Ultrasonic image analysis and diagnosis system based on artificial intelligence

Abstract

The invention discloses an ultrasonic image analysis and diagnosis system based on artificial intelligence, and the system comprises an image collection module which is used for obtaining a real-time ultrasonic image, an image processing module which carries out the preprocessing of the image, an artificial intelligence analysis module which analyzes the image through a deep learning model, and a feature extraction training module which extracts features and trains the model. The classification and prediction module judges whether an image is abnormal and outputs a prediction result, the verification module evaluates system performance to verify accuracy, sensitivity and specificity, the diagnosis module performs feature extraction and classification prediction on a new image and provides a diagnosis result, and the user interaction module displays the diagnosis result and supports a doctor to further make a decision. Compared with the prior art, the system utilizes a deep learning technology to realize rapid image processing and automatic analysis, and can generate a diagnosis result in a short time.

Description

Technical Field

[0001] The present application relates to the field of medical imaging technology, and specifically to an ultrasound image analysis and diagnosis system based on artificial intelligence. Background Art

[0002] As a non-invasive, radiation-free medical imaging device, ultrasound diagnostic equipment is widely used in clinical diagnosis. Traditional ultrasound diagnosis relies on the doctor's experience and expertise to observe and interpret images. However, since ultrasound images are affected by many factors (such as patient body shape, probe position, operating techniques, etc.), the image quality is uneven, and the accuracy and consistency of the diagnostic results are limited. In addition, the interpretation process of ultrasound images is complicated and time-consuming, and is prone to human errors.

[0003] With the rapid development of artificial intelligence (AI) technology, especially the application of deep learning algorithms in the field of image processing, new solutions have been provided for ultrasound image analysis. By training deep learning models, lesions in ultrasound images can be automatically identified and classified, improving the accuracy and efficiency of diagnosis. Summary of the invention

[0004] In view of the shortcomings of the prior art, the purpose of this application is to provide an ultrasound image analysis and diagnosis system based on artificial intelligence to solve the problems raised in the above background technology.

[0005] According to one aspect of the present application, an ultrasound image analysis and diagnosis system based on artificial intelligence includes:

[0006] An image acquisition module, used to acquire real-time ultrasound images, including images under normal and abnormal conditions, through an ultrasound probe, and transmit the image data to a processing unit;

[0007] An image processing module, used for preprocessing the acquired ultrasound images, including denoising, standardization and image enhancement operations;

[0008] An artificial intelligence analysis module, which is used to select and design deep learning models to analyze processed ultrasound images;

[0009] A feature extraction training module is used to extract features from the preprocessed ultrasound image using a deep learning model, and input the extracted features into the deep learning model for training;

[0010] The classification and prediction module is used to determine whether there are abnormalities or lesions in the image based on the trained features and patterns, and output the prediction results;

[0011] The verification module is used to verify and evaluate the system performance and verify the accuracy, sensitivity and specificity of the system through test sets or actual clinical data;

[0012] The diagnostic module is used to extract features, classify and predict new input ultrasound images after the system is trained, and provide diagnostic results or suggestions based on the model output;

[0013] The user interaction module is used to receive and display the diagnosis results, provide a user interface for doctors, and support further diagnosis and decision-making.

[0014] Preferably, the artificial intelligence analysis module uses a convolutional neural network as the main deep learning model and is trained using a large amount of annotated ultrasound image data.

[0015] Preferably, the feature extraction training module includes:

[0016] A data collection submodule, used for collecting diverse ultrasound image data;

[0017] The data annotation submodule is used by professional doctors to annotate the lesion area and type;

[0018] Model training submodule, used to train deep learning models based on labeled data and optimize model parameters;

[0019] The model validation submodule is used to evaluate model performance through cross-validation and optimize the model.

[0020] Preferably, the image processing module includes:

[0021] A target image determination module is used to locate the lesion area on the original ultrasound image and determine the main processing area;

[0022] An image enhancement module is used to decompose, sharpen and superimpose pixels in the high-frequency direction of the main processing area to obtain an enhanced ultrasound image;

[0023] The image magnification module is used to magnify the enhanced image to obtain a target processed ultrasonic image.

[0024] Preferably, the image enhancement module generates a sharpened image by performing multi-directional decomposition on each pixel point of the ultrasound image and then combining the decomposed pixels, and superimposes the sharpened image with the original image to enhance the details of the lesion area.

[0025] Preferably, it also includes:

[0026] An image recognition module, used to identify the clarity and integrity of the collected ultrasound images;

[0027] Image judgment module, used to analyze the reasons for image blur, including insufficient light, inaccurate viewing angle, device tilt or device malfunction;

[0028] The image adjustment module is used to issue adjustment instructions according to the analysis results of the image judgment module to optimize the acquisition parameters of the ultrasound equipment.

[0029] Preferably, the image adjustment module can issue adjustment instructions for the angle, light conditions or focal length parameters of the ultrasound device to improve the clarity and accuracy of subsequent acquired images.

[0030] Preferably, the user interaction module provides a visual interface for the diagnosis results, supporting doctors to further mark, analyze and record abnormal areas.

[0031] The advantages of this application compared with the prior art are:

[0032] 1. Improve diagnostic accuracy: AI models can identify small lesions and reduce the possibility of missed diagnosis and misdiagnosis;

[0033] 2. Shorten diagnosis time: The automated image analysis process significantly reduces doctors’ working time, enabling them to make diagnoses faster;

[0034] 3. Reduce the burden on doctors: By providing auxiliary diagnosis, it reduces the workload of doctors and improves work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a system schematic diagram of an ultrasound image analysis and diagnosis system based on artificial intelligence according to an embodiment of the present application. DETAILED DESCRIPTION

[0036] In order to make the contents of the present application easier to understand, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0037] like Figure 1 As shown, an ultrasound image analysis and diagnosis system based on artificial intelligence includes an image acquisition module, an image processing module, an artificial intelligence analysis module, a feature extraction training module, a classification and prediction module, a verification module, a diagnosis module, a user interaction module, as well as an image recognition module, an image judgment module and an image adjustment module. It can realize the whole process from image acquisition to the output of diagnosis results, and has the characteristics of automation and intelligence.

[0038] The image acquisition module uses an ultrasound probe to collect ultrasound images of the patient in real time, including image data under normal and abnormal conditions. When acquiring images, the system supports two-dimensional and three-dimensional imaging modes to ensure that detailed information of the target area is covered. The acquired images are transmitted to the image processing module through the data interface for further analysis.

[0039] The image processing module is used to pre-process the collected ultrasound images, including denoising, standardization and image enhancement operations. Specifically, the image processing module includes a target image determination module, an image enhancement module and an image magnification module; wherein the target image determination module automatically locates the lesion area of ​​the input original ultrasound image, adopts an algorithm based on edge detection and feature segmentation, quickly determines the lesion area, and marks it as the main processing area; the image enhancement module performs pixel-level enhancement on the located main processing area, calculates the components of each pixel in the area in different directions through a high-frequency directional decomposition method, generates a component image, sharpens the component image, and then performs pixel superposition with the original image to generate an enhanced ultrasound image; the image magnification module performs high-resolution magnification on the enhanced main processing area to generate a target processed ultrasound image, which is convenient for doctors to observe the details of the lesion.

[0040] In the artificial intelligence analysis module, the system uses convolutional neural network (CNN) as the main deep learning model and designs a multi-layer structure, including feature extraction layer, classification layer, etc. The model is trained with a large amount of annotated ultrasound image data and can identify and classify various lesion types.

[0041] The feature extraction training module is used to extract features from the preprocessed ultrasound image using a deep learning model, and input the extracted features into the deep learning model for training; specifically, the training process of the deep learning model by the feature extraction training module includes the following steps:

[0042] Data collection: Different types of ultrasound images were collected, including normal and abnormal cases, to ensure the diversity and representativeness of the data.

[0043] Data annotation: Professional radiologists annotate the images to identify the lesion areas and their types.

[0044] Model training: Use labeled data to train the CNN model, optimize the model parameters, and improve its ability to recognize ultrasound images.

[0045] Model validation: The performance of the model is evaluated through cross-validation and test sets to ensure its reliability in practical applications.

[0046] The classification and prediction module classifies the input ultrasound image according to the trained model and outputs the prediction result of whether there are abnormalities or lesions, as well as the specific classification label.

[0047] The verification module verifies the system through actual clinical data and test sets to evaluate its accuracy, sensitivity and specificity. The verification results are used to optimize the model or adjust the diagnostic algorithm to improve the reliability of the system.

[0048] Diagnosis module, in the system deployment stage, after a new ultrasound image is input, the model automatically extracts image features and generates diagnosis results. The diagnosis results include disease classification, risk probability or related measurement values.

[0049] The user interaction module is designed with a friendly user interface to support doctors to view the diagnosis results and provide visual analysis charts and text descriptions. Doctors can further confirm or adjust the diagnosis according to the system prompts and finally generate a diagnosis report.

[0050] In one embodiment, it also includes an image recognition, judgment and adjustment module; wherein, in the image recognition module, the system performs preliminary recognition on the collected ultrasound image to determine whether the image is clear and complete. If the image is blurred or missing, it enters the image judgment module to perform cause analysis; the image judgment module identifies the specific cause of the unclear image by analyzing factors such as light distribution, device posture and acquisition viewing angle. For example, the image may be blurred due to device tilt, insufficient light or viewing angle offset; the image adjustment module issues corresponding adjustment instructions based on the output results of the image judgment module, such as adjusting the device angle or changing the light intensity. The operator adjusts the ultrasound equipment according to the prompts to ensure that the subsequently collected images meet the diagnostic requirements.

[0051] The diagnostic process is as follows:

[0052] 1. Use the image acquisition module to obtain the patient's ultrasound image and transmit it.

[0053] 2. The image processing module preprocesses the input ultrasound image, including target positioning, image enhancement and image magnification.

[0054] 3. The image recognition module makes a preliminary judgment on the image quality, and unqualified images are optimized through the image adjustment module.

[0055] 4. The preprocessed images are input into the artificial intelligence analysis module, and the trained deep learning model is used for feature extraction and diagnosis prediction.

[0056] 5. The verification module evaluates system performance through clinical data to ensure the accuracy of diagnostic results.

[0057] 6. The diagnosis module outputs the diagnosis results and displays them to the doctor through the user interaction module. The doctor can further confirm and record the diagnosis opinions.

[0058] The above embodiments are only used to illustrate the technical solutions of the embodiments of the present application, rather than to limit them. Although the embodiments of the present application are described in detail with reference to the above embodiments, it should be understood by those skilled in the art that the technical solutions described in the above embodiments can still be modified, or some of the technical features can be replaced by equivalents, without departing from the spirit and scope defined by the claims of the present application.

Claims

1. An ultrasonic image analysis and diagnosis system based on artificial intelligence, characterized in that: include: An image acquisition module, used to acquire real-time ultrasound images, including images under normal and abnormal conditions, through an ultrasound probe, and transmit the image data to a processing unit; An image processing module, used for preprocessing the acquired ultrasound images, including denoising, standardization and image enhancement operations; An artificial intelligence analysis module, which is used to select and design deep learning models to analyze processed ultrasound images; A feature extraction training module is used to extract features from the preprocessed ultrasound image using a deep learning model, and input the extracted features into the deep learning model for training; The classification and prediction module is used to determine whether there are abnormalities or lesions in the image based on the trained features and patterns, and output the prediction results; The verification module is used to verify and evaluate the system performance and verify the accuracy, sensitivity and specificity of the system through test sets or actual clinical data; The diagnostic module is used to extract features, classify and predict new input ultrasound images after the system is trained, and provide diagnostic results or suggestions based on the model output; The user interaction module is used to receive and display the diagnosis results, provide a user interface for doctors, and support further diagnosis and decision-making.

2. The ultrasonic image analysis and diagnosis system based on artificial intelligence according to claim 1, characterized in that: The artificial intelligence analysis module uses a convolutional neural network as the main deep learning model and is trained using a large amount of annotated ultrasound image data.

3. The ultrasonic image analysis and diagnosis system based on artificial intelligence according to claim 2, characterized in that: The feature extraction training module includes: A data collection submodule, used for collecting diverse ultrasound image data; The data annotation submodule is used by professional doctors to annotate the lesion area and type; Model training submodule, used to train deep learning models based on labeled data and optimize model parameters; The model validation submodule is used to evaluate model performance through cross-validation and optimize the model.

4. The ultrasonic image analysis and diagnosis system based on artificial intelligence according to claim 1, characterized in that: The image processing module comprises: A target image determination module is used to locate the lesion area on the original ultrasound image and determine the main processing area; An image enhancement module is used to decompose, sharpen and superimpose pixels in the high-frequency direction of the main processing area to obtain an enhanced ultrasound image; The image magnification module is used to magnify the enhanced image to obtain a target processed ultrasound image.

5. The ultrasonic image analysis and diagnosis system based on artificial intelligence according to claim 4, characterized in that: The image enhancement module generates a sharpened image by performing multi-directional decomposition on each pixel point of the ultrasound image and then combining the decomposition and the sharpened image, and superimposes the image with the original image to enhance the details of the lesion area.

6. The ultrasonic image analysis and diagnosis system based on artificial intelligence according to claim 1, characterized in that: Also includes: An image recognition module, used to identify the clarity and integrity of the collected ultrasound images; Image judgment module, used to analyze the reasons for image blur, including insufficient light, inaccurate viewing angle, device tilt or device malfunction; The image adjustment module is used to issue adjustment instructions according to the analysis results of the image judgment module to optimize the acquisition parameters of the ultrasound equipment.

7. The ultrasonic image analysis and diagnosis system based on artificial intelligence according to claim 6, characterized in that: The image adjustment module can issue adjustment instructions for the angle, light conditions or focal length parameters of the ultrasound device to improve the clarity and accuracy of subsequent acquired images.

8. The ultrasonic image analysis and diagnosis system based on artificial intelligence according to claim 1, characterized in that: The user interaction module provides a visual interface for the diagnosis results, supporting doctors to further mark, analyze and record abnormal areas.

Images