultrasound medical reports, improves the efficiency and accuracy of report generation, and assists doctors in making accurate decisions; It includes image data acquisition module, ultrasound image data processing module, image data depth analysis module, ultrasound report automatic generation module, and user interaction feedback module; The image data acquisition module is responsible for collecting and acquiring ultrasound images, and the ultrasound image data processing module pre-processes the acquired ultrasound images; The image data in-depth analysis module identifies disease information in ultrasound images, the ultrasound report automatic generation module converts the analysis results into artificial intelligence ultrasound reports that meet clinical standards, and the user interaction feedback module builds a two-way collaborative closed loop between doctors and the system, achieving system optimization through feedback.

[0026] Based on the above technical solution, the image data acquisition module serves as the front-end data entry of the system. It is responsible for connecting with various ultrasound imaging devices in the hospital, collecting and acquiring raw ultrasound image information from ultrasound examination equipment, and receiving and acquiring equipment examination parameters and relevant patient information. Among them, the equipment inspection parameters include the probe frequency of the ultrasound inspection and the setting information of the inspection equipment, and the relevant patient information includes the patient's medical record number, name, and ultrasound inspection site; After acquiring image data and patient information, it is necessary to ensure the consistency and quality of the data format. At the same time, the acquired information needs to be classified and organized for subsequent processing. When classifying and organizing the ultrasound image data information, the ultrasound image data information is associated with a unified database. The ultrasound image data information is arranged in the database according to the order of the time series. Under the same time series, the ultrasound image information is divided and stored according to the examination part of the patient's body; In addition, during the image data acquisition process, a data integrity check mechanism is introduced to automatically detect whether the received image data and patient information are missing, formatted abnormally, or have content errors; Once data anomalies are detected, the system will automatically record and generate a prompt log so that operators can handle them in a timely manner; At the same time, in order to improve the efficiency of data interaction, the system supports interface docking with the hospital information system (HIS) and the picture archiving and communication system (PACS) to realize automatic acquisition of patient examination form information, equipment configuration parameters and examination images, and reduce manual input errors.

[0027] After the data is stored in a unified manner, the system will automatically generate a unique identification code (such as UID) for each set of images and their corresponding information, and organize and manage them as follows: Primary index: sorted by check timestamp; Secondary index: Under examination time, classified by patient ID; Level 3 index: under patient ID, classified by examination site (e.g., "liver," "breast," "thyroid"); Four-level indexing: For each part, image frames and device parameter information are recorded, such as probe type, frequency, grayscale gain settings, etc.

[0028] This structure enables subsequent modules such as image preprocessing, feature extraction, and report generation to quickly and accurately locate target image data, and supports retrospective and statistical analysis by conditions (time, patient, and location), significantly improving the system's data organization efficiency and clinical application scalability.

[0029] Based on the above technical solution, the ultrasound image data processing module pre-processes the collected original ultrasound images, including image denoising, image enhancement and image segmentation; Image denoising is the process of removing noise from the original ultrasound image to improve image quality. Image enhancement is used to improve image quality and enhance detailed information, thereby optimizing it for further analysis. Image segmentation uses deep learning technology to achieve accurate target detection and segmentation, distinguishing different tissues or structures in ultrasound images, and specifically accurately segmenting organs, blood vessels, and tumors in the image. By processing the original ultrasound image information, the accuracy and efficiency of subsequent processing are improved, the image quality is optimized and the medical features are highlighted, laying the foundation for subsequent recognition and analysis, and facilitating subsequent image analysis and intelligent report generation; During the ultrasound image processing process, the ultrasound images also need to be calibrated to a standard so that the size, color, and grayscale of each ultrasound image are calibrated to form a unified standard, which facilitates further analysis of the subsequent ultrasound images. To improve the quality of ultrasound images, the original image must first be denoised and enhanced. Adaptive denoising factors are used to adjust the denoising intensity based on the local contrast of the image, thus avoiding uniform denoising of the entire image and ensuring the preservation of image details. in: : Original ultrasound image; : Image enhancement part, which can increase detail information; : Denoising factor, dynamically adjusted according to the local contrast of the image; The calculation formula is as follows:

[0030] in, Calculated by the following formula, it reflects the change in local contrast of the image:

[0031] Denoising Factor According to the local contrast Dynamic calculation, low-contrast areas rely more on denoising, while high-contrast areas enhance image details; in, Represents the local contrast of the image around the pixel (x, y), which can be obtained by calculating the local standard deviation or grayscale gradient; Image segmentation is achieved by extracting key features through a deep learning model. To improve segmentation accuracy, we fuse features extracted by different convolutional layers with image spatial information, enabling the model to better cope with complex ultrasound image structures. in: : The feature map of the i-th layer is extracted from the enhanced image; : Image spatial information (such as the local structure of the image); : The weight of each feature map is learned according to its importance to the segmentation task; The calculation formula is as follows:

[0032] Features extracted using convolutional layers and the spatial information of the image Fusion, weight Used to dynamically adjust the importance of each feature and obtain the final segmentation result through the sigmoid function σ ; Ultrasound images may have information lost due to noise or other reasons. To address this issue, we use a generative adversarial network (GAN) to repair the lost parts and use the image context to infer the content of the missing area. G: Generative Adversarial Network (GAN); : Enhanced ultrasound image; : missing image part;

[0033] Generate network G based on known enhanced image and missing regions Information, perform image restoration, and generate a complete image; In order to ensure the effect of image processing, we introduced a quality assessment mechanism to evaluate the quality of each image in real time and adjust the subsequent image processing steps according to the quality score Q. If the image quality is lower than the set threshold, the system will automatically trigger the reprocessing process. : processed image; : original image; : The weight of each pixel is adjusted according to the importance of the image area;

[0034] The quality score Q is used to evaluate the image quality by calculating the difference between the processed image and the original image. Dynamically adjusts based on the complexity of image features to help the system determine which areas require more processing.

[0035] For example, a frame of abdominal ultrasound image from a hospital ultrasound examination system has an image size of: 512×5125 pixels, Grayscale range: [0, 255], Image ID: US_20250106_00123.

[0036] The original grayscale value of the local image area (x = 250, y = 300) is:

[0037] After the local image enhancement module is enhanced by histogram, the output is:

[0038] The local contrast of the image is calculated using a 7×7 window to calculate the standard deviation, which is:

[0039] Set the parameter β=4.0 and substitute into the adaptive factor formula:

[0040] Therefore, the image output is:

[0041] The enhanced image is input into the deep neural network to extract 3 layers of convolution features (shallow , middle level , deep ), and spatial structure information :

[0042] The segmentation output passes through the sigmoid activation function: Z=0.25*0.6232+0.45*0.5544+0.30*0.6188=0.5909;

[0043] The pixel was determined to be the edge area of ​​the lesion (the segmentation threshold was set to 0.5).

[0044] Based on the above technical solution, the ultrasound image data processing module can also manually mark medical information with significant features on the ultrasound image through a manual annotation method during the process of processing the ultrasound image, and record the manually marked information; Special manual annotation is used to help quickly, significantly and accurately identify medical information with significant features, making it easy to quickly analyze the medical feature information corresponding to the feature type.

[0045] Based on the above technical solution, the image data deep analysis module uses a deep learning model to identify and extract key features of the pre-processed ultrasound images. The deep learning model uses a convolutional neural network (CNN) to detect and quantify the medical feature information fed back in the image information by identifying and extracting key features in the ultrasound images. The key features include the location of nodules and cysts in the image, as well as the size of nodules and the area of ​​cysts. It is also necessary to identify and locate the distribution of various organs in the human body in the ultrasound image.

[0046] Based on the above technical solutions, the image data deep analysis module also supports pattern-specific image recognition. By identifying specific features in ultrasound images, it can be used to assist in the diagnosis of complex diseases and quickly and accurately identify tumor growth trends and types. In addition, the image data deep analysis module integrates a feature recognition network based on a multi-task learning framework. While extracting image structural features, it can also simultaneously predict multiple medical indicators such as lesion type, boundary morphology, and tissue structural damage. This module supports the identification and quantification of key features such as tumor edge irregularity, internal echo distribution, calcification point characteristics, and blood flow distribution patterns. During tumor recognition, the system uses a trained deep learning model to compare and analyze lesion change trends in historical and current images, automatically estimating information such as the rate of tumor volume change and morphological evolution. Combined with time series analysis, it can further infer the tumor's possible growth behavior (such as slow growth, rapid enlargement, or potential malignancy). The module also has a built-in library of typical pathological feature maps, which supports similarity matching between image features and typical benign and malignant templates. It is used to assist in judging the benign and malignant risk level of tumors, improve the early screening and accurate judgment capabilities of complex cases, and provide doctors with more specific decision-making support.

[0047] Based on the above technical solution, the automatic ultrasound report generation module converts automatically identified and extracted medical features into natural and coherent medical text. This allows complex analysis results to be converted into readable AI ultrasound reports, allowing clinicians to quickly and accurately review and use them. It can also combine long-term accumulated experience and clinical knowledge to identify the nature of lesions, generate more specific diagnostic recommendations, and achieve in-depth analysis of lesions. After generating an AI ultrasound report, pattern recognition and semantic analysis of the report content are required to ensure that the generated report can be clearly conveyed to the doctor. Pattern recognition is to introduce a review mechanism to perform pattern recognition on the report content to generate a standard format intelligent ultrasound examination report, so that the report meets the formal requirements of scientific reports and ensures the accuracy and professionalism of the report content. Semantic analysis uses natural language processing technology to generate clear and accurate report text from the report content, avoiding ambiguity in the report content and ensuring that the generated report text contains accurate diagnostic results, patient information, and treatment recommendations.

[0048] Based on the above technical solution, the automatic ultrasound report generation module can further predict the patient's lesion risk based on the patient's current symptoms during the process of generating an intelligent ultrasound examination report. At the same time, based on the current symptoms and lesion risk, it can provide the patient with recommendations for further examination and treatment. The current ultrasound examination report can also be compared and analyzed with historical case data, and the comparison results can provide diagnostic references to assist doctors in making decisions.

[0049] Based on the above technical solution, the user interaction feedback module includes a user interaction unit and a user feedback unit; The user interaction unit is the information exchange interface between patients, doctors and the system. It facilitates doctors and patients to have a deeper understanding of and provide feedback on the content of reports, enabling patients to more accurately understand the content of reports, clarify the lesion predictions and subsequent treatment recommendations in the reports, and support doctors to modify the results of automatic reports. The user interaction unit also supports visualization functions and user authentication management. The visualization function facilitates the establishment of a channel for real-time interaction between patients and doctors, and facilitates further discussion and communication between patients and doctors. User authentication management uses access control policies to allocate user permissions according to requirements, preventing unauthorized personnel from accessing sensitive information, and ensuring the security of patient diagnostic information by authorizing and authenticating users.

[0050] Based on the above technical solution, the user feedback unit collects user feedback information and converts it into a basis for system improvement. By building a convenient user feedback channel, patients can provide suggestions for improvement on any aspect of the report, and doctors can provide more professional or detailed interpretation requirements. Users are encouraged to provide feedback on their usage experience and error information. By building a dedicated channel for user feedback, we can promptly identify and resolve user feedback issues. We also establish a user feedback database to store common error messages, record and track the status of each feedback request, and ensure that all feedback is handled promptly and in a targeted manner. Common reporting error messages can also be automatically identified within the database to provide users with timely responses.

[0051] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. The artificial intelligence-based ultrasound automatic report generation system is characterized by: Realize the data processing from ultrasound images to automatically generate accurate and detailed ultrasound medical reports, assisting doctors in making precise decisions; It includes image data acquisition module, ultrasound image data processing module, image data depth analysis module, ultrasound report automatic generation module, and user interaction feedback module; The image data acquisition module is responsible for collecting and acquiring ultrasound images, and the ultrasound image data processing module pre-processes the acquired ultrasound images; The image data in-depth analysis module identifies disease information in ultrasound images, the ultrasound report automatic generation module converts the analysis results into artificial intelligence ultrasound reports that meet clinical standards, and the user interaction feedback module establishes a two-way collaborative closed loop between doctors and the system; The ultrasound image data processing module denoises and enhances the original image, and adjusts the denoising intensity according to the local contrast of the image through an adaptive denoising factor, thereby ensuring the preservation of image details; The calculation formula is as follows: ； in, Calculated by the following formula, it reflects the change in local contrast of the image: ； Denoising Factor According to the local contrast Dynamic calculation, low-contrast areas rely more on denoising, while high-contrast areas enhance image details; in, It represents the local contrast of the image around the pixel (x,y), which can be obtained by calculating the local standard deviation or grayscale gradient.

2. The artificial intelligence-based automatic ultrasound report generation system according to claim 1, characterized in that: The image data acquisition module serves as the front-end data entry of the system and is responsible for connecting with various ultrasonic imaging equipment in the hospital, collecting and acquiring original ultrasonic image information from the ultrasonic examination equipment, and receiving and acquiring equipment examination parameters and related patient information.

3. The artificial intelligence-based automatic ultrasound report generation system according to claim 2, characterized in that: The ultrasound image data processing module pre-processes the collected original ultrasound image, including image denoising, image enhancement and image segmentation; By processing the original ultrasound image information, during the ultrasound image processing process, the ultrasound image also needs to be calibrated to a standard so that the size, color, and grayscale of each ultrasound image form a unified standard after calibration.

4. The artificial intelligence-based automatic ultrasound report generation system according to claim 3, characterized in that: The ultrasound image data processing module manually marks medical information with significant features on the ultrasound image by a manual annotation method during the process of processing the ultrasound image, and records the manually marked information; Image segmentation is achieved by extracting key features through a deep learning model. To improve segmentation accuracy, the features extracted by different convolutional layers are integrated with image spatial information, enabling the model to better cope with complex ultrasound image structures. The calculation formula is as follows: ； In ultrasound images, some parts may be lost due to noise or other reasons. Generative Adversarial Networks (GANs) are used to repair the lost parts and infer the content of the missing areas using the context of the image. ； Generate network G based on known enhanced image and missing regions Information, perform image restoration, and generate a complete image; In order to ensure the effect of image processing, a quality assessment mechanism is introduced to evaluate the quality of each image in real time and adjust the subsequent image processing steps according to the quality score Q. If the image quality is lower than the set threshold, the system will automatically trigger the reprocessing process. 。 5. The artificial intelligence-based automatic ultrasound report generation system according to claim 1, characterized in that: The image data deep analysis module uses a deep learning model to identify and extract key features of the pre-processed ultrasound image, and detects and quantifies the medical feature information fed back in the image information by identifying and extracting key features in the ultrasound image; The key features include the location of nodules and cysts in the image, as well as the size of nodules and the area of ​​cysts. It is also necessary to identify and locate the distribution of various organs in the human body in the ultrasound image.

6. The artificial intelligence-based automatic ultrasound report generation system according to claim 5, characterized in that: The image data in-depth analysis module also supports a specific pattern of image recognition function, which is used to assist in the diagnosis of complex diseases by identifying specific features in ultrasound images, and quickly and accurately identify the growth trend and type of tumors.

7. The artificial intelligence-based automatic ultrasound report generation system according to claim 1, characterized in that: The automatic ultrasound report generation module converts automatically identified and extracted medical features into natural and coherent medical text, realizing the conversion of complex analysis results into readable artificial intelligence ultrasound reports, allowing clinicians to view and use them quickly and accurately. It can also combine long-term accumulated experience and clinical knowledge to identify the nature of lesions, generate more specific diagnostic recommendations, and achieve in-depth analysis of lesions.

8. The artificial intelligence-based automatic ultrasound report generation system according to claim 7, characterized in that: The automatic ultrasound report generation module can also further predict the patient's lesion risk based on the patient's current symptoms during the process of generating the intelligent ultrasound examination report, and provide the patient with recommendations for further examination and treatment based on the current symptoms and lesion risk; The current ultrasound examination report can also be compared and analyzed with historical case data, and the comparison results can provide diagnostic references to assist doctors in making decisions.

9. The artificial intelligence-based automatic ultrasound report generation system according to claim 1, characterized in that: The user interaction feedback module includes a user interaction unit and a user feedback unit; The user interaction unit is an information interaction interface between patients, doctors and the system, which facilitates doctors and patients to have a deeper understanding and feedback of the content information of the report, enables patients to understand the content of the report information more accurately, and at the same time clarifies the lesion prediction and subsequent treatment recommendations in the report information, and also supports doctors to modify the result information of the automatic report.

10. The artificial intelligence-based automatic ultrasound report generation system according to claim 9, characterized in that: The user feedback unit collects user feedback information and converts the user feedback information into a basis for system improvement. By building a convenient user feedback channel, patients can put forward improvement suggestions for any aspect of the report, and doctors can put forward more professional or detailed interpretation requirements, and encourage users to feedback usage experience and error information.

Images