Ultrasonic image digital intelligent management system, method and equipment and storage medium

Abstract

The invention provides an ultrasonic image digital intelligent management system, method and device and a storage medium, the management system comprises a data governance module, an ultrasonic image intelligent analysis module and an updating iteration optimization module, and an ultrasonic image digital intelligent management workflow integrating data governance and image intelligent analysis is developed. A closed-loop management system is formed from image data storage and standardization processing to analysis result output, and the automation degree and the intelligent level of ultrasonic image data processing are greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of ultrasonic imaging technology, and in particular to an ultrasonic imaging digital intelligent management system, method, device and storage medium. Background Art

[0002] In recent years, with the widespread application of ultrasound imaging technology, medical imaging has become a vital tool for clinical diagnosis, treatment evaluation, and disease monitoring. Ultrasound, due to its non-invasive, real-time, and cost-effective nature, has become the preferred imaging technique for diagnosing cardiovascular, liver, and thyroid diseases. However, with the widespread adoption of ultrasound equipment, ultrasound imaging data has experienced explosive growth, and traditional data management and processing methods face the following challenges: 1) Data standardization and governance issues: Ultrasound image data is usually stored in the DICOM (Digital Imaging and Communications in Medicine) format. The data governance process often involves image format conversion, desensitization, denoising, segmentation and standardization processing. These operations rely on manual completion, which is inefficient and prone to errors. 2) Bottlenecks in the application of intelligent analysis technology: Current intelligent analysis of ultrasound images mainly relies on deep learning technologies, such as convolutional neural networks (CNNs), which are used to distinguish ultrasound image data of different examination sites, different examination sections of the same examination site, and automatically segment or detect diseased tissues. Existing technologies often design models for a single task and cannot fully cover the entire process from data governance to intelligent analysis. 3) The dilemma of data management: With the rapid growth of ultrasound image data, traditional PACS systems face bottlenecks in data query and storage. They lack support for integrated management and intelligent analysis and are unable to meet clinical and scientific research needs.

[0003] Some improvement solutions have been proposed in the prior art for the above problems, among which:

[0004] The invention application with application number CN 201880100205.3 provides an automated clinical workflow, including classification, segmentation and measurement value calculation of echocardiographic images taken by an ultrasound device. However, the image analysis and feature extraction of this solution are only for ultrasound images, mainly relying on a multi-network architecture with a fixed process, and does not integrate data governance functions such as format conversion, desensitization and standardization processing, and cannot form a complete closed loop from data governance to result output.

[0005] The invention application with application number CN202210714067.6 discloses a clinical and scientific research data analysis and management system based on medical testing. The system forms an integrated management model of data collection, storage, analysis and feedback by connecting the clinical testing equipment system, clinical data management system, scientific research data management system and login management system. However, this solution has the following limitations: 1) Insufficient data governance: The application does not involve the desensitization, standardization and format conversion of medical data. It only focuses on the capture and stacking integration of data, and does not solve the problem of low data governance efficiency. 2) Limited intelligent analysis capabilities: The risk prediction function of the analysis module is based on a simple stacking of case information, lacks the application of multimodal data fusion and deep learning algorithms, and cannot meet the high-precision analysis requirements of complex medical imaging data. 3) Insufficient support for imaging data: The application focuses on the analysis and integration of case information and scientific research data. There is a lack of a clear solution for the processing and analysis of unstructured data such as ultrasound images, which limits the applicability of the system in imaging data management scenarios.

[0006] The invention application with application number CN202011378853.0 discloses an end-to-end traceable multi-heterogeneous medical data management platform. The system realizes a closed loop of the entire process from data application and approval to data storage and analysis through modules such as the user data application terminal, the data use approval system, and the policy system. Its features include encrypted distribution of standardized configuration policies, multi-module collaboration, and system log monitoring and alarms. It can provide secure, efficient, and traceable data management services, and is particularly suitable for the storage and remote access of medical research data. However, this application mainly focuses on the storage, access, and security management of heterogeneous data, and does not include specific governance processes for medical imaging data (such as desensitization, format conversion, and standardization), which limits its application in complex imaging data scenarios. In addition, the application lacks intelligent analysis support and does not combine AI models to implement tasks such as medical image classification, feature extraction, and measurement value acquisition, making it difficult to meet clinical needs for intelligent data analysis.

[0007] To address the shortcomings of existing technologies, this application proposes a digital ultrasound image management workflow that integrates data governance and intelligent image analysis. This method, through the deep integration of data governance and intelligent image recognition, achieves closed-loop management of the entire ultrasound image data process, from data upload, storage, processing, to intelligent analysis and output, significantly improving data processing efficiency, analysis accuracy, and clinical applicability. Summary of the Invention

[0008] The present invention proposes a digital and intelligent ultrasound image management system, method, and device that solves the problems of low data standardization and management efficiency in the existing technology and the inability to fully cover the entire process from data management to intelligent analysis. The technical solution of the present invention is achieved as follows:

[0009] An ultrasonic image digital intelligent management system, comprising:

[0010] The data management module is used to parse the DICOM file and pass the parsed DICOM metadata to the ultrasound image intelligent analysis module;

[0011] Ultrasound image intelligent analysis module, which is used to distinguish ultrasound images of different examination sites and sections, segment organ structures and pathological tissues; detect abnormal features or pathological tissues in ultrasound images, and provide accurate location information in real time;

[0012] An update iterative optimization module is used to update and optimize the data governance module and the ultrasonic image intelligent analysis module.

[0013] As a preferred technical solution, the data governance module parses the DICOM file based on the open source library pydicom, and the DICOM metadata includes patient ID, examination date, examination location, examination equipment, executor ID, and patient type classification number.

[0014] As a preferred technical solution, the data governance module includes an image processing submodule and a video processing submodule: the image processing submodule combines with the image processing library OpenCV to realize the input and format conversion of a single image; the video processing module uses inter-frame gradients to create a mask to extract the region of interest ROI, and samples images according to the cardiac cycle, and then converts the format of the collected images.

[0015] As a preferred technical solution, the ultrasound image intelligent analysis module includes an image classification AI model, an image segmentation AI model and an image detection AI model, which are used to perform image classification, segmentation and detection functions respectively.

[0016] As a preferred technical solution, the update iterative optimization module includes:

[0017] Pseudo-label generation module: This module uses the initial ultrasound image intelligent analysis model to generate pseudo-label datasets and supports manual adjustment of labels to reduce the workload of manual labeling;

[0018] Self-learning and model update module: It uses pseudo-label data to update iterative model parameters, improve model performance, and enhance the model's ability to generalize to new data.

[0019] Model deployment and feedback: The model is deployed on the AI ​​server and regular evaluation of model performance is performed to ensure continuous optimization and efficient operation of the system.

[0020] A method for digital and intelligent management of ultrasound images uses the above-mentioned digital and intelligent management system for ultrasound images, including the following steps: receiving ultrasound image detection data from each hospital area and clinic, parsing DICOM files based on an open source library, parsing the DICOM metadata and passing it to an ultrasound image intelligent analysis module, the ultrasound image intelligent analysis module distinguishes ultrasound images of different examination parts and section types, and segments organ structures and diseased tissues; detects abnormal features or diseased tissues in ultrasound images, and provides accurate location information in real time, and updates and optimizes the data governance module and the ultrasound image intelligent analysis module through the update iterative optimization module.

[0021] As a preferred technical solution, the data governance module includes an image processing submodule and a video processing submodule. The workflow of the image processing module is as follows: input a single-frame image, remove the patient's sensitive information above the image according to the device model, and convert the image format as required; the workflow of the video processing submodule is as follows: input a video, use the inter-frame gradient to create a mask to extract the region of interest ROI, sample the image according to the cardiac cycle, and then perform format conversion.

[0022] As a preferred technical solution, the workflow of the image classification AI model is as follows: input image / video, if the input is video, sample the image, normalize the image, distinguish the modalities, coarse cross-section classification, fine cross-section classification, output the top1, top2 predicted categories and confidence; the workflow of the image segmentation AI model is as follows: input image / video, segmentation algorithm, output segmentation results, post-processing, overlay the original image, produce visualization results, calculate the area and volume indicators of the segmented area; the workflow of the image detection AI model is as follows: input image / video, target detection algorithm, locate the target detection structure, filter the detection box according to the confidence threshold, output the target category and location information, overlay the original image, and produce visualization results.

[0023] An ultrasonic image digital intelligent management device, used to execute an ultrasonic image digital intelligent management method, characterized by comprising:

[0024] Memory, used to store original DICOM data and analysis results, supporting high-concurrency queries and data calls;

[0025] GPU servers, which provide computing resources for running deep learning models and implementing intelligent analysis;

[0026] Database: stores ultrasound report measurement data and ultrasound image intelligent analysis results, and allows for quick query;

[0027] Interactive device: provides a friendly interactive interface and supports custom data query and batch export functions.

[0028] A non-temporary storage medium is used to store a program, which is used to enable an ultrasonic image digital intelligent management device to perform the following actions: executing an ultrasonic image digital intelligent management method.

[0029] Compared with the existing technology, this solution has the following beneficial effects: (1) Improvement of automated data management. By integrating DICOM file parsing, format conversion, data desensitization, image denoising and standardization functions, an efficient automated data management process has been built, which is compatible with ultrasound images collected by various ultrasound imaging devices from multiple manufacturers. By reducing manual intervention, processing efficiency and accuracy are improved, ensuring the high quality of input data and the traceability of data, providing a reliable foundation for subsequent analysis; (2) Full-process closed-loop management: Through modular design, a closed-loop system from data collection and governance to analysis and result output has been built, eliminating the problem of link separation, ensuring the consistency and efficiency of data processing, and greatly optimizing the smoothness and efficiency of the workflow; (3) Deep integration of intelligent analysis technology. Through the multimodal AI model, the image features and clinical text information are deeply integrated, covering the entire chain of tasks from image classification and segmentation to lesion detection. It realizes functions such as examination site identification, section classification, lesion tissue segmentation and abnormal target detection, comprehensively improving the breadth and depth of intelligent analysis, and providing important support for accurate diagnosis and scientific research; (4) Enhanced data security and management capabilities: Through hierarchical permission management and encrypted transmission technology, the security and privacy of sensitive data are ensured. At the same time, customized query and batch export functions are supported to meet the needs of users with different roles, achieving flexibility and controllability in data management. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0031] Figure 1 This is the workflow for digital management of ultrasound images in the present invention;

[0032] Figure 2 This is a schematic diagram of a data management module for an ultrasonic image digital intelligent management system according to the present invention;

[0033] Figure 3 This is a schematic diagram of an intelligent analysis module of an ultrasonic image digital intelligent management system of the present invention;

[0034] Figure 4 Schematic diagram of an update iterative optimization module of an ultrasonic image digital intelligent management system of the present invention. DETAILED DESCRIPTION

[0035] The following will clearly and completely describe the technical solutions of the present invention in conjunction with the embodiments of the present invention. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0036] Reference Figure 1 The present invention provides a method for digital management of ultrasound images, comprising the following steps: receiving ultrasound image detection data from each hospital area and clinic, parsing DICOM files based on an open source library, parsing the DICOM metadata and passing it to an ultrasound image intelligent analysis module, wherein the ultrasound image intelligent analysis module distinguishes ultrasound images of different examination parts and section types, and segments organ structures and diseased tissues; detects abnormal features or diseased tissues in ultrasound images, and provides accurate location information in real time, and updates and optimizes the data governance module and the ultrasound image intelligent analysis module through the update iterative optimization module.

[0037] Reference Figures 2-4 As shown, the data governance module includes an image processing submodule and a video processing submodule. The workflow of the image processing module is as follows: input a single-frame image, remove the patient's sensitive information above the image according to the device model, and convert the image format as required; the workflow of the video processing submodule is as follows: input a video, use the inter-frame gradient to create a mask to extract the region of interest ROI, sample the image according to the cardiac cycle, and then perform format conversion.

[0038] The workflow of the image classification AI model is as follows: input image / video, if the input is video, sample the image, normalize the image, distinguish the modalities, coarse cross-section classification, fine cross-section classification, output the top1, top2 predicted categories and confidence; the workflow of the image segmentation AI model is as follows: input image / video, segmentation algorithm, output segmentation results, post-processing, overlay the original image, produce visualization results, calculate the area and volume indicators of the segmented area; the workflow of the image detection AI model is as follows: input image / video, target detection algorithm, locate the target detection structure, filter the detection box according to the confidence threshold, output the target category and location information, overlay the original image, and produce visualization results.

[0039] Specifically, the module design is as follows:

[0040] (1) Data governance module: DICOM file parsing is based on the open source library (pydicom), and DICOM metadata (such as patient ID, examination date, etc.) is parsed and passed to subsequent modules; 1) Image processing module: Combined with the image processing library (OpenCV), it realizes the input and format conversion of single images (such as JPG, PNG, etc.); 2) Video processing module: supports video input, uses inter-frame gradients to create masks to extract regions of interest (ROIs), samples images according to the cardiac cycle, and then performs format conversion (supports AVI, MP4, etc.).

[0041] (2) Ultrasound image intelligent analysis module: 1) Image classification AI model: used to automatically distinguish ultrasound images of different examination sites (such as the heart, liver, and kidneys) and view types (such as the apical four-chamber view and the apical two-chamber view in cardiac ultrasound); 2) Image segmentation AI model: used to automatically segment organ structures and diseased tissues, providing basic support for disease detection and quantitative analysis; 3) Image Detection AI Model: This model is used to automatically detect abnormal features or pathological tissues in ultrasound images, such as tumors and thrombi. Based on the target detection framework, the model can provide accurate location information while performing real-time detection, assisting clinical diagnosis.

[0042] (3) Update iterative optimization module: 1) Pseudo-label generation module: This module uses the initial ultrasound image intelligent analysis model to generate a pseudo-label dataset and supports manual label adjustment to reduce the workload of manual labeling. 2) Self-learning and model updating module: This module uses pseudo-labeled data to iteratively update model parameters, improve model performance, and enhance the model's ability to generalize to new data. 3) Model deployment and feedback: Deploy the model on the AI ​​server and regularly evaluate model performance to ensure continuous optimization and efficient operation of the system.

[0043] The core hardware includes: (1) Memory: stores original DICOM data and analysis results, supports high concurrent query and data call; (2) GPU server: provides high-performance computing resources for running deep learning models and achieving fast and efficient intelligent analysis; (3) Database: stores ultrasound report measurement data and ultrasound image intelligent analysis results, and allows for quick query; (4) Interactive device: provides a friendly interactive interface and supports customized data query and batch export functions.

[0044] Functional design:

[0045] (1) Centralized storage of ultrasound image data: standardized ultrasound image data, image intelligent analysis results, and desensitized ultrasound report measurement data.

[0046] (2) Intelligent analysis of ultrasound images: including automatic differentiation of ultrasound image data of different examination sites and different examination sections of the same examination site; automatic segmentation or detection of diseased tissues, etc.

[0047] (3) Model update and iterative optimization: Use the pseudo-label dataset generated by the initial ultrasound image intelligent analysis model to continuously optimize the ultrasound image intelligent analysis model to improve the model's adaptability and generalization ability to new data.

[0048] (4) Customize data query conditions and export data in batches: Users can customize queries based on examination site, date, lesion type and other conditions, and export relevant data in batches.

[0049] (5) Hierarchical authority management: supports hierarchical user authority settings to ensure data security and controllability, and adapt to the usage needs of different roles.

[0050] Compared with the prior art, the following improvements have been made:

[0051] 1. Comprehensive optimization of automated data governance: This application introduces a variety of automated data governance technologies, including DICOM file parsing, format conversion, data desensitization, image denoising, and standardized processing, which reduces reliance on manual operations, significantly improves the efficiency and accuracy of data governance, and ensures that the quality of input data meets the high standards of clinical and scientific research.

[0052] 2. Innovative Design for Full-Process Closed-Loop Management: This application implements a complete closed-loop process from data standardization and intelligent analysis to output results. Compared to traditional approaches that separate data governance and analysis, this application, through a modular design, deeply integrates data governance and intelligent analysis capabilities, significantly improving data processing efficiency and the overall intelligence level of the system.

[0053] 3. Deep integration of intelligent analysis technology: This application realizes comprehensive intelligent analysis of ultrasound images by combining AI models for image classification, segmentation and detection, including automatic differentiation of examination sites, identification of examination sections, segmentation of diseased tissues, etc., covering the entire chain of tasks from data management to image analysis.

[0054] 4. Self-learning capability and model iterative optimization

[0055] By continuously optimizing the intelligent analysis model using pseudo-labeled datasets, we built an AI system with self-learning capabilities. This not only improves the generalization ability of the model, but also ensures that it can continuously adapt to new clinical and scientific research needs.

[0056] The beneficial effects achieved are as follows:

[0057] 1. Improve the efficiency and intelligence of ultrasound image processing

[0058] By integrating data governance and intelligent analysis technologies, this application significantly reduces manual intervention, improves the efficiency and accuracy of data processing, and provides clinicians and researchers with efficient intelligent tools.

[0059] 2. Enhance the system's scalability and continuous optimization capabilities

[0060] The modular design and pseudo-label self-learning mechanism of this application make the system scalable, adaptable to different clinical scenarios and scientific research needs, and maintain the technological advancement through continuous iteration.

[0061] 3. Optimize clinical workflow and decision-making efficiency

[0062] This application optimizes the workflow of ultrasound imaging from acquisition to result application through full-process automation and intelligence, providing clinicians with more efficient decision support tools and improving diagnostic efficiency and accuracy.

[0063] Based on the above innovations, this application achieves innovation and breakthroughs in ultrasound imaging data processing and intelligent analysis methods by integrating automated data governance, intelligent analysis technology, and full-process closed-loop management. By introducing multi-task AI models and automated data governance technology, the system not only significantly improves data processing efficiency and analysis accuracy, but also enhances the interpretability and clinical applicability of analysis results, providing more efficient and accurate intelligent tools for clinical diagnosis and scientific research applications, helping to promote the intelligent development of ultrasound medical imaging.

[0064] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An ultrasonic image digital intelligent management system, characterized in that: include: The data management module is used to parse the DICOM file and pass the parsed DICOM metadata to the ultrasound image intelligent analysis module; Ultrasound image intelligent analysis module, which is used to distinguish ultrasound images of different examination sites and sections, segment organ structures and pathological tissues; detect abnormal features or pathological tissues in ultrasound images, and provide accurate location information in real time; An update iterative optimization module is used to update and optimize the data governance module and the ultrasonic image intelligent analysis module.

2. The ultrasonic image digital intelligent management system according to claim 1, characterized in that: It also includes the data governance module parsing the DICOM file based on the open source library pydicom, and the DICOM meta-information includes patient ID, examination date, examination location, examination equipment, executor ID, and patient type classification number.

3. The ultrasonic image digital intelligent management system according to claim 1, characterized in that: The data management module includes an image processing submodule and a video processing submodule: the image processing submodule combines the image processing library OpenCV to realize the input and format conversion of a single image; The video processing module uses inter-frame gradients to create a mask to extract the region of interest (ROI), samples images according to the cardiac cycle, and then performs format conversion on the collected images.

4. The ultrasonic image digital intelligent management system according to claim 1, characterized in that: The ultrasonic image intelligent analysis module includes an image classification AI model, an image segmentation AI model and an image detection AI model, which are used to perform image classification, segmentation and detection functions respectively.

5. The ultrasonic image digital intelligent management system according to claim 1, characterized in that: The update iterative optimization module includes: Pseudo-label generation module: This module uses the initial ultrasound image intelligent analysis model to generate pseudo-label datasets and supports manual adjustment of labels to reduce the workload of manual labeling; Self-learning and model updating module: It uses pseudo-labeled data to update iterative model parameters, improve model performance, and enhance the model's ability to generalize to new data; Model deployment and feedback: The model is deployed on the AI ​​server and regular evaluation of model performance is performed to ensure continuous optimization and efficient operation of the system.

6. A method for digital management of ultrasound images, characterized in that: An ultrasound image digital intelligent management system as described in any one of claims 1 to 5 above is used, comprising the following steps: receiving ultrasound image detection data from each hospital area and clinic, parsing DICOM files based on an open source library, parsing the DICOM metadata and passing it to an ultrasound image intelligent analysis module, the ultrasound image intelligent analysis module distinguishes ultrasound images of different examination parts and section types, and segments organ structures and diseased tissues; detects abnormal features or diseased tissues in ultrasound images, and provides accurate location information in real time, and updates and optimizes the data governance module and the ultrasound image intelligent analysis module through the update iterative optimization module.

7. The method for digital management of ultrasound images according to claim 6, characterized in that: The data governance module includes an image processing submodule and a video processing submodule. The workflow of the image processing module is as follows: input a single-frame image, remove the patient's sensitive information above the image according to the device model, and convert the image format as required; the workflow of the video processing submodule is as follows: input a video, use the inter-frame gradient to create a mask to extract the region of interest (ROI), sample the image according to the cardiac cycle, and then perform format conversion.

8. The method for digital management of ultrasound images according to claim 6, wherein: The workflow of the image classification AI model is as follows: input image / video, if the input is video, sample the image, normalize the image, distinguish the modalities, perform coarse and fine cross-sectional classification, and output the top 1 and top 2 predicted categories and confidence levels. The workflow of the image segmentation AI model is as follows: input image / video, segmentation algorithm, output segmentation results, post-processing, overlaying the original image, producing visualization results, and calculating the area and volume indicators of the segmented area; The workflow of the image detection AI model is as follows: input image / video, target detection algorithm, locate target detection structure, filter detection box according to confidence threshold, output target category and location information, overlay the original image, and generate visualization results.

9. An ultrasonic image digital intelligent management device, used to execute the ultrasonic image digital intelligent management method according to any one of claims 6 to 8, characterized in that: include: Memory, used to store original DICOM data and analysis results, supporting high-concurrency queries and data calls; GPU servers, which provide computing resources for running deep learning models and implementing intelligent analysis; Database: stores ultrasound report measurement data and ultrasound image intelligent analysis results, and allows for quick query; Interactive device: provides a friendly interactive interface and supports custom data query and batch export functions.

10. A non-temporary storage medium, characterized in that: It is used to store a program, which is used to enable the ultrasonic image digital management device as described in claim 9 to perform the following actions: execute an ultrasonic image digital management method as described in any one of claims 6 to 8 above.

Images