Ultrasonic image classification enhancement method and system and storage medium

Abstract

The invention discloses an ultrasonic image classification enhancement method and system and a storage medium, and relates to the technical field of medical ultrasonic image processing, and the method comprises the steps of ultrasonic image optimization, key structure labeling, key frame selection and ultrasonic image classification enhancement. Data collection processing and image quality factor extraction are carried out on an original ultrasonic image, multi-factor fusion classification evaluation is carried out, and real-time parameter feedback type automatic optimization of the ultrasonic image is realized; performing accurate segmentation, key point fitting and structural integrity evaluation on a target structure by adopting a structural marking point layout method combining morphological perception prediction and structural evaluation; in the key frame selection stage, sequential structure stability analysis and image registration improvement are combined, comprehensive scoring is performed on continuous frame images, and key frame automatic selection in a stable state is realized; the method can effectively improve the ultrasonic image quality stability, the structure labeling accuracy and the key frame selection reliability, and is suitable for real-time ultrasonic imaging and auxiliary diagnosis scenes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ultrasound image processing, in particular to an ultrasound image classification enhancement method, system and storage medium. BACKGROUND

[0002] The ultrasound image classification enhancement method refers to a system tool that uses ultrasound imaging technology to obtain image information of an ultrasound detection target, and combines image processing, structure recognition and intelligent analysis to assist in judging the classification of the image. The system can help doctors observe the key structures of the target (such as the heart, limbs, etc.) more clearly, improve the accuracy of section recognition, measurement and diagnosis efficiency, and is particularly suitable for routine screening and abnormal detection, and has the functions of improving diagnosis quality, reducing the burden of doctors and enhancing the standardization of operation.

[0003] However, in the existing ultrasound image processing method, there are problems of unstable image quality, image blur and inability to meet the feature extraction accuracy of the processing result;

[0004] In the existing ultrasound image optimization enhancement method, there are problems of low efficiency in the optimization process and long processing time;

[0005] In the existing target structure recognition and labeling process, the traditional ultrasound image structure labeling method relies on manual labeling or simple image processing algorithm, which often leads to low labeling accuracy, especially in the recognition of complex geometric shapes or weak boundary structures, which is prone to errors;

[0006] In the existing key frame selection process of the ultrasound image, due to the movement of the target or improper operation, the optimal frame image is often blurred or unable to capture the stable state of the target structure, which causes great difficulty for subsequent analysis. SUMMARY

[0007] In view of the above problems, in order to overcome the defects of the prior art, the present application provides an ultrasound image classification enhancement method, system and storage medium. The technical scheme adopted by the present application is as follows: the ultrasound image classification enhancement method provided by the present application comprises the following steps:

[0008] Step S1: ultrasound image optimization;

[0009] Step S2: key structure labeling;

[0010] Step S3: key frame optimization based on time sequence stability;

[0011] Step S4: ultrasound image classification enhancement.

[0012] Further, in step S1, the ultrasound image is optimized for obtaining original ultrasound images and pre-processing and image enhancement, specifically, through ultrasound image acquisition and basic preprocessing, original ultrasound image data is obtained, and through the ultrasound image intelligent optimization scheme combining quality factor fusion and parameter reverse mapping, real-time parameter feedback type optimization of ultrasound image is carried out, and optimized ultrasound image data is obtained, including the following steps:

[0013] Step S11: data collection processing, specifically, through continuous image frame acquisition, the target ultrasound original image is collected, and the basic enhanced image data is obtained through image basic enhancement processing;

[0014] Step S12: image quality factor extraction, specifically, according to the basic enhanced image data, the image quality factor is extracted by sequentially performing signal-to-noise ratio estimation, contrast estimation and boundary sharpness estimation, and the image quality factor parameter set is obtained; The image quality factor parameter set specifically includes signal-to-noise ratio parameter, contrast parameter and boundary sharpness parameter;

[0015] The signal-to-noise ratio estimation is specifically performed by image frequency energy ratio method;

[0016] The contrast estimation is specifically performed by global contrast measurement;

[0017] The boundary sharpness estimation is specifically performed by Laplacian gradient variance method;

[0018] Step S13: multi-factor fusion classification evaluation, specifically, according to the image quality factor parameter set, the multi-factor fusion image quality screening threshold is obtained by parameter index normalization and comprehensive image quality weighted calculation, and the image classification is performed on the basic enhanced image data according to the multi-factor fusion image quality screening threshold, and the image classification state label is obtained;

[0019] The parameter index normalization is specifically performed by maximum and minimum normalization calculation according to the image quality factor parameter set to obtain the normalized image quality parameter, and the multi-factor fusion image quality screening threshold is obtained by comprehensive image quality weighted calculation according to the normalized image quality parameter, and the calculation formula is:

[0020] ;

[0021] In the formula, Q total is the multi-factor fusion image quality screening threshold, is the signal-to-noise ratio weight, is the normalized signal-to-noise ratio parameter, is the contrast weight, is the normalized contrast parameter, is a boundary sharpness weight, is a normalized boundary sharpness parameter;

[0022] Step S14: parameter optimization suggestion generation, specifically, according to the image classification state label, constructing an image optimization parameter mapping table, and performing image optimization parameter matching based on a rule-based method to obtain parameter optimization suggestion data;

[0023] Step S15: automatic optimization of ultrasound images, specifically, according to the parameter optimization suggestion data, performing image optimization operations to automatically optimize the basic enhanced image data to obtain optimized ultrasound image data.

[0024] Further, in step S2, the key structure labeling is used to identify and label the key structure of the ultrasound image subject, specifically, according to the optimized ultrasound image data, using an ultrasound image structure labeling point layout method combining morphological perception prediction and structure evaluation to perform key structure labeling, to obtain structure labeling image data, including the following steps:

[0025] Step S21: structure segmentation, specifically, using a lightweight semantic segmentation model, according to the optimized ultrasound image data, performing structure segmentation, and using an edge extraction algorithm to extract structure edges to obtain structure segmentation basic data;

[0026] The lightweight semantic segmentation model specifically uses an improved DeepLabv3+ model; the edge extraction algorithm specifically uses a boundary tracking algorithm improved by introducing a direction gradient constraint;

[0027] The improved DeepLabv3+ model specifically replaces the standard ResNet backbone with a lightweight MobileNetV3 backbone network, and increases a structure perception module in the decoder and integrates a joint attention module in the spatial pyramid pooling layer to perform lightweight improvement of the semantic segmentation model;

[0028] Step S22: key structure point fitting, specifically, by constructing a structure instance object, and by performing geometric fitting and key point prediction on the instance object, to obtain structure key point pair data;

[0029] Step S23: structure labeling point value generation, specifically, according to the structure key point pair data, calling a drawing function to generate structure point values and cursors to obtain structure labeling measurement values;

[0030] Step S24: structure integrity evaluation, specifically, performing weighted summation on image quality evaluation reference data corresponding to each structure key point pair data to obtain structure completeness score reference data, which is used to assist key frame selection;

[0031] The calculation formula of the structure completeness reference score is:

[0032]

[0033] In the formula, Q f is the structure completeness reference score, is the structure fitting error weight, Q shape is the structure fitting error parameter, is the structure region gradient variance weight, Q clarity is the structure region gradient variance parameter, is the symmetry score weight, Q symmetry is the fitting symmetry score parameter.

[0034] Step S25: key structure labeling, specifically, according to the structure labeling measurement value and the structure completeness reference data, key structure labeling is performed on each structure instance object to obtain structure labeling image data.

[0035] Further, in step S3, the key frame selection is used to automatically identify the standard target observation section and automatically select the key frame at the relative stable position of the target structure for selection. Specifically, according to the structure labeling image data, an improved ultrasound image key frame selection determination method combining time sequence structure stability analysis and image registration is used for key frame selection to obtain key frame image optimization data, including the following steps:

[0036] Step S31: structure trajectory tracking, specifically, according to the structure labeling image data, structure gravity calculation and structure motion trajectory recording are sequentially performed to obtain instance structure trajectory tracking data;

[0037] Step S32: structure stability score, specifically, by sequentially calculating the translation amplitude score, the contour deformation score and the image texture change score, structure stability weighted score is performed to obtain comprehensive frame selection score reference data;

[0038] Step S33: selection judgment prompt, specifically, by setting a frame selection threshold, according to the comprehensive frame selection score reference data, frame selection judgment is performed, and combined with the image quality evaluation reference data, image frame selection prompt is generated in real time;

[0039] Step S34: motion blur registration improvement, specifically, by detecting the structure of the current frame and the previous frame, a motion blur compensation method combined with result region registration enhancement is used for registration improvement to obtain registration optimization frame selection image data;

[0040] ​Step S35: key frame selection, specifically, according to the image frame selection prompt, semi-automatic image frame selection is performed, and the cut surface instantaneous frame image is optimized by optimizing the frame image data according to the registration, to obtain the frame image optimization data.

[0041] Further, in step S4, the ultrasound image classification enhancement is enhanced for combining the optimized ultrasound image, the structure annotation information and the frame image optimization data to assist the doctor in ultrasound image classification judgment, specifically, according to the optimized ultrasound image data, the structure annotation image data and the frame image optimization data, the ultrasound image classification enhancement is performed by sequentially analyzing the morphological parameters and the image parameters, to obtain the ultrasound image comprehensive classification reference data.

[0042] The application provides an ultrasound image classification enhancement system, which comprises an ultrasound image optimization module, a key structure annotation module, a key frame selection module and an image classification enhancement module.

[0043] The ultrasound image optimization module is used for ultrasound image optimization, and the optimized ultrasound image data is obtained through the ultrasound image optimization, and the optimized ultrasound image data is sent to the key structure annotation module and the image classification enhancement module.

[0044] The key structure annotation module is used for key structure annotation, and the structure annotation image data is obtained through the key structure annotation, and the structure annotation image data is sent to the key frame selection module and the image classification enhancement module.

[0045] The key frame selection module is used for key frame selection, and the frame image optimization data is obtained through the key frame selection, and the frame image optimization data is sent to the image classification enhancement module.

[0046] The image classification enhancement module is used for image classification enhancement, and the diagnostic auxiliary measurement parameter data is obtained through the image classification enhancement.

[0047] The application has the following beneficial effects by adopting the above scheme:

[0048] (1) In the existing ultrasound image processing method, there are problems of unstable image quality, blurred image and processing result cannot meet the feature extraction accuracy. Specifically, the traditional ultrasound imaging data often has the problems of blur and noise interference. In dynamic scanning image, blur or noise caused by motion often occurs, which leads to difficulty in structure recognition of target object. The scheme creatively adopts a comprehensive intelligent method combining image quality factor intelligent classification, key area perception optimization and time sequence key frame optimization support. Through the introduction of intelligent evaluation and real-time optimization of image quality factors (signal-to-noise ratio, contrast and boundary definition), the scheme can dynamically adjust the image optimization parameters according to real-time feedback, so that the image can maintain high quality in different imaging environments, thereby greatly improving the accuracy and reliability of detection;

[0049] (2) In the existing ultrasound image optimization enhancement method, there are problems of low efficiency in optimization process and long processing time. In the traditional scheme, ultrasound image optimization usually relies on static parameters for post-processing, which is relatively cumbersome and slow. This often cannot meet the real-time flow processing requirements in actual application, especially in high-throughput scanning or dynamic continuous monitoring. The scheme creatively adopts an ultrasound image intelligent optimization scheme combining quality factor fusion and parameter reverse mapping for real-time parameter feedback optimization of ultrasound image, so that the ultrasound image optimization process is more efficient, and high-quality images can be obtained in a short time, ensuring fast and accurate data processing support;

[0050] (3) In the existing target structure recognition and labeling process, the traditional ultrasound image structure labeling method relies on manual labeling or simple image processing algorithm, which often leads to low labeling accuracy, especially in the recognition of complex geometric shapes or weak boundary structures. Specifically, the labeling of low-contrast areas or overlapping tissues is easily affected by image quality and target position changes, causing inaccurate or missed labeling. The scheme creatively adopts an ultrasound image structure labeling point layout method combining morphological perception prediction and structure evaluation to label key features, which can more accurately locate and label the key structure of the detected target, thereby providing more reliable basic data for subsequent quantitative analysis;

[0051] (4) In the existing key frame selection process of ultrasound images, due to the movement of the target or improper operation, the preferred frame image is often blurred or the stable state of the target structure cannot be captured, which causes great difficulty in subsequent analysis. Specifically, a small change in the position of the target may cause the image selection to fail to align with the most stable part of the observed subject, affecting the calculation and judgment of the system. The present scheme creatively adopts an improved ultrasound image key frame selection method combining time sequence structure stability analysis and image registration to lock the time sequence key frame, which can accurately judge and select the best section when the target is in a stable position for locking, ensure that the image is clear and meets the detection requirements, and at the same time reduce the interference caused by motion blur, improve the processing accuracy and efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 A flowchart of an ultrasound image classification enhancement method provided by the present application is shown in the figure.

[0053] Figure 2 A schematic diagram of an ultrasound image classification enhancement system provided by the present application is shown in the figure.

[0054] Figure 3 A flowchart of step S1 ultrasound image optimization is shown in the figure.

[0055] Figure 4 A flowchart of step S2 key structure labeling is shown in the figure.

[0056] Figure 5 A flowchart of step S3 key frame selection is shown in the figure.

[0057] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation on the present application. DETAILED DESCRIPTION

[0058] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0059] In the description of the present application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0060] Embodiment one, refer to Figure 1 The present application provides an ultrasound image classification enhancement method, which comprises the following steps:

[0061] Step S1: ultrasound image optimization;

[0062] Step S2: key structure labeling;

[0063] Step S3: key frame selection;

[0064] Step S4: ultrasound image classification enhancement.

[0065] By performing the above operation, in the existing ultrasound image processing method, there are problems of unstable image quality, image blur and processing results that cannot meet the feature extraction accuracy. Specifically, the traditional ultrasound imaging data often has the problems of blur and noise interference. In dynamic scanning images, blur or noise caused by motion often occurs, which leads to difficulty in recognizing the structure of the target object. The present scheme creatively adopts a comprehensive intelligent method combining image quality factor intelligent classification, key region perception optimization and time sequence key frame optimization support. By introducing intelligent evaluation and real-time optimization of image quality factors (signal-to-noise ratio, contrast and boundary definition), the present scheme can dynamically adjust the image optimization parameters according to real-time feedback, so that the image can maintain high quality in different imaging environments, thereby greatly improving the accuracy and reliability of detection.

[0066] Embodiment two, refer to Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S1, the ultrasound image optimization is used to obtain the original ultrasound image and perform preprocessing and image enhancement. Specifically, through ultrasound image acquisition and basic preprocessing, the original ultrasound image data is obtained, and through the ultrasound image intelligent optimization scheme combining quality factor fusion and parameter reverse mapping, real-time parameter feedback type optimization of ultrasound image is performed to obtain optimized ultrasound image data, comprising the following steps:

[0067] Step S11: data collection and processing, specifically, through continuous image frame acquisition, the target ultrasound original image is collected and obtained, and image basic enhancement processing is performed to obtain basic enhanced image data;

[0068] Step S12: Image quality factor extraction, specifically, based on the basic enhanced image data, image quality factors are extracted by sequentially performing signal-to-noise ratio estimation, contrast estimation, and boundary sharpness estimation to obtain an image quality factor parameter set; the image quality factor parameter set specifically includes signal-to-noise ratio parameters, contrast parameters, and boundary sharpness parameters;

[0069] The signal-to-noise ratio (SNR) estimation is specifically performed using the image frequency domain energy ratio method, and the calculation formula is as follows:

[0070] ;

[0071] In the formula, Q SNR Here, is the signal-to-noise ratio parameter, u is the horizontal coordinate in the frequency domain, used to represent the frequency domain representation of horizontal pixels in the image, and v is the vertical coordinate in the frequency domain. It is a set of effective frequency domain region indices used to represent the main content of the image. It is the Fourier transform result of the image, used to represent the complex representation of the image in the frequency domain. It is a set of noise frequency domain region indices used to represent high-frequency edges. It is the local minimum constant, specifically taking the value 10. -8 ;

[0072] The contrast estimation is specifically performed through global contrast measurement, and the calculation formula is as follows:

[0073] ;

[0074] In the formula, Q contrast It's a contrast parameter. is the standard deviation of image pixel values, used to represent the degree of dispersion of image gray levels; I is the image identifier. It is the average pixel value of the image, used to represent the average brightness;

[0075] Boundary sharpness estimation is specifically performed using the Laplacian gradient variance method, and the calculation formula is as follows:

[0076] ;

[0077] In the formula, Q sharpness It is the boundary sharpness parameter, and Var is the variance function. The overall value is the Laplacian operator response value of the image, used to highlight edge regions in the image, where x is the horizontal pixel index of the original image and y is the vertical pixel index of the original image.

[0078] Step S13: Multi-factor fusion classification evaluation, specifically by obtaining a multi-factor fusion image quality screening threshold based on the image quality factor parameter set through parameter index normalization and comprehensive image quality weighting calculation, and then performing image classification on the basic enhanced image data based on the multi-factor fusion image quality screening threshold to obtain image classification status labels;

[0079] The parameter normalization is specifically based on the image quality factor parameter set, performing max-min normalization calculations to obtain normalized image quality parameters. Then, based on these normalized image quality parameters, a comprehensive image quality weighted calculation is performed to obtain the multi-factor fusion image quality screening threshold. The calculation formula is as follows:

[0080] ;

[0081] In the formula, Q total It is a multi-factor fusion image quality screening threshold. It is the signal-to-noise ratio weight. It is the normalized signal-to-noise ratio parameter. It is the contrast weight. It is the normalized contrast parameter. It is the boundary clarity weight. It is a normalized boundary sharpness parameter;

[0082] Preferably, the signal-to-noise ratio weight The default value is 0.4; the contrast weight The default value is set to 0.3; the boundary sharpness weight The default value is set to 0.3;

[0083] Preferably, Table 1 is a reference example table for image classification of basic enhanced image data. As shown in the table, when the multi-factor fusion image quality screening threshold is less than 0.4, the image classification level is low quality; when the multi-factor fusion image quality screening threshold is less than 0.7 and greater than or equal to 0.4, the image classification level is medium quality; and when the multi-factor fusion image quality screening threshold is greater than or equal to 0.7, the image classification level is high quality.

[0084] Table 1. Reference Examples of Image Classification Based on Basic Augmented Image Data

[0085]

[0086] Step S14: Parameter optimization suggestion generation, specifically, based on the image classification state label, an image optimization parameter mapping table is constructed, and image optimization parameters are matched using a rule-based method to obtain parameter optimization suggestion data;

[0087] Preferably, Table 2 is a rule-based table for matching image optimization parameters. As shown in the table, the quality factor anomalies include signal-to-noise ratio, contrast and boundary sharpness. The optimization suggestion example is generated by inputting the optimization suggestion parameters as keywords into the expert model.

[0088] Table 2. Schematic diagram of rules for image optimization parameter matching using rule-based methods.

[0089]

[0090] Step S15: Automatic optimization of ultrasound images, specifically, based on the parameter optimization suggestion data, an image optimization operation is performed to automatically optimize the basic enhanced image data to obtain optimized ultrasound image data.

[0091] By performing the above operations, this solution addresses the technical problems of low efficiency and long processing time in existing ultrasound image optimization and enhancement methods. In traditional schemes, ultrasound image optimization usually relies on statically set parameters for post-processing, which is cumbersome and slow. This often fails to meet the requirements of real-time streaming processing in practical applications, especially in high-throughput scanning or dynamic continuous monitoring. This solution creatively adopts an intelligent ultrasound image optimization scheme that combines quality factor fusion and parameter back mapping to perform real-time parameter feedback optimization of ultrasound images. This makes the ultrasound image optimization process more efficient, enabling the acquisition of high-quality images in a short time and ensuring fast and accurate data processing support.

[0092] Example 3, see Figure 1 , Figure 2 and Figure 4 This embodiment is based on the above embodiment. In step S2, the key structure annotation is used to identify and annotate the key structures of the main body of the ultrasound image. Specifically, based on the optimized ultrasound image data, a method for setting up ultrasound image structure annotation points that combines morphological perception prediction and structural evaluation is used to annotate the key structures and obtain structure-annotated image data. This includes the following steps:

[0093] Step S21: Structural segmentation, specifically, using a lightweight semantic segmentation model to perform structural segmentation based on the optimized ultrasound image data, and using an edge extraction algorithm to extract structural edges to obtain basic structural segmentation data;

[0094] The lightweight semantic segmentation model specifically adopts the improved DeepLabv3+ model; the edge extraction algorithm specifically adopts the boundary tracking algorithm improved by introducing directional gradient constraints.

[0095] The improved DeepLabv3+ model specifically improves the semantic segmentation model by replacing the standard ResNet backbone with a lightweight MobileNetV3 backbone network, adding a structure-aware module to the decoder, and integrating a joint attention module into the spatial pyramid pooling layer. The calculation formula is as follows:

[0096] ;

[0097] In the formula, It is the structural segmentation mask output by a lightweight semantic segmentation model, D SAR It is a structure-aware residual decoder module, used to represent adding a structure-aware module to the decoder and enhancing the shape awareness of boundaries and target objects, F ASPPCBAM It is a spatial pyramid pooling module that integrates channel spatial attention, used to optimize ultrasound image features, B MobileNetV3 It is a lightweight, improved backbone network used to replace ResNet and extract image features; I is the image identifier.

[0098] The improved boundary tracking algorithm with introduced directional gradient constraints specifically introduces directional gradient constraints on top of the standard boundary tracking algorithm. These constraints include gradient energy, directional consistency, and path smoothing terms, used to prevent the contour of the target structure from crossing weak edge regions. The calculation formula is as follows:

[0099] ;

[0100] In the formula, B is the continuous boundary path output, used to represent the boundary path extracted by the improved boundary tracing algorithm with introduced directional gradient constraints, P is the sequence of sampling points on the boundary tracing path, and p j Here, J is the total number of sampling points, j is the index of the sampling point, and E is the index of the sampling point. grad It is the gradient energy term, used to represent marginal significance, E angle It is a direction consistency term used to measure changes in edge direction and prevent boundary jumps, E smooth It is a path smoothing term used to limit the jagged inflection points of the boundary;

[0101] Preferably, the gradient energy term is specifically extracted from the gradient map using the standard Scharr operator, and the gradient magnitude of the sampling points is calculated as the gradient energy term.

[0102] The direction consistency term is specifically calculated by taking the cosine similarity of the vector directions of three consecutive sampling points on the boundary trajectory and the angle between the vector directions of the three sampling points as the direction consistency term.

[0103] The path smoothing term is specifically calculated by taking the second-order difference curvature of three consecutive annual sampling points as the path smoothing term.

[0104] Step S22: Key structural point fitting, specifically by constructing structural instance objects and performing geometric fitting and key point prediction on the instance objects to obtain structural key point pair data;

[0105] Preferably, in this embodiment, taking fetal ultrasound images as an example, the structural instance objects specifically include the fetal head, abdominal contour, femoral shaft, cerebellum contour, four-chamber heart region, and spine;

[0106] The measurements of the tire head include biparietal diameter and head circumference;

[0107] The tire head is specifically measured by fitting a least-squares ellipse;

[0108] The formula for calculating the bicornuate diameter is as follows:

[0109] ;

[0110] In the formula, These are the coordinates of the left end point of the biparietal diameter head. The coordinates of the right end point of the biparietal diameter head, x c The horizontal coordinate of the center of the fitted ellipse is y. c is the vertical coordinate of the center of the fitted ellipse, and b is the radius of the minor axis of the fitted ellipse. It is the rotation angle of the fitted ellipse;

[0111] The formula for calculating the head circumference is:

[0112] ;

[0113] In the formula, HC is the estimated head circumference, and a is the radius of the major axis of the fitted ellipse;

[0114] The measurement item for the abdominal circumference specifically refers to the chest circumference;

[0115] The abdominal circumference contour is specifically calculated using a least-squares ellipse fitting that is identical to that of the fetal head;

[0116] The measurement items for the femoral shaft specifically refer to femoral length;

[0117] The formula for calculating the femur length is:

[0118] ;

[0119] In the formula, FL is the femoral length value, x1 is the horizontal coordinate of the proximal end of the femur, x2 is the horizontal coordinate of the caudal end of the femur, y1 is the vertical coordinate of the proximal end of the femur, and y2 is the vertical coordinate of the caudal end of the femur.

[0120] The measurement of the cerebellar contour specifically refers to the transverse diameter of the cerebellum;

[0121] The cerebellar transverse diameter is specifically measured by fitting the maximum inscribed transverse diameter;

[0122] The measurement items for the four chambers of the heart specifically refer to the cardiothoracic ratio.

[0123] The cardiothoracic ratio is specifically determined by extracting the morphology of the four-chamber region of the heart.

[0124] The spine is used as an auxiliary marker for fetal position and to locate the back area;

[0125] Step S23: Generate structural annotation point values. Specifically, based on the data of the structural key points, the drawing function is called to generate the structural point values ​​and cursors to obtain the structural annotation measurement values.

[0126] Preferably, the drawing function, specifically the function call function, is shown in the following pseudocode:

[0127] ;

[0128] In the formula, DrawLine is the function to draw the structural measurement line segment, P1 is the first coordinate index of the drawing point, P2 is the second coordinate index of the drawing point, DisplayValue is the function to display the calculated measurement value, and R is the resolution conversion factor, specifically set as the conversion factor for converting pixel units to millimeter units;

[0129] The cursor generation is as follows: the specific cursor type corresponding to biparietal diameter and femur length is a straight line with two ends and a value; the specific cursor type corresponding to head circumference and chest circumference is a closed elliptical curve and a circumference value; and the specific cursor type corresponding to cerebellar transverse diameter is a horizontal line and a value.

[0130] Step S24: Structural integrity assessment, specifically, weighted summation of the image quality assessment reference data corresponding to each structural key point data to obtain structural integrity score reference data, which is used to assist in key frame selection;

[0131] The formula for calculating the structural integrity reference score is as follows:

[0132] ;

[0133] In the formula, Q f It is a structural integrity reference score. It is the structural fitting error weight, Q shape It is the structural fitting error parameter. Qclarity is the gradient variance weight of the structural region, and Qclarity is the gradient variance parameter of the structural region. It is the symmetry scoring weight, Q symmetry It is the fitting symmetry score parameter;

[0134] Preferably, the structural fitting error weights The default value is 0.4, and the gradient variance weight of the structural region is... The default value is 0.3, and the symmetry scoring weight is... The default value is 0.3;

[0135] Preferably, when the structural integrity reference score is 0.8 or more, the system determines it as a recommended keyframe;

[0136] Step S25: Key structure annotation, specifically, based on the structure annotation measurement values ​​and the structure completeness score reference data, key structure annotation is performed on each structure instance object to obtain structure annotation image data.

[0137] By performing the above operations, this solution addresses the technical problem that traditional ultrasound image structure annotation methods rely on manual annotation or simple image processing algorithms in the existing target structure identification and annotation process. This often leads to low annotation accuracy, especially in the identification of complex geometric shapes or weak boundary structures, where errors are prone to occur. Specifically, the annotation of low-contrast regions or overlapping tissues is easily affected by changes in image quality and target position, resulting in inaccurate or missed annotations. This solution creatively adopts an ultrasound image structure annotation point layout method that combines morphological perception prediction and structural evaluation to annotate key features. This method can more accurately locate and annotate the key structures of the target to be detected, thereby providing more reliable basic data for subsequent quantitative analysis.

[0138] Example 4, see Figure 1 , Figure 2 and Figure 5 This embodiment is based on the above embodiment. In step S3, the keyframe selection is used to automatically identify the standard target observation section and automatically select keyframes at relatively stable positions of the target structure. Specifically, based on the structure-annotated image data, a keyframe selection and determination method for ultrasound images that combines temporal structure stability analysis and image registration improvement is used to select keyframes and obtain optimized keyframe image data. This includes the following steps:

[0139] Step S31: Structural trajectory tracking, specifically, based on the structural annotation image data, the structural center of gravity is calculated and the structural motion trajectory is recorded sequentially to obtain the instance structural trajectory tracking data;

[0140] Step S32: Structural stability scoring, specifically by sequentially calculating translation amplitude score, contour deformation score and image texture change score, performing a weighted structural stability score, and obtaining comprehensive frame selection score reference data;

[0141] The calculation of the translation amplitude score is specifically performed by calculating the translation amplitude of the frame image, and the calculation formula is as follows:

[0142] ;

[0143] In the formula, Q move This is a translational amplitude score, used to represent the structural stability during motion. Where i is the total duration, i is the time index, and C is the total duration. t-i+1 C is the geometric center coordinate of the structure at the current time t minus the coordinates of index frame i+1. t-i It is the geometric center coordinate of the structure at the current time t minus the time of index frame i, and MaxMove is the maximum translation parameter;

[0144] The maximum translational parameter is set based on the statistical upper limit of the maximum normal translational distance and is used to normalize the actual inter-frame displacement amplitude to adapt to different probe operation amplitudes.

[0145] The contour deformation score is specifically calculated by averaging the structural masks across consecutive frames. The calculation formula is as follows:

[0146] ;

[0147] In the formula, Q sh It is the profile deformation score, S t-i It is the structural mask region at the current time t minus the index frame i, used to represent the binary contour of the structure, S t-i+1 It is the structure mask region at the current time t minus the time of index frame i+1. It is the intersection symbol. It is the union symbol;

[0148] The image texture change score is specifically calculated using structural region feature descriptors, and the formula is as follows:

[0149] ;

[0150] In the formula, Q texture It is the image texture variation score, mean i It is the mean calculation function, cos sim It is the cosine similarity function, F t-i It is the structural region feature vector at current time t minus index frame i, F t-i+1 It is the current time t minus the structural region feature vector at index frame i+1;

[0151] The comprehensive frame selection score reference data is specifically obtained by weighted summation of the translation amplitude score, contour deformation score, and image texture change score, combined with the structural integrity reference score. The calculation formula is as follows:

[0152] ;

[0153] In the formula, Q select It is a comprehensive frame selection scoring reference data, where a1 is the translation amplitude weight, a2 is the contour deformation weight, a3 is the image texture change weight, and a4 is the structural integrity weight.

[0154] Preferably, the default value of the translation amplitude weight a1 is set to 0.3, the default value of the contour deformation weight a2 is set to 0.2, the default value of the image texture change weight a3 is set to 0.2, and the default value of the structural integrity weight a4 is set to 0.3.

[0155] Preferably, when the value of the comprehensive frame selection score reference data is 0.85 or higher, the system automatically selects frames; when the value of the comprehensive frame selection score reference data is [0.7, 0.85), the system provides suggestions for frame selection and assists manual frame selection decision-making.

[0156] Step S33: Select a judgment prompt, specifically by setting a frame selection threshold, making a frame selection judgment based on the comprehensive frame selection score reference data, and generating an image frame selection prompt in real time by combining the image quality assessment reference data;

[0157] Step S34: Motion blur registration improvement, specifically by detecting the structural relationship between the current frame and the previous frame, and using a motion blur compensation method that combines the result region registration enhancement to improve the registration and obtain the registration-optimized frame image data;

[0158] Step S35: Keyframe selection, specifically, based on the image frame selection prompts, semi-automatic image frame selection is performed, and based on the registration optimized frame selection image data, the instantaneous cross-sectional frame selection image is optimized to obtain frame selection image optimization data.

[0159] By performing the above operations, this solution addresses the technical problem in the existing ultrasound image keyframe selection process where the selected frame image is often blurry or fails to capture the stable state of the target structure due to target movement or improper operation. This poses a significant challenge to subsequent analysis. Specifically, even slight changes in the target position may prevent the image frame selection from aligning with the most stable part of the observed subject, affecting the system's calculations and judgments. This solution creatively adopts an ultrasound image keyframe selection and determination method that combines temporal structure stability analysis and image registration improvement to lock the temporal keyframe. This method can accurately determine and select the optimal section when the target is in a stable position for locking, ensuring that the image is clear and meets the detection requirements. At the same time, it reduces interference caused by motion blur, improving the accuracy and efficiency of the processing.

[0160] Example 5, see Figure 1 , Figure 2 This embodiment is based on the above embodiment. In step S4, the ultrasound image classification enhancement is used to combine optimized ultrasound images, structural annotation information and frame selection image optimization data to assist doctors in ultrasound image classification judgment. Specifically, based on the optimized ultrasound image data, the structural annotation image data and the frame selection image optimization data, ultrasound image classification enhancement is performed by sequentially analyzing morphological parameters and image parameters to obtain comprehensive ultrasound image classification reference data.

[0161] Example 6, see Figure 1 and Figure 2 Based on the above embodiments, this embodiment provides an ultrasound image classification and enhancement system, including an ultrasound image optimization module, a key structure annotation module, a key frame selection module, and an image classification and enhancement module.

[0162] The ultrasound image optimization module is used to optimize ultrasound images. Through ultrasound image optimization, optimized ultrasound image data is obtained, and the optimized ultrasound image data is sent to the key structure annotation module and the image classification enhancement module.

[0163] The key structure annotation module is used for key structure annotation. Through key structure annotation, it obtains structure annotation image data and sends the structure annotation image data to the key frame selection module and the image classification enhancement module.

[0164] The keyframe selection module is used for keyframe selection. Through keyframe selection, it obtains frame selection image optimization data and sends the frame selection image optimization data to the image classification enhancement module.

[0165] The image classification enhancement module is used for image classification enhancement, and through image classification enhancement, diagnostic auxiliary measurement parameter data is obtained.

[0166] Example 7, see Figure 1and Figure 2 This embodiment is based on the above embodiments. The present invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of an ultrasound image classification enhancement method as described in any of the above embodiments.

[0167] The computer-readable storage media involved in this invention include random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs (compact disc read-only memory), or any other form of storage media known in the art.

[0168] The present invention provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the ultrasound image classification enhancement method as described in any of the above embodiments.

[0169] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0170] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0171] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A method for classifying and enhancing ultrasound images, characterized in that: The method includes the following steps: Step S1: Ultrasound image optimization. Obtain the original ultrasound image and perform preprocessing and image enhancement. Extract image quality factors based on the basic enhanced image data, and obtain the multi-factor fusion image quality screening threshold through parameter index normalization and comprehensive image quality weighted calculation. Based on the multi-factor fusion image quality screening threshold, the basic enhanced image data is classified to obtain image classification status labels, and parameter optimization suggestion data is generated based on the image classification status labels. Based on the parameter optimization suggestion data, an image optimization operation is performed to automatically optimize the basic enhanced image data to obtain optimized ultrasound image data; Step S2: Key structure annotation. Based on the optimized ultrasound image data, key structures are annotated to obtain annotated structure image data. The structural integrity of the key structural points is evaluated to obtain structural integrity score reference data. Step S3: Keyframe selection. Based on the structure-annotated image data and the structure completeness score reference data, keyframes are selected to obtain optimized frame image data. Step S4: Ultrasound image classification enhancement. Combining the optimized ultrasound image data, the structurally labeled image data, and the frame-selected image optimization data, ultrasound image classification enhancement is performed by sequentially analyzing morphological parameters and image parameters to obtain comprehensive ultrasound image classification reference data.

2. The ultrasound image classification and enhancement method according to claim 1, characterized in that: The image quality factor extraction specifically involves extracting image quality factors based on the base enhanced image data by sequentially performing signal-to-noise ratio estimation, contrast estimation, and boundary sharpness estimation, thereby obtaining an image quality factor parameter set. The image quality factor parameter set specifically includes signal-to-noise ratio parameters, contrast parameters, and boundary sharpness parameters.

3. The ultrasound image classification and enhancement method according to claim 2, characterized in that: The formula for calculating the multi-factor fusion image quality screening threshold is as follows: ； In the formula, Q total It is a multi-factor fusion image quality screening threshold. It is the signal-to-noise ratio weight. It is the normalized signal-to-noise ratio parameter. It is the contrast weight. It is the normalized contrast parameter. It is the boundary clarity weight. It is the normalized boundary sharpness parameter.

4. The ultrasound image classification and enhancement method according to claim 3, characterized in that: The parameter optimization suggestion generation specifically involves constructing an image optimization parameter mapping table based on the image classification status label, and then matching the image optimization parameters using a rule-based method to obtain parameter optimization suggestion data. The automatic ultrasound image optimization specifically involves performing image optimization operations based on the parameter optimization suggestion data to automatically optimize the basic enhanced image data and obtain optimized ultrasound image data.

5. The ultrasound image classification and enhancement method according to claim 4, characterized in that: The key structure annotation is specifically performed by using an ultrasound image structure annotation point layout method that combines morphological perception prediction and structural evaluation based on the optimized ultrasound image data, to obtain structure annotation image data and generate structure annotation measurement values.

6. The ultrasound image classification and enhancement method according to claim 5, characterized in that: The structural integrity assessment specifically involves weighted summation of the image quality assessment reference data corresponding to each structural key point to obtain structural integrity score reference data, which is then used to assist in key frame selection. The formula for calculating the structural integrity reference score is as follows: ； In the formula, Q f It is a structural integrity reference score. It is the structural fitting error weight, Q shape It is the structural fitting error parameter. It is the gradient variance weight of the structural region, Q clarity It is the gradient variance parameter of the structural region. It is the symmetry scoring weight, Q symmetry It is the fitting symmetry score parameter.

7. The ultrasound image classification and enhancement method according to claim 6, characterized in that: The keyframe selection is specifically based on the structure-annotated image data, using an improved ultrasound image keyframe selection and determination method that combines temporal structure stability analysis and image registration. This involves sequentially calculating translational amplitude scores, contour deformation scores, and image texture change scores, and then weighting and summing these scores in conjunction with the structure integrity score reference data to obtain comprehensive frame selection score reference data. Keyframes are then selected based on this comprehensive frame selection score reference data.

8. An ultrasound image classification and enhancement system for implementing the ultrasound image classification and enhancement method as described in any one of claims 1-7, characterized in that: It includes an ultrasound image optimization module, a key structure annotation module, a key frame selection module, and an image classification enhancement module.

9. The ultrasound image classification and enhancement system according to claim 8, characterized in that: The ultrasound image optimization module is used to optimize ultrasound images. Through ultrasound image optimization, optimized ultrasound image data is obtained, and the optimized ultrasound image data is sent to the key structure annotation module and the image classification enhancement module. The key structure annotation module is used for key structure annotation. Through key structure annotation, it obtains structure annotation image data and sends the structure annotation image data to the key frame selection module and the image classification enhancement module. The keyframe selection module is used for keyframe selection. Through keyframe selection, it obtains frame selection image optimization data and sends the frame selection image optimization data to the image classification enhancement module. The image classification enhancement module is used for image classification enhancement, and through image classification enhancement, diagnostic auxiliary measurement parameter data is obtained.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the ultrasound image classification enhancement method as described in any one of claims 1 to 7.

Images