Artificial intelligence-based thyroid gland scanning quality control system

Abstract

The invention relates to the technical field of ultrasonic image scanning, in particular to a thyroid scanning quality control system based on artificial intelligence, and the system comprises ultrasonic scanning equipment which is used for obtaining an ultrasonic image; the image quality evaluation module is used for evaluating the quality of the ultrasonic images based on multi-scale analysis and a deep learning model and screening the ultrasonic images to obtain screened ultrasonic images; the feature recognition and classification module receives the screened ultrasonic images and is used for extracting target features of the thyroid gland from the screened ultrasonic images and analyzing the change trend of the target features in combination with historical ultrasonic images; and the report generation module is used for generating a scanning report according to the change trend analysis result of the target characteristics. The problems of high dependence on professional physicians, complex operation, high resource requirements and poor diagnosis consistency in the prior art are solved. The method is suitable for non-professionals to use, can screen and control the quality of ultrasonic images, and remarkably improves the accuracy of thyroid examination.

Description

Technical Field

[0001] The present invention relates to the technical field of ultrasonic image scanning, and in particular to a thyroid scanning quality control system. Background Art

[0002] Traditional thyroid examinations mainly rely on ultrasonic imaging equipment and are performed by professional physicians. The ultrasound probe is connected to the ultrasound equipment, and the ultrasound equipment display screen displays real-time images. The physician diagnoses and marks thyroid nodules based on the images. However, traditional scanning methods rely on the experience and skills of professional physicians, and the operation is complicated and not suitable for non-professionals. For beginners and other non-professionals, the quality of the ultrasound images obtained is poor, and it does not have the function of screening the quality of ultrasound images. It is difficult to identify the subsequent thyroid target features for ultrasound images of poor quality, which affects the later diagnostic results. Summary of the invention

[0003] The purpose of the present invention is to provide a thyroid scan quality control system based on artificial intelligence to solve the above technical problems;

[0004] The technical problem solved by the present invention can be achieved by adopting the following technical solutions:

[0005] A thyroid scan quality control system based on artificial intelligence, comprising:

[0006] Ultrasonic scanning equipment, including an ultrasonic probe and an image acquisition module, for acquiring ultrasonic images;

[0007] An image quality assessment module, connected to the ultrasound scanning device, and used to assess the quality of the ultrasound image based on multi-scale analysis and a deep learning model and screen the ultrasound image to obtain the screened ultrasound image;

[0008] a feature recognition and classification module, connected to the image quality assessment module to receive the screened ultrasound image, for extracting target features of the thyroid gland from the screened ultrasound image, and analyzing a change trend of the target features in combination with historical ultrasound images;

[0009] A report generation module is connected to the feature recognition and classification module and is used to generate a scan report based on the change trend analysis results of the target features.

[0010] Preferably, the image quality assessment module includes a multi-scale analysis module for assessing the quality of the ultrasound image based on multi-scale analysis, and the multi-scale analysis module includes:

[0011] A multi-scale decomposition unit, used for performing multi-layer decomposition on the ultrasonic image by wavelet transform to obtain frequency components of different scales of the ultrasonic image;

[0012] A feature extraction unit, connected to the multi-scale decomposition unit, for extracting a clarity feature and a contrast feature of each of the frequency components;

[0013] The weighted evaluation unit is used to perform weighted calculation on the clarity feature and the contrast feature at each scale to evaluate the quality of the ultrasound image.

[0014] Preferably, the multi-scale decomposition unit uses discrete wavelet transform to perform a preset number of wavelet decomposition layers on the ultrasound image, and each layer of decomposition of the ultrasound image includes low-frequency components, horizontal high-frequency components, vertical high-frequency components and diagonal high-frequency components, and the wavelet decomposition of the next layer is repeated for the low-frequency components until the preset number of layers is reached.

[0015] Preferably, the feature extraction unit calculates the energy of the frequency component and uses the energy calculation result as the clarity feature. The calculation formula is:

[0016]

[0017] Among them C s represents the clarity feature of the sth layer, E s represents the energy of the sth layer, represents the high-frequency coefficient, i represents the vertical pixel position, and j represents the horizontal pixel position;

[0018] The feature extraction unit calculates the contrast feature of the frequency component using the following calculation formula:

[0019]

[0020] Among them, K s represents the contrast feature of the sth layer, max(I s ) represents the maximum pixel value of the local area of ​​the sth layer, min(I s ) represents the minimum pixel value of the local area of ​​the sth layer.

[0021] Preferably, the weighted evaluation unit is used to assign a weight coefficient to each layer, and the calculation formula of the weight coefficient is:

[0022]

[0023] ∑W s =1

[0024] Among them, W s represents the weight coefficient assigned to the sth layer, C s represents the clarity feature of the sth layer, K s represents the contrast feature of the sth layer;

[0025] The weighted calculation formula of the weighted evaluation unit for the clarity feature is:

[0026] C=∑(C s ·W s )

[0027] Wherein, C represents the weighted calculation result of the clarity feature;

[0028] The weighted calculation formula of the contrast feature by the weighted evaluation unit is:

[0029] K=∑(K s ·W s )

[0030] Wherein, K represents the weighted calculation result of the contrast feature.

[0031] Preferably, the image quality assessment module further includes an image screening unit connected to the multi-scale analysis module, and configured to screen the ultrasound image according to the weighted calculation result of the clarity feature and the weighted calculation result of the contrast feature. The image screening unit includes a scorer, configured to score the clarity and contrast of the ultrasound image. The calculation formula of the scorer for scoring the clarity of the ultrasound image is:

[0032]

[0033] Among them C p represents the clarity score of the current ultrasound image, C represents the weighted calculation result of the clarity feature of the current ultrasound image, and C min represents the weighted calculation result of the smallest clarity feature in all the ultrasound images, C max A weighted calculation result representing the maximum clarity feature in all the ultrasound images;

[0034] The calculation formula for the contrast score of the ultrasound image by the scorer is:

[0035]

[0036] Where K p represents the contrast score of the current ultrasound image, K represents the weighted calculation result of the contrast feature of the current ultrasound image, and K min represents the weighted calculation result of the smallest contrast feature in all the ultrasound images, K max A weighted calculation result representing the maximum contrast feature in all the ultrasound images;

[0037] The image screening unit (22) also includes a filter connected to the scorer, and is used to screen according to a preset clarity score C0 and a preset contrast score K0, and to set the clarity score C0 top Exceeds the preset clarity score C0 and contrast score K p The ultrasound image exceeding the preset contrast score K0 is used as the screened ultrasound image.

[0038] Preferably, the image quality assessment module also includes a metadata collection unit connected to the filter and used to collect user information and scanning information, and the filter dynamically adjusts the screening threshold according to the user information and the scanning information.

[0039] Preferably, the deep learning model adopts a trained convolutional neural network, which extracts contrast features, signal-to-noise ratio features, edge sharpness features and texture features for evaluating the quality of the ultrasound image through multi-layer convolution and pooling. The last layer of the deep learning model is a regression layer, which is used to output the quality evaluation score of the ultrasound image.

[0040] Preferably, the feature recognition and classification module extracts the target features of the thyroid gland from the screened ultrasound image, and also includes a historical data integration module connected to the feature recognition and classification module, which is used to establish a data set of the target features at each time point. The feature recognition and classification module calculates the mean and standard deviation of each target feature in the historical time series, calculates the rate of change, and outputs trend analysis results.

[0041] Preferably, it also includes,

[0042] A decision support module, connected to the report generation module, for analyzing the scan report through an artificial intelligence model to generate a clinical decision support result;

[0043] A remote collaboration platform, connected to the report generation module, used for remote consultation and collaboration with an external medical platform and outputting the scan report to the external medical platform;

[0044] An education and training module, connected to the report generation module, having a standard case library and an auxiliary learning model for scanning teaching;

[0045] A monitoring and alarm module, connected to the ultrasonic scanning device, for monitoring the scanning process of the ultrasonic scanning device (1), identifying abnormal scanning conditions and issuing an alarm prompt;

[0046] The data management module is connected to the report generation module and is used to store historical data in a data encryption, access control and audit tracking manner.

[0047] Beneficial effects of the present invention: Due to the adoption of the above technical solution, the present invention solves the problems of high reliance on professional physicians, complex operation, high resource requirements and poor diagnostic consistency in the prior art through the intelligent image quality assessment module and feature recognition and classification module. It is suitable for use by non-professionals, can screen and control the quality of ultrasound images, and significantly improve the accuracy of thyroid examination. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is an architecture diagram of a thyroid scan quality control system based on artificial intelligence in an embodiment of the present invention;

[0049] Figure 2 is a structural diagram of an image quality assessment module in an embodiment of the present invention;

[0050] Figure 3 4 is an architecture diagram of a multi-scale analysis module in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0053] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0054] An artificial intelligence-based thyroid scan quality control system, such as Figures 1 to 3 As shown, including,

[0055] Ultrasonic scanning equipment 1, including an ultrasonic probe and an image acquisition module, for acquiring ultrasonic images;

[0056] An image quality assessment module 2 is connected to the ultrasound scanning device 1, and is used to assess the quality of ultrasound images based on multi-scale analysis and a deep learning model and screen the ultrasound images to obtain screened ultrasound images;

[0057] The feature recognition and classification module 3 is connected to the image quality assessment module 2 to receive the screened ultrasound image, and is used to extract the target feature of the thyroid gland from the screened ultrasound image, and analyze the change trend of the target feature in combination with the historical ultrasound image;

[0058] The report generation module 4 is connected to the feature recognition and classification module 3 and is used to generate a scanning report based on the change trend analysis results of the target features.

[0059] Specifically, the present invention introduces an intelligent auxiliary tool through the image quality assessment module 2, which can screen the thyroid ultrasound images obtained by non-professionals, select effective thyroid ultrasound images, and enable non-professionals to perform thyroid examinations and obtain accurate results.

[0060] More specifically, the feature recognition and classification module 3 can automatically analyze ultrasound images and give analysis results based on machine learning and image processing technology, and can analyze the changing trend of target features in combination with historical ultrasound images.

[0061] The ultrasonic scanning device 1 of the present invention includes a lightweight and portable ultrasonic probe, and integrates an efficient image acquisition module, with built-in basic image processing functions, such as denoising, contrast enhancement, etc., to ensure image quality, making the ultrasonic device easy to carry and operate, suitable for use by non-professionals, and ensuring that the collected ultrasonic images are clear and accurate.

[0062] The image quality assessment module 2 is used to automatically assess the quality of ultrasound images and identify blurred, noisy or incomplete images. The present invention combines the image quality assessment module 2 with the feature recognition and classification module 3 to ensure that only high-quality images are used for thyroid lesion detection, thereby improving feature recognition accuracy.

[0063] The feature recognition and classification module 3 adopts a deep learning algorithm such as a convolutional neural network (CNN). The present invention uses a large amount of labeled data for model pre-training, automatically analyzes and extracts features from ultrasound images, automatically identifies and classifies thyroid lesions, such as nodules, cysts, calcifications, etc., and combines them with the patient's historical ultrasound images to provide dynamic change trends of the lesions. The feature recognition and classification module 3 continuously performs online learning and optimization in actual use. The report generation module 4 generates a detailed scanning report based on the analysis results, provides detailed lesion descriptions and classification information, and reduces the workload of personnel.

[0064] The system of the present invention can be applied to portable medical imaging devices, such as portable MRI or CT devices. In addition to the convolutional neural network (CNN) in the feature recognition and classification module 3, other types of machine learning algorithms can also be considered in other embodiments, such as support vector machine (SVM), random forest (RF), etc.

[0065] The present invention solves the problems of high reliance on professional physicians, complex operation, high resource requirements and poor diagnostic consistency in the prior art by providing a portable ultrasonic examination device, an intelligent image quality assessment module 2 and a feature recognition and classification module 3. The system of the present invention is particularly suitable for use by non-professionals, can screen and control the quality of ultrasonic images, and significantly improve the accuracy of thyroid examination.

[0066] In a preferred embodiment, the image quality assessment module 2 includes a multi-scale analysis module 21, which is used to assess the quality of the ultrasound image based on multi-scale analysis. The multi-scale analysis module 21 includes:

[0067] A multi-scale decomposition unit 211 is used to perform multi-layer decomposition on the ultrasonic image through wavelet transform to obtain frequency components of different scales of the ultrasonic image;

[0068] A feature extraction unit 212, connected to the multi-scale decomposition unit 211, for extracting a clarity feature and a contrast feature of each frequency component;

[0069] The weighted evaluation unit 213 is used to perform weighted calculation on the clarity feature and the contrast feature at each scale to evaluate the quality of the ultrasound image.

[0070] Specifically, the present invention evaluates the quality of ultrasound images through multi-scale analysis, analyzes ultrasound image signals at different scales and resolutions to capture features at different levels in the ultrasound image, helps identify image details, texture and structural information, and evaluates quality indicators of image clarity and contrast.

[0071] In a preferred embodiment, the multi-scale decomposition unit 211 uses discrete wavelet transform to perform a preset number of wavelet decompositions on the ultrasound image. Each layer of the ultrasound image decomposition includes low-frequency components, horizontal high-frequency components, vertical high-frequency components, and diagonal high-frequency components. The wavelet decomposition of the next layer is repeated for the low-frequency components until the preset number of layers is reached.

[0072] Specifically, the present invention uses an ultrasonic device to obtain an original ultrasonic image, performs basic image preprocessing, such as denoising, normalization, and contrast enhancement, and applies a wavelet transform (such as a discrete wavelet transform) to decompose the image. The wavelet transform can decompose the image into different frequency components, each frequency component corresponds to a different scale of the image, the low-frequency component contains the overall rough structure of the image, and the high-frequency component contains more detailed detail information.

[0073] In a preferred embodiment, the feature extraction unit 212 calculates the energy of the frequency component and uses the energy calculation result as the clarity feature. The calculation formula is:

[0074]

[0075] Among them C s represents the clarity feature of the sth layer, E s represents the energy of the sth layer, represents the high-frequency coefficient, i represents the vertical pixel position, and j represents the horizontal pixel position;

[0076] The feature extraction unit 212 calculates the contrast feature of the frequency component using the following formula:

[0077]

[0078] Among them, K s represents the contrast feature of the sth layer, max(I s ) represents the maximum pixel value of the local area of ​​the sth layer, min(I s ) represents the minimum pixel value of the local area of ​​the sth layer.

[0079] Specifically, the present invention can judge the clarity of an image by analyzing the intensity and changes of high-frequency components. The richer the high-frequency components, the clearer the image. The contrast is evaluated by calculating the pixel difference between pixels in different areas.

[0080] First, after preliminary filtering of the image, discrete wavelet transform (DWT) is used to perform wavelet decomposition on the image, and the image is decomposed into four parts: low frequency (LL), horizontal high frequency (LH), vertical high frequency (HL), and diagonal high frequency (HH).

[0081] Among them, the low-frequency component (LL) contains the overall structure and contour information of the image and represents the general brightness changes of the image.

[0082] The horizontal high frequency component (LH) contains the details of the horizontal edges in the image.

[0083] The vertical high frequency component (HL) contains the details of the vertical edges in the image.

[0084] The diagonal high frequency components (HH) contain details of diagonal edges in the image.

[0085] The low frequency part (LL) is subjected to wavelet transformation again, and this step is repeated. The number of decomposition layers can be set according to the requirements, and in the present invention, 2 to 3 layers of decomposition are generally used.

[0086] The clarity is often proportional to the intensity and number of high-frequency components. If the absolute value of the high-frequency coefficient is large, it means that there are more details and clear edges in the image, so the image appears clearer. By calculating the energy and variance of the high-frequency part, the clarity of the image can be quantitatively evaluated. High energy and large variance usually indicate that the image contains rich details. It can be calculated using the following formula.

[0087]

[0088] Among them C s represents the clarity feature of the sth layer, E s represents the energy of the sth layer, represents the high-frequency coefficient, i represents the vertical pixel position, and j represents the horizontal pixel position;

[0089] In the low-frequency and high-frequency parts, the local contrast is calculated. The definition of local contrast can be used,

[0090]

[0091] Among them, K s represents the contrast feature of the sth layer, max(I s ) represents the maximum pixel value of the local area of ​​the sth layer, min(I s ) represents the minimum pixel value of the local area of ​​the sth layer.

[0092] In a preferred embodiment, the weighted evaluation unit 213 is used to assign a weight coefficient to each layer. The calculation formula of the weight coefficient is:

[0093]

[0094] ∑W s =1

[0095] Among them, W s represents the weight coefficient assigned to the sth layer, C s represents the clarity feature of the sth layer, K s represents the contrast feature of the sth layer;

[0096] The weighted calculation formula for the clarity feature by the weighted evaluation unit 213 is:

[0097] C=∑(C s ·W s )

[0098] Wherein, C represents the weighted calculation result of the clarity feature;

[0099] The weighted calculation formula for the contrast feature by the weighted evaluation unit 213 is:

[0100] K=∑(K s ·W s )

[0101] Wherein, K represents the weighted calculation result of the contrast feature.

[0102] In a preferred embodiment, the image quality assessment module 2 further includes an image screening unit 22, connected to the multi-scale analysis module 21, for screening the ultrasound image according to the weighted calculation result of the clarity feature and the weighted calculation result of the contrast feature. The image screening unit 22 includes a scorer for scoring the clarity and contrast of the ultrasound image. The scorer calculates the clarity score of the ultrasound image using the following formula:

[0103]

[0104] Among them C p represents the clarity score of the current ultrasound image, C represents the weighted calculation result of the clarity feature of the current ultrasound image, and C min Represents the weighted calculation result of the smallest clarity feature in all ultrasound images, C max Represents the weighted calculation result of the maximum clarity feature in all ultrasound images;

[0105] The calculation formula of the contrast score of ultrasound images by the scorer is:

[0106]

[0107] Where K p represents the contrast score of the current ultrasound image, K represents the weighted calculation result of the contrast feature of the current ultrasound image, and K min Represents the weighted calculation result of the smallest contrast feature in all ultrasound images, K max Represents the weighted calculation result of the maximum contrast feature in all ultrasound images;

[0108] The image screening unit 22 also includes a filter connected to the scorer for screening according to a preset clarity score C0 and a preset contrast score K0, and p Exceeds the preset clarity score C0 and contrast score K p Ultrasound images exceeding a preset contrast score K0 are used as screened ultrasound images.

[0109] Specifically, the present invention summarizes the clarity and contrast indicators of each scale, adopts a weighted average method for comprehensive evaluation, sets the clarity and contrast thresholds according to actual applications, and adopts the image screening unit 22 to automatically screen images.

[0110] The present invention can effectively integrate the clarity and contrast information of each scale in the multi-scale analysis to obtain a comprehensive image quality evaluation, thereby ensuring the accuracy of screening.

[0111] In a preferred embodiment, the image quality assessment module 2 also includes a metadata collection unit 23 connected to a filter for collecting user information and scanning information, and the filter dynamically adjusts the screening threshold according to the user information and scanning information.

[0112] Specifically, during the image acquisition process, the metadata recorded by the present invention includes basic information of the patient, such as age, gender, medical history, etc., and scanning information, such as ultrasound equipment model, probe type, examination environment, etc. The screening threshold is dynamically adjusted according to the collected metadata. For example, for elderly patients, the contrast threshold needs to be lowered to adapt to their physiological characteristics.

[0113] In a preferred embodiment, the deep learning model uses a trained convolutional neural network, which extracts contrast features, signal-to-noise ratio features, edge sharpness features and texture features for evaluating the quality of ultrasound images through multiple layers of convolution and pooling. The last layer of the deep learning model is a regression layer, which is used to output the quality evaluation score of the ultrasound image.

[0114] Specifically, the deep learning model adopts a pre-trained convolutional neural network model, such as Inception or EfficientNet, and is trained on a large data set, such as ImageNet. It has good feature extraction capabilities and can effectively capture multiple features in ultrasound images. The present invention uses a practical deep learning framework to load the pre-trained deep learning model.

[0115] The input layer of the deep learning model is used to receive the filtered ultrasound image as input.

[0116] The multi-layer convolutional layers of the deep learning model are used to extract image features. Each convolution layer is followed by an activation function (ReLU) and a pooling layer. The convolution layer extracts contrast features, signal-to-noise ratio features, edge sharpness features, and texture features. After multiple layers of convolution and pooling, the features are flattened and input to the fully connected layer for higher-level feature combination.

[0117] A regression layer is set at the last layer of the model, and a linear activation function is used to output the image quality score. The regression layer can be weighted according to the previously extracted features and output a continuous value representing the image quality evaluation score.

[0118] In a preferred embodiment, the feature recognition and classification module 3 extracts the target features of the thyroid gland from the screened ultrasound image, and also includes a historical data integration module connected to the feature recognition and classification module 3, which is used to establish a data set of target features at each time point. The feature recognition and classification module 3 calculates the mean and standard deviation of each target feature in the historical time series, calculates the rate of change, and outputs trend analysis results.

[0119] Specifically, the present invention collects all historical ultrasound images and their feature data, extracts relevant features from each image, and establishes a feature data set for each time point. The time series analysis method is used to calculate the mean and standard deviation of each feature in the historical time series, and the change rate is calculated based on the difference between the current feature and the historical mean, where the change rate = (current feature - historical mean) / historical mean × 100%.

[0120] The present invention plots the historical characteristic values, mean, standard deviation and change rate into a time series graph to visualize the characteristic change trend. A threshold is set based on clinical experience or statistical results to determine whether the characteristic change is significant. For example, when the change rate exceeds a certain percentage, the lesion is considered to have progressed. The change trend analysis is combined with the statistical analysis results to comprehensively evaluate the progress or improvement of the lesion. The report generation module 4 generates a scan report to display the change trend, statistical results and clinical recommendations for the doctor's reference.

[0121] In a preferred embodiment, it also includes:

[0122] The decision support module 5 is connected to the report generation module 4 and is used to analyze the scan report through an artificial intelligence model to generate clinical decision support results;

[0123] The remote collaboration platform 6 is connected to the report generation module 4 and is used for remote consultation and collaboration with an external medical platform and outputs a scan report to the external medical platform;

[0124] Education and training module 7, connected to report generation module 4, has a standard case library and auxiliary learning model for scanning teaching;

[0125] The monitoring and alarm module 8 is connected to the ultrasonic scanning device 1 and is used to monitor the scanning process of the ultrasonic scanning device 1, identify abnormal scanning conditions and issue an alarm prompt;

[0126] The data management module 9 is connected to the report generation module 4 and is used to store historical data in a manner of data encryption, access control and audit tracking.

[0127] Specifically, the decision support module 5 is used to provide clinical decision support, such as recommending further examination plans to generate clinical decision support results; it can provide personalized auxiliary treatment suggestions based on the thyroid feature identification results to assist doctors in completing clinical decisions.

[0128] The remote collaboration platform 6 is a service platform that supports remote consultation and expert collaboration, facilitating communication and discussion between doctors. The present invention utilizes video conferencing technology and a collaboration platform to achieve remote consultation and expert collaboration, and realizes data transmission and sharing in remote consultation through data sharing and synchronization mechanisms. It also introduces virtual reality technology and collaboration tools to provide an immersive remote consultation experience and improve the effect of expert collaboration. The collaboration tools provide, for example, real-time marking, annotation, and discussion functions to facilitate communication and discussion between doctors.

[0129] The system of the present invention adopts a graphical user interface (GUI), and the education and training module 7 provides clear operation instructions and feedback, and designs standardized operation steps. The user only needs to follow the prompts to complete the inspection, which reduces the learning curve and enables non-professionals to quickly get started and use it.

[0130] In addition to the graphical user interface, the present invention also uses a voice user interface (VUI) to simplify user operations by operating through voice commands.

[0131] The education and training module 7 establishes a standardized thyroid ultrasound image case library and provides a platform for online learning and training. The education and training module 7 of the present invention uses AI technology to provide real-time feedback and case analysis to help doctors improve their diagnostic skills. At the same time, it provides a virtual patient simulation function for actual combat exercises and skill training. It can make learning paths and recommendations based on the user's learning progress to improve the learning effect in a personalized way. At the same time, it provides online examination and evaluation functions to help users test their learning effects and conduct skill evaluations.

[0132] The monitoring and alarm module 8 monitors the ultrasonic inspection process in real time, identifies abnormal situations and issues alarm prompts, and uses real-time monitoring technology to monitor the image quality and operation specifications during the ultrasonic inspection process. The present invention identifies abnormal situations through sensors, such as the hand posture, movement direction, and angle of holding the probe of the scanning personnel. The sensors include acceleration sensors, gyroscopes, and optical sensors. The present invention uses machine learning models to extract and identify features such as hand posture and holding angle from the data collected by the sensors, and combines with the abnormal detection model to identify abnormal situations by setting thresholds or using classification algorithms, such as improper hand position, irregular probe angle, etc., and immediately issues alarm prompts to help operators correct problems in a timely manner.

[0133] The present invention can also adaptively adjust monitoring parameters according to different patients and examination conditions, and provide a multi-level alarm mechanism, which issues different levels of alarm prompts according to the severity of abnormal conditions, helping operators to prioritize important issues. It also provides real-time feedback on image quality during image acquisition and prompts operators to make adjustments immediately.

[0134] The data management module 9 uses cloud storage or local storage to safely and reliably save the scanned data, and uses data encryption, access control and audit tracking to ensure the security and privacy of user data. The present invention uses data encryption technology and access control mechanism to protect the privacy and security of data. Data encryption can use differential privacy and homomorphic encryption. The access control mechanism dynamically adjusts data access rights according to user roles and operation requirements to ensure the principle of minimum permissions for data use. Through audit tracking and log management, the access and operation of data are recorded to ensure compliance and transparency of data use. The present invention also introduces blockchain technology to ensure data transmission security and non-tamperability. The above is only a preferred embodiment of the present invention, and does not limit the implementation method and protection scope of the present invention. For those skilled in the art, it should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the specification and diagrams of the present invention should be included in the protection scope of the present invention.

Claims

1. A thyroid scan quality control system based on artificial intelligence, characterized in that: include, An ultrasonic scanning device (1), comprising an ultrasonic probe and an image acquisition module, for acquiring an ultrasonic image; An image quality assessment module (2) connected to the ultrasound scanning device (1), the image quality assessment module (2) being used to assess the quality of the ultrasound image based on multi-scale analysis and a deep learning model and to screen the ultrasound image to obtain the screened ultrasound image; A feature recognition and classification module (3) is connected to the image quality assessment module (2), receives the screened ultrasound image, is used to extract the target feature of the thyroid gland from the screened ultrasound image, and analyzes the change trend of the target feature in combination with historical ultrasound images; A report generation module (4) is connected to the feature recognition and classification module (3) and is used to generate a scanning report based on the change trend analysis results of the target features.

2. The artificial intelligence-based thyroid scan quality control system according to claim 1, characterized in that: The image quality assessment module (2) comprises a multi-scale analysis module (21) for assessing the quality of the ultrasound image based on multi-scale analysis. The multi-scale analysis module (21) comprises: A multi-scale decomposition unit (211), used for performing multi-layer decomposition on the ultrasonic image by wavelet transform to obtain frequency components of different scales of the ultrasonic image; A feature extraction unit (212), connected to the multi-scale decomposition unit (211), used to extract a clarity feature and a contrast feature of each of the frequency components; The weighted evaluation unit (213) is used to perform weighted calculation on the clarity feature and the contrast feature at each scale to evaluate the quality of the ultrasound image.

3. The artificial intelligence-based thyroid scan quality control system according to claim 2, characterized in that: The multi-scale decomposition unit (211) uses discrete wavelet transform to perform a preset number of wavelet decompositions on the ultrasonic image, wherein each layer of the ultrasonic image decomposition includes low-frequency components, horizontal high-frequency components, vertical high-frequency components and diagonal high-frequency components, and the wavelet decomposition of the next layer is repeated for the low-frequency components until the preset number of layers is reached.

4. The artificial intelligence-based thyroid scan quality control system according to claim 2, characterized in that: The feature extraction unit (212) calculates the energy of the frequency component and uses the energy calculation result as the clarity feature. The calculation formula is: Among them C s represents the clarity feature of the sth layer, E s represents the energy of the sth layer, represents the high-frequency coefficient, i represents the vertical pixel position, and j represents the horizontal pixel position; The feature extraction unit (212) calculates the contrast feature of the frequency component using the following calculation formula: Among them, K s represents the contrast feature of the sth layer, max(I s ) represents the maximum pixel value of the local area of ​​the sth layer, min(I s ) represents the minimum pixel value of the local area of ​​the sth layer.

5. The artificial intelligence-based thyroid scan quality control system according to claim 4, characterized in that: The weighted evaluation unit (213) is used to assign a weight coefficient to each layer. The calculation formula of the weight coefficient is: ∑W s ＝1 Among them, W s represents the weight coefficient assigned to the sth layer, C s represents the clarity feature of the sth layer, K s represents the contrast feature of the sth layer; The weighted calculation formula of the weighted evaluation unit (213) for the clarity feature is: C=∑(C s ·W s ) Wherein, C represents the weighted calculation result of the clarity feature; The weighted calculation formula of the contrast feature by the weighted evaluation unit (213) is: K=∑(K s ·W s ) Wherein, K represents the weighted calculation result of the contrast feature.

6. The artificial intelligence-based thyroid scan quality control system according to claim 5, characterized in that: The image quality assessment module (2) further comprises an image screening unit (22), connected to the multi-scale analysis module (21), and configured to screen the ultrasound image according to a weighted calculation result of a clarity feature and a weighted calculation result of a contrast feature. The image screening unit (22) comprises a scorer, configured to perform a clarity score and a contrast score on the ultrasound image. The scorer calculates the clarity score of the ultrasound image using the following formula: Among them C p represents the clarity score of the current ultrasound image, C represents the weighted calculation result of the clarity feature of the current ultrasound image, and C min represents the weighted calculation result of the smallest clarity feature in all the ultrasound images, C max A weighted calculation result representing the maximum clarity feature in all the ultrasound images; The calculation formula for the contrast score of the ultrasound image by the scorer is: Where K p represents the contrast score of the current ultrasound image, K represents the weighted calculation result of the contrast feature of the current ultrasound image, and K min represents the weighted calculation result of the smallest contrast feature in all the ultrasound images, K max A weighted calculation result representing the maximum contrast feature in all the ultrasound images; The image screening unit (22) also includes a filter connected to the scorer, and is used to screen according to a preset clarity score C0 and a preset contrast score K0, and to set the clarity score C0 to p Exceeds the preset clarity score C0 and contrast score K p The ultrasound image exceeding the preset contrast score K0 is used as the screened ultrasound image.

7. The artificial intelligence-based thyroid scan quality control system according to claim 6, characterized in that: The image quality assessment module (2) also includes a metadata collection unit (23) connected to the filter and used to collect user information and scanning information, and the filter dynamically adjusts the screening threshold according to the user information and the scanning information.

8. The artificial intelligence-based thyroid scan quality control system according to claim 1, characterized in that: The deep learning model uses a trained convolutional neural network, which extracts contrast features, signal-to-noise ratio features, edge sharpness features and texture features for evaluating the quality of the ultrasound image through multi-layer convolution and pooling. The last layer of the deep learning model is a regression layer, which is used to output the quality evaluation score of the ultrasound image.

9. The artificial intelligence-based thyroid scan quality control system according to claim 1, characterized in that: The feature recognition and classification module (3) extracts the target feature of the thyroid gland from the screened ultrasound image, and also includes a historical data integration module connected to the feature recognition and classification module (3) for establishing a data set of the target feature at each time point. The feature recognition and classification module (3) calculates the mean and standard deviation of each target feature in the historical time series, calculates the rate of change, and outputs a trend analysis result.

10. The artificial intelligence-based thyroid scan quality control system according to claim 1, characterized in that: Also includes, A decision support module (5), connected to the report generation module (4), for analyzing the scan report through an artificial intelligence model to generate a clinical decision support result; A remote collaboration platform (6), connected to the report generation module (4), used for remote consultation and collaboration with an external medical platform and outputting the scan report to the external medical platform; An education and training module (7), connected to the report generation module (4), having a standard case library and an auxiliary learning model for scanning teaching; A monitoring and alarm module (8), connected to the ultrasonic scanning device (1), used to monitor the scanning process of the ultrasonic scanning device (1), identify abnormal scanning conditions and issue an alarm prompt; The data management module (9) is connected to the report generation module (4) and is used to store historical data in a data encryption, access control and audit tracking manner.

Images