Gynecological ultrasonic image analysis method and system

Abstract

The invention provides a gynecological ultrasonic image analysis method and system, and the method comprises the steps: obtaining an anatomical structure feature set including endometrial boundary features, ovarian follicle distribution features and uterine muscular layer texture features through a standardized ultrasonic image, and generating a structured segmentation map of a key anatomical structure of a target subject based on the anatomical structure feature set; performing dynamic parameter quantization on the structured segmentation map to obtain physiological dynamic parameters of the target subject, and further obtaining a functional state evolution curve of the reproductive organ of the target subject; performing feature level fusion on the structured segmentation map and the functional state evolution curve of the reproductive organ, and determining a multi-dimensional evaluation matrix of the reproductive health state of the target subject according to a fusion result; and generating a comprehensive diagnosis report of the target subject according to the multi-dimensional evaluation matrix, and using the comprehensive diagnosis report for auxiliary diagnosis of clinicians. By means of the scheme, automatic early screening of gynecological diseases can be achieved based on combined dynamic analysis of the anatomical structure and the physiological function.

Description

Technical Field

[0001] This application relates to the field of image analysis technology, and more specifically, to a gynecological ultrasound image analysis method and system. Background Technology

[0002] Ultrasound image analysis refers to the process of digitally processing, extracting features, and quantitatively analyzing image data acquired by a medical ultrasound imaging system. This data is used to identify tissue structures, determine the nature of lesions, measure organ size, or assess physiological function. Gynecological ultrasound image analysis refers to the process of using ultrasound technology to acquire images of female pelvic organs such as the uterus, ovaries, fallopian tubes, and cervix, which are then observed, measured, evaluated, and diagnosed by a professional physician, typically an ultrasound specialist or obstetrician-gynecologist.

[0003] Gynecological ultrasound image analysis is an indispensable "eye" in modern obstetrics and gynecology. It is a safe, radiation-free, real-time, and efficient imaging tool that plays a crucial role in the prevention, screening, diagnosis, treatment, and follow-up of gynecological diseases. However, current gynecological ultrasound image analysis largely relies on physicians manually interpreting two-dimensional ultrasound images, evaluating them by manually measuring single parameters such as endometrial thickness and follicle diameter. This method suffers from high subjectivity, insufficient quantitative analysis, and limited utilization of dynamic information. Especially when facing complex clinical scenarios requiring comprehensive assessment of multiple anatomical structures and analysis of temporal dynamic changes, current methods struggle to achieve satisfactory results. Precise differentiation between the endometrial boundary and myometrial texture, quantitative analysis of the spatial distribution of multiple follicles, and dynamic tracking of reproductive organ function have resulted in a single assessment dimension, insufficient utilization of spatial features, and weak ability to capture dynamic evolution patterns. In recent years, although some studies have attempted to apply image processing algorithms to ultrasound analysis, they are still mostly limited to static feature extraction at a single time point and lack the ability to collaboratively analyze multi-scale anatomical features, resulting in insufficient sensitivity in identifying abnormal states and difficulty in achieving a comprehensive assessment of reproductive health status. Therefore, how to achieve automated early screening of gynecological diseases based on the joint dynamic analysis of anatomical structure and physiological function has become a challenge for the industry. Summary of the Invention

[0004] This application provides a gynecological ultrasound image analysis method and system, which can realize automated early screening of gynecological diseases based on the combined dynamic analysis of anatomical structure and physiological function.

[0005] In a first aspect, this application provides a method for analyzing gynecological ultrasound images, comprising the following steps: The raw ultrasound image data of the target subject during the gynecological ultrasound examination is acquired, and the raw ultrasound image data is preprocessed to obtain standardized ultrasound images. Multi-scale feature extraction is performed on the standardized ultrasound images to obtain an anatomical feature set including endometrial boundary features, ovarian follicle distribution features, and uterine myometrial texture features. Based on the anatomical feature set, a structured segmentation atlas of key anatomical structures of the target subject is generated. The structured segmentation map is dynamically parameterized to obtain the physiological dynamic parameters of the target subject. The physiological dynamic parameters are then fitted with trends at multiple consecutive time points to obtain the functional state evolution curve of the target subject's reproductive organs. The structured segmentation map is fused with the functional state evolution curve of the reproductive organs at the feature level, and then a multi-dimensional assessment matrix of the reproductive health status of the target subject is determined based on the fusion result. The multi-dimensional assessment matrix is ​​compared and analyzed with the preset clinical diagnostic thresholds to generate a comprehensive diagnostic report for the target subject, which is then sent to a medical analysis terminal for clinical doctors to assist in diagnosis.

[0006] In some embodiments, multi-scale feature extraction is performed on the standardized ultrasound image to obtain an anatomical feature set including endometrial boundary features, ovarian follicle distribution features, and uterine myometrial texture features. Specifically, this includes: Extract multi-scale image features from the standardized ultrasound images; Based on the multi-scale image features, the endometrial boundary was identified, the ovarian follicle region was located, and the texture of the myometrium was analyzed. The obtained identification results, localization results, and analysis results are integrated into an anatomical structure feature set.

[0007] In some embodiments, generating a structured segmentation atlas of key anatomical structures of the target subject based on the anatomical feature set specifically includes: Pixel-level classification feature vectors are constructed based on the anatomical structure feature set; Based on the pixel-level classification feature vector, tissue category classification is performed on each pixel in the standardized ultrasound image; The classification results generate a structured segmentation map of the key anatomical structures of the target subject.

[0008] In some embodiments, dynamic parameter quantization of the structured segmentation map to obtain the target subject's dynamic physiological parameters specifically includes: Quantitative indicators of key anatomical structures are extracted from the structured segmentation map; The physiological dynamic parameters of the target subject are calculated based on the quantitative indicators.

[0009] In some embodiments, trend fitting of the physiological dynamic parameters over multiple consecutive time points to obtain the functional state evolution curve of the target subject's reproductive organs specifically includes: Organize physiological dynamic parameters at multiple consecutive time points into time series data; The time series data is subjected to curve fitting processing; Based on the fitting results, the functional state evolution curve of the reproductive organs of the target subject is generated.

[0010] In some embodiments, feature-level fusion of the structured segmentation map and the functional state evolution curve of the reproductive organ specifically includes: The spatial morphological feature vector of the target subject's reproductive organs is extracted from the structured segmentation map; The temporal dynamic feature vector of the reproductive organ functional state of the target subject is determined based on the aforementioned functional state evolution curve; The spatial morphological feature vector and the temporal dynamic feature vector are fused to obtain the fusion result.

[0011] In some embodiments, determining the multi-dimensional assessment matrix of the reproductive health status of the target subject based on the obtained fusion results specifically includes: A multi-dimensional assessment matrix was determined based on clinical diagnostic criteria. The fusion results are mapped to the corresponding dimensions of the multi-dimensional evaluation matrix; The mapped evaluation indicators are standardized and scored to generate a multi-dimensional evaluation matrix of the reproductive health status of the target examinee.

[0012] Secondly, this application provides a gynecological ultrasound image analysis system, comprising: The acquisition module is used to acquire the original ultrasound image data of the target subject during the gynecological ultrasound examination, and to preprocess the original ultrasound image data to obtain standardized ultrasound images. The processing module is used to perform multi-scale feature extraction on the standardized ultrasound image to obtain an anatomical structure feature set including endometrial boundary features, ovarian follicle distribution features and uterine myometrial texture features, and to generate a structured segmentation atlas of key anatomical structures of the target subject based on the anatomical structure feature set. The processing module is also used to perform dynamic parameter quantization on the structured segmentation map to obtain the physiological dynamic parameters of the target subject, and to perform trend fitting on the physiological dynamic parameters over multiple consecutive time points to obtain the functional state evolution curve of the target subject's reproductive organs. The processing module is also used to perform feature-level fusion of the structured segmentation map and the functional state evolution curve of the reproductive organs, and then determine the multi-dimensional assessment matrix of the reproductive health status of the target subject based on the obtained fusion result. The execution module is used to compare and analyze the multi-dimensional evaluation matrix with the preset clinical diagnostic threshold, generate a comprehensive diagnostic report for the target subject, and send it to the medical analysis terminal for clinical doctors to assist in diagnosis.

[0013] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory for storing a computer program, and the processor for calling and running the computer program from the memory, so that the computer device performs the above-described gynecological ultrasound image analysis method.

[0014] Fourthly, this application provides a computer-readable storage medium storing instructions or code that, when executed on a computer, cause the computer to perform the aforementioned gynecological ultrasound image analysis method.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: In this application, raw ultrasound image data of the target subject during a gynecological ultrasound examination is acquired, and the raw ultrasound image data is preprocessed to obtain standardized ultrasound images. Multi-scale feature extraction is performed on the standardized ultrasound images to obtain an anatomical feature set including endometrial boundary features, ovarian follicle distribution features, and uterine myometrial texture features. Based on the anatomical feature set, a structured segmentation atlas of the target subject's key anatomical structures is generated. Dynamic parameters of the structured segmentation atlas are quantified to obtain the target subject's physiological dynamic parameters. Trend fitting of the physiological dynamic parameters is performed over multiple consecutive time points to obtain the functional state evolution curve of the target subject's reproductive organs. The structured segmentation atlas and the functional state evolution curve of the reproductive organs are fused at the feature level, and a multi-dimensional assessment matrix of the target subject's reproductive health status is determined based on the fusion result. The multi-dimensional assessment matrix is ​​compared and analyzed with a preset clinical diagnostic threshold to generate a comprehensive diagnostic report for the target subject, which is then sent to a medical analysis terminal for clinical physician-assisted diagnosis.

[0016] Therefore, in this application, firstly, a structured segmentation map of the key anatomical structures of the target subject is generated based on the anatomical feature set. This allows for the understanding of the spatial topological relationships of different anatomical regions at the pixel level, providing a high-resolution structural foundation for subsequent parameter extraction and dynamic modeling. This improves the accuracy and consistency of image structure analysis, laying a precise anatomical basis for automated screening of gynecological diseases. Secondly, trend fitting is performed on the physiological dynamic parameters over multiple consecutive time points to obtain the functional state evolution curve of the target subject's reproductive organs. This enables a shift from static measurement to dynamic process analysis, providing temporal perception and functional prediction capabilities. This allows for the early detection of lesion signs, thereby improving the timeliness and sensitivity of early screening for gynecological diseases. Then, the structured segmentation map and the functional state evolution curve of the reproductive organs are fused at the feature level, and the results are then used to determine the optimal method for data processing. A multi-dimensional assessment matrix for the reproductive health status of target subjects can establish a unified representation system between spatial features and temporal dynamics, achieving synergistic representation at the anatomical and physiological levels. This strengthens the multi-angle identification of pathological features and improves the ability to make global judgments on disease evolution in complex clinical scenarios. Finally, by comparing and analyzing the multi-dimensional assessment matrix with preset clinical diagnostic thresholds, an automatic closed loop from feature extraction to clinical interpretation can be achieved. This digitizes and parameterizes doctors' experience standards, improving screening efficiency and consistency, and generating a comprehensive diagnostic report for the target subjects for clinical doctors' auxiliary diagnosis. This effectively reduces misdiagnosis and missed diagnosis, ultimately achieving intelligent and automated early screening of gynecological diseases and enhancing the objectivity and reliability of clinical diagnosis. In summary, this scheme can achieve automated early screening of gynecological diseases based on the joint dynamic analysis of anatomical structure and physiological function. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is an exemplary flowchart of a gynecological ultrasound image analysis method according to some embodiments of this application; Figure 2 This is an exemplary flowchart illustrating the determination of a structured segmentation map according to some embodiments of this application; Figure 3 This is an exemplary flowchart illustrating the implementation of feature-level fusion according to some embodiments of this application; Figure 4This is a schematic diagram of the structure of a gynecological ultrasound image analysis system according to some embodiments of this application; Figure 5 This is a schematic diagram of the structure of a computer device for implementing a gynecological ultrasound image analysis method according to some embodiments of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] refer to Figure 1 The figure is an exemplary flowchart of a gynecological ultrasound image analysis method according to some embodiments of this application. The gynecological ultrasound image analysis method mainly includes the following steps: In step 101, the original ultrasound image data of the target subject during the gynecological ultrasound examination is acquired, and the original ultrasound image data is preprocessed to obtain standardized ultrasound images.

[0021] In specific implementation, the acquisition of raw ultrasound image data of the target subject during gynecological ultrasound examination can be achieved in the following ways: raw DICOM format image data of the target subject including the uterus, ovaries and accessory structures can be directly acquired through the image output interface of the gynecological ultrasound diagnostic instrument, or the raw ultrasound image dataset of the corresponding examination sequence can be retrieved through the medical image archiving and communication system using the patient's ID number; wherein, as a preferred embodiment, the raw ultrasound image data is three-dimensional ultrasound volume data, and in other embodiments, two-dimensional ultrasound sequence images acquired at preset time intervals can also be used, which is not specifically limited here.

[0022] It should be noted that the original ultrasound image data in this application refers to the initial image data directly generated during the gynecological ultrasound examination.

[0023] In specific implementation, the preprocessing of the original ultrasound image data to obtain a standardized ultrasound image can be achieved in the following way: First, the original image data is parsed to obtain a pixel matrix. Then, a median filtering algorithm is used to suppress the speckle noise unique to ultrasound, and a contrast-limited adaptive histogram equalization algorithm is used to enhance the contrast of tissue boundaries. Next, depth attenuation compensation is performed based on the ultrasound probe frequency and acquisition depth. Finally, the image is uniformly resampled to a size of 512×512 pixels and normalized to a grayscale range of 0-255 to generate a standardized ultrasound image for feature extraction. In a preferred embodiment, the depth attenuation compensation can be simulated using a time gain compensation curve. In other embodiments, anisotropic diffusion filtering can also be used for noise suppression, which is not specifically limited here.

[0024] It should be noted that the standardized ultrasound image in this application refers to the image generated after a series of standardization processes on the original ultrasound image data, which is used to eliminate image quality fluctuations caused by differences in equipment, acquisition parameters and scanning conditions.

[0025] In step 102, multi-scale feature extraction is performed on the standardized ultrasound image to obtain an anatomical feature set including endometrial boundary features, ovarian follicle distribution features, and uterine myometrial texture features. Based on the anatomical feature set, a structured segmentation atlas of key anatomical structures of the target subject is generated.

[0026] In some embodiments, multi-scale feature extraction of the standardized ultrasound image to obtain an anatomical feature set including endometrial boundary features, ovarian follicle distribution features, and myometrial texture features can be achieved through the following steps: Extract multi-scale image features from the standardized ultrasound images; Based on the multi-scale image features, the endometrial boundary was identified, the ovarian follicle region was located, and the texture of the myometrium was analyzed. The obtained identification results, localization results, and analysis results are integrated into an anatomical structure feature set.

[0027] In specific implementation, the extraction of multi-scale image features from the standardized ultrasound image is achieved in the following manner: First, a Gaussian pyramid can be constructed to decompose the standardized ultrasound image into multiple scales, obtaining image sequences at different resolution scales; then, Gabor filter banks are applied at each scale level to extract texture features in different directions, and local binary pattern features are calculated to characterize the micro-texture structure; finally, the texture feature maps extracted at each scale are fused with the original scale image to form multi-scale image features containing macroscopic contour information and micro-texture details; wherein, as a preferred embodiment, the Gaussian pyramid can adopt a 4-layer decomposition structure, and in other embodiments, it can also be... Multi-scale feature extraction can be achieved using wavelet transform or multi-layer feature maps from deep convolutional neural networks, without limitation. Based on the multi-scale image features, the identification of endometrial boundaries, the localization of ovarian follicle regions, and the analysis of myometrial texture can be achieved in the following ways: For endometrial boundary identification, based on the edge gradient information in the multi-scale features, an active contour model can be used to iteratively evolve the initial contour to the endometrial-myometrial junction, extracting morphological features such as endometrial thickness and boundary curvature, which are then used as the identified endometrial boundary features. For ovarian follicle region localization, multi-scale texture features can be used to detect candidate follicle regions through circular Hough transform, combined with region growth... The algorithm accurately locates follicle boundaries, extracts follicle quantity, diameter distribution, and spatial arrangement features, and uses these as the obtained ovarian follicle distribution feature results. For uterine myometrial texture analysis, statistical measures such as contrast, correlation, and entropy of the gray-level co-occurrence matrix can be calculated based on multi-scale features to quantify the uniformity and texture complexity of the myometrial tissue, and these are used as the obtained uterine myometrial texture feature results. In a preferred embodiment, the active contour model can adopt an edge-based geometric active contour; in other embodiments, the level set method can also be used for endometrial boundary evolution, which is not limited here. The obtained recognition results, localization results, and analysis results are integrated into anatomical structure features. The collection can be achieved in the following way: First, the identified endometrial boundary features can be represented as a numerical vector containing thickness sequence and contour curvature parameters; the located ovarian follicle distribution features can be encoded as a feature matrix containing follicle location coordinates, diameter size, and distribution density; and the analyzed myometrial texture features can be quantified as a feature vector of texture statistics. Then, the above heterogeneous features are standardized to eliminate the influence of dimensions, and then the above features are concatenated into a unified multidimensional feature vector according to a predetermined feature encoding specification. Finally, a feature index mapping table is established for this multidimensional feature vector to clarify the anatomical structural attributes corresponding to each feature dimension, so as to form a structured anatomical structural feature set.In one preferred embodiment, the feature encoding specification can arrange the feature dimensions sequentially according to the anatomical order of the endometrium, ovary, and myometrium. In other embodiments, principal component analysis can be used to reduce the dimensionality of the features before integration; this is not limited here.

[0028] It should be noted that the multi-scale image features in this application refer to the feature set obtained by analyzing standardized ultrasound images at different resolution scales, which is used to simultaneously capture different levels of information in the image, from macroscopic anatomical contours to microscopic tissue textures; the endometrial boundary features in this application refer to quantitative parameters used to describe the morphological characteristics at the junction of the endometrium and myometrium; the ovarian follicle distribution features in this application refer to the set of geometric parameters used to quantify the spatial arrangement and developmental status of follicles in the ovary; the myometrial texture features in this application refer to statistical quantities used to characterize the uniformity and structural complexity of the echo pattern of the myometrial tissue, which can quantify the microscopic structural features of the myometrial tissue; the anatomical structure feature set in this application refers to a unified structured data set formed by integrating the quantitative features of the endometrium, ovary, and myometrium, which is used to provide a complete data foundation containing multiple anatomical structures and multidimensional features for the analysis of the functional status of reproductive organs.

[0029] In some embodiments, reference Figure 2 As shown, this figure is an exemplary flowchart of determining a structured segmentation map in some embodiments of this application. In this embodiment, generating a structured segmentation map of the key anatomical structures of the target subject based on the anatomical feature set can be achieved by the following steps: First, in step 1021, a pixel-level classification feature vector is constructed based on the anatomical structure feature set; Secondly, in step 1022, the tissue category of each pixel in the standardized ultrasound image is classified according to the pixel-level classification feature vector; Finally, in step 1023, a structured segmentation map of the key anatomical structures of the target subject is generated from the classification results.

[0030] In specific implementation, the pixel-level classification feature vector based on the anatomical structure feature set can be constructed in the following way: First, the endometrial boundary features, ovarian follicle distribution features, and myometrial texture features included in the anatomical structure feature set can be spatially aligned with the pixel coordinates of the standardized ultrasound image; then, taking each pixel as the center, the statistics of multi-scale features in its local neighborhood are extracted, including the mean, standard deviation, and gradient direction histogram, and concatenated with the original anatomical features corresponding to that pixel; finally, the dimensionality of the concatenated high-dimensional features is reduced by principal component analysis to form a pixel-level classification feature vector for classification; wherein, as a preferred embodiment, the local neighborhood can be set as The three scales are 3×3, 7×7, and 15×15. In other implementations, redundant feature dimensions can also be removed using feature selection algorithms; no specific limitation is made here. The tissue category classification of each pixel in the standardized ultrasound image based on the pixel-level classification feature vector can be achieved as follows: a pre-trained fully connected neural network classifier can be used to classify the dimensionality-reduced feature vector of each pixel in the standardized ultrasound image into four categories: endometrium, follicles, myometrium, and background. For example, the pixel-level classification feature vector can first be input into a neural network with two hidden layers, the hidden layer output can be calculated using the ReLU activation function, and finally, the four-class probability scores can be output using the Softmax function. The classification process involves using the category with the highest probability as the predicted tissue category label for each pixel in the standardized ultrasound image, thereby obtaining the pixel-level classification result of the standardized ultrasound image. In a preferred embodiment, the number of nodes in the hidden layer of the neural network can be set to 128 or 64. In other embodiments, random forests or gradient boosting trees can be used instead of the neural network classifier; no specific limitation is made here. The generation of a structured segmentation map of the key anatomical structures of the target subject from the classification result can be achieved as follows: First, the pixel-level classification result can be reconstructed into a label matrix of the same size as the standardized ultrasound image, where each element represents the predicted tissue category number; then, for each... Morphological post-processing is performed on the connected regions of each category. This involves using an opening operation (erosion followed by dilation) to eliminate isolated noise points, and then a closing operation (dilation followed by erosion) to fill small holes. Finally, a boundary tracing algorithm is used to extract the continuous contours of each tissue region, converting them into vector graphics. A map containing the boundaries and labels of different tissue regions is generated according to a preset color coding scheme, serving as a structured segmentation map of the key anatomical structures of the target subject. In a preferred embodiment, the color coding can use red to mark the endometrial boundary, green to mark the follicular region, and blue to mark the myometrial region. In other embodiments, contour extraction can be used instead of the boundary tracing algorithm; this is not a limitation.

[0031] It should be noted that the pixel-level classification feature vector in this application refers to a multi-dimensional feature representation constructed for each pixel in a standardized ultrasound image to distinguish different tissue types. It is used to comprehensively characterize the morphological, textural, and spatial distribution attributes of each pixel and its neighborhood. The tissue category classification in this application refers to the decision-making process of classifying each pixel in a standardized ultrasound image into a predefined tissue category. It is used to achieve accurate differentiation of endometrium, follicles, myometrium, and background tissues. The classification result in this application refers to the set of labels obtained after completing pixel-level tissue category classification of the standardized ultrasound image. The structured segmentation atlas in this application refers to a visualization atlas containing different tissue boundaries and category labels generated based on the pixel-level classification result. It is used to intuitively display the spatial distribution and morphological characteristics of key anatomical structures such as endometrium, follicles, and myometrium in a machine-readable vector format.

[0032] In step 103, the structured segmentation map is dynamically parameterized to obtain the physiological dynamic parameters of the target subject. The physiological dynamic parameters are then fitted with trends over multiple consecutive time points to obtain the functional state evolution curve of the target subject's reproductive organs.

[0033] In some embodiments, the dynamic parameter quantization of the structured segmentation map to obtain the physiological dynamic parameters of the target subject can be achieved by the following steps: Quantitative indicators of key anatomical structures are extracted from the structured segmentation map; The physiological dynamic parameters of the target subject are calculated based on the quantitative indicators.

[0034] In specific implementation, the quantitative indicators of key anatomical structures extracted from the structured segmentation map can be achieved in the following way: First, calculate the maximum thickness, average thickness, and endometrial area of ​​the endometrial region from the vector boundary of the endometrial region in the structured segmentation map; extract the equivalent diameter, central coordinates, and total number of follicles of each follicle from the follicular region; calculate the contrast, correlation, and entropy value of the gray-level co-occurrence matrix of the myometrium region to obtain the quantitative indicators of key anatomical structures. In a preferred embodiment, the endometrial thickness calculation can be performed using a vertical distance measurement method based on the boundary midline. In other embodiments, the thickness distribution can also be calculated after extracting the endometrial centerline through regional skeletonization; no specific limitations are imposed here. The physiological dynamic parameters of the target subject can be calculated based on the quantitative indicators in the following ways: the maximum endometrial thickness and endometrial area can be used as direct parameters of the degree of endometrial development; the diameter of the dominant follicle and the average diameter of the follicles can be calculated based on the equivalent diameter of all follicles; the follicle distribution density can be calculated based on the spatial coordinates of the follicles; and the three statistical measures of myometrial texture can be combined into a myometrial uniformity index. In a preferred embodiment, the diameter of the dominant follicle is taken as the maximum diameter among all follicles, and the myometrial uniformity index can be calculated by weighted summation of three texture features. In other embodiments, principal component analysis can also be used to fuse multiple texture features into a single index, which is not limited here.

[0035] It should be noted that the quantitative indicators of key anatomical structures in this application refer to the numerical measurement results that reflect the morphological and textural characteristics of key reproductive organs such as the endometrium, ovarian follicles, and myometrium, which are directly measured from the structured segmentation atlas; the physiological dynamic parameters in this application refer to the dynamic characteristic quantities that can characterize the changes in the functional state of the reproductive organs of the target subject over time, and are used to transform static anatomical structure measurements into dynamic indicators that reflect changes in physiological function.

[0036] In some embodiments, the following steps can be used to obtain the functional state evolution curve of the target subject's reproductive organs by performing trend fitting on the physiological dynamic parameters over multiple consecutive time points: Organize physiological dynamic parameters at multiple consecutive time points into time series data; The time series data is subjected to curve fitting processing; Based on the fitting results, the functional state evolution curve of the reproductive organs of the target subject is generated.

[0037] In specific implementation, organizing physiological dynamic parameters from multiple consecutive time points into time series data can be achieved in the following way: The physiological dynamic parameters of the target subject at multiple consecutive time points can be arranged in the order of collection time, namely: endometrial thickness, dominant follicle diameter, follicle distribution density, and myometrial homogeneity index parameters at multiple examination time points within a complete menstrual cycle. Then, the arranged parameter sequence is smoothed, for example, by using a moving average method to eliminate random fluctuations. Finally, the processed parameters are aligned by timestamps to form a regular time series data. In a preferred embodiment, the moving average window can be set to three time points. In other embodiments, a Savitzky-Golay filter can also be used for smoothing; no specific limitation is made here. Curve fitting of the time series data can be achieved in the following way: a cubic spline interpolation algorithm can be used to smoothly fit the time series data, ensuring the continuity of the second derivative of the curve while accurately passing through all valid data points. For the dominant follicle diameter sequence, a Logistic growth model is additionally used for fitting. Its S-shaped growth pattern quantifies the accelerated and stable phases of follicle development. In a preferred embodiment, the cubic spline interpolation can use natural boundary conditions. In other embodiments, non-uniform rational B-splines can be used for more flexible curve fitting; no specific limitation is made here. The functional state evolution curve of the target subject's reproductive organs based on the fitting results can be achieved by superimposing and integrating the fitted endometrial thickness change curve, dominant follicle development curve, follicle distribution density change curve, and myometrial uniformity index curve along a unified time axis. Coordinate axis labels, numerical scales, and key physiological event markers are added to each curve, including ovulation day markers and endometrial transformation period markers. Finally, the complete curve set is output as a vector graphic format, generating a functional state evolution curve that visually displays the dynamic changes in the target subject's reproductive organ function. In a preferred embodiment, the key physiological event markers can use vertical dashed lines to indicate ovulation day, endometrial transformation period, etc. In other embodiments, different background colors can be used to distinguish the physiological stages of the menstrual cycle; no limitation is made here.

[0038] It should be noted that the time series data in this application refers to a regular dataset formed by organizing the physiological dynamic parameters of the target subject at multiple consecutive time points in chronological order; the fitting result in this application refers to the continuous mathematical function and its corresponding curve obtained after curve fitting processing of the time series data, which is used to reveal the inherent laws and trend characteristics of the physiological dynamic parameters changing over time, and to transform discrete measurement data into a function expression that can be continuously analyzed; the functional state evolution curve of the reproductive organs in this application refers to a set of visual curves that integrate the changing trends of multiple physiological dynamic parameters, generated based on the fitting result, which is used to intuitively display the dynamic changes in the functional state of key reproductive structures such as the endometrium, follicles and myometrium of the target subject during the observation period, and can provide a time-series graphical basis for clinical assessment.

[0039] In step 104, the structured segmentation map is fused with the functional state evolution curve of the reproductive organs at the feature level, and then a multi-dimensional assessment matrix of the reproductive health status of the target subject is determined based on the fusion result.

[0040] In some embodiments, reference Figure 3 As shown, this figure is an exemplary flowchart of feature-level fusion in some embodiments of this application. The feature-level fusion of the structured segmentation map and the functional state evolution curve of the reproductive organ can be achieved by the following steps: The spatial morphological feature vector of the target subject's reproductive organs is extracted from the structured segmentation map; The temporal dynamic feature vector of the reproductive organ functional state of the target subject is determined based on the aforementioned functional state evolution curve; The spatial morphological feature vector and the temporal dynamic feature vector are fused to obtain the fusion result.

[0041] In specific implementation, the spatial morphological feature vector of the target subject's reproductive organs can be extracted from the structured segmentation map in the following way: First, the area-to-perimeter ratio and roundness of the endometrial region in the structured segmentation map can be calculated as morphological features, the standard deviation of the centroid coordinates of the follicle region can be statistically analyzed as spatial distribution features, and the texture energy and moment of inertia of the myometrial region can be extracted as texture features; then, the above features are combined into a spatial morphological feature vector; wherein, as a preferred embodiment, the roundness can be calculated by 4π×surface, and in other embodiments, Fourier descriptors can also be used to characterize the shape of the endometrial boundary, which is not specifically limited here; the temporal dynamic feature vector of the target subject's reproductive organ functional state based on the functional state evolution curve can be determined in the following way: the maximum slope and area under the curve of the endometrial thickness change curve in the functional state evolution curve can be calculated, the duration of the growth acceleration phase and the peak value of the stationary phase of the dominant follicle development curve can be extracted, and the myometrial layer can be extracted as a texture feature vector. The coefficient of variation of the layer uniformity index curve is calculated; then the above dynamic parameters are combined into a time-series dynamic feature vector; wherein, as a preferred embodiment, the area under the curve can be calculated using the trapezoidal method, and in other embodiments, the wavelet transform coefficients of the curve can also be extracted as corresponding features, which is not specifically limited here; the spatial morphology feature vector and the time-series dynamic feature vector are fused to obtain the fusion result, which can be achieved in the following way, namely: first, the spatial morphology feature vector and the time-series dynamic feature vector are subjected to min-max normalization to eliminate the difference in dimensions; then, the normalized feature vectors are directly concatenated into a fused feature vector; finally, the dimensionality of the fused feature vector is reduced by principal component analysis to obtain the fused features used to construct a multi-dimensional evaluation matrix, which is used as the fusion result; wherein, as a preferred embodiment, the dimensionality of the reduced feature can be set to 8 dimensions, and in other embodiments, canonical correlation analysis can also be used for feature fusion, which is not limited here.

[0042] It should be noted that the spatial morphological feature vector in this application refers to a set of numerical features that quantifies the spatial structure and morphological characteristics of reproductive organs. It is used to transform the geometric morphology and texture characteristics of the endometrium, follicle distribution, and myometrium into machine-readable feature representations. The temporal dynamic feature vector in this application refers to a set of numerical features that characterizes the changes in physiological parameters over time. It is used to quantify the dynamic evolution of the functional state of reproductive organs. The fusion result in this application refers to a unified feature representation that integrates the spatial structural information and temporal dynamic information of reproductive organs, which can simultaneously reflect changes in morphological characteristics and functional state.

[0043] In some embodiments, determining the multidimensional assessment matrix of the reproductive health status of the target subject based on the obtained fusion results can be achieved by the following steps: A multi-dimensional assessment matrix was determined based on clinical diagnostic criteria. The fusion results are mapped to the corresponding dimensions of the multi-dimensional evaluation matrix; The mapped evaluation indicators are standardized and scored to generate a multi-dimensional evaluation matrix of the reproductive health status of the target examinee.

[0044] In specific implementation, the multi-dimensional assessment matrix determined based on clinical diagnostic criteria can be implemented in the following way: First, four assessment dimensions can be established according to the gynecological ultrasound diagnostic specifications in the clinical diagnostic criteria, such as: endometrial receptivity dimension, follicular development synchronicity dimension, ovarian reserve function dimension, and myometrial health status dimension; 2-3 key assessment indicators are set for each dimension, and the weight coefficients of each indicator are determined; a blank matrix template with assessment dimensions as rows and assessment indicators as columns is constructed as the multi-dimensional assessment matrix; wherein, as a preferred embodiment, the weight coefficients can be determined by the analytic hierarchy process (AHP), and in other embodiments, the weight allocation can also be determined by the expert scoring method, which is not limited here; the fusion result is mapped to the corresponding dimension of the multi-dimensional assessment matrix in the following way: the dimensionality-reduced fusion feature vectors in the fusion result can be allocated to each cell of the multi-dimensional assessment matrix according to a predetermined mapping rule, for example: the features reflecting endometrial thickness changes and morphology are mapped to the endometrial receptivity dimension, the features reflecting follicular development speed and distribution are mapped to the follicular development synchronicity dimension, and the features reflecting follicular number and growth curve are mapped to the dimensionality-reduced fusion feature vectors. The features are mapped to the ovarian reserve function dimension, and the features reflecting the uniformity of myometrial texture are mapped to the myometrial health status dimension. In a preferred embodiment, the mapping rule can be implemented using a pre-trained regression model. In other embodiments, a lookup table method can also be used for feature-to-indicator mapping, which is not limited here. The standardized scoring of the mapped evaluation indicators to generate a multi-dimensional evaluation matrix of the target subject's reproductive health status can be achieved as follows: Each mapped evaluation indicator can be scored using a percentage system, converting the original values ​​to a score of 0-100 based on the clinical normal range; then, the scores of indicators within the same dimension are weighted and summed according to preset weight coefficients to obtain the total dimension score; finally, all dimension scores are filled into the corresponding positions of the evaluation matrix in dimensional order to generate a 4×2 multi-dimensional evaluation matrix containing specific numerical scores, which serves as the multi-dimensional evaluation matrix of the target subject's reproductive health status. In a preferred embodiment, the percentage scoring can be achieved using a piecewise linear transformation method. In other embodiments, a normal distribution percentile method can also be used for score conversion, which is not limited here.

[0045] It should be noted that the corresponding dimension of the assessment matrix in this application refers to the classification unit in the multi-dimensional assessment structure that is directly related to a specific clinical assessment goal. It is used to accurately classify various features in the fusion results into the corresponding clinical assessment categories to ensure the accurate correspondence between features and clinical significance. The mapped assessment index in this application refers to the quantitative parameter assigned to a specific position in the assessment matrix after feature-dimensional mapping processing. It is used to transform abstract fusion features into assessable indicators with clear clinical significance. The multi-dimensional assessment matrix in this application refers to a complete assessment table containing standardized scoring results for each dimension. It is used to comprehensively present the multi-dimensional quantitative assessment results of the reproductive health status of the target examinee in the form of structured data. The dimensions include key clinical dimensions such as endometrial receptivity, follicular development synchronicity, ovarian reserve function, and myometrial health status.

[0046] In step 105, the multi-dimensional evaluation matrix is ​​compared and analyzed with the preset clinical diagnostic threshold to generate a comprehensive diagnostic report for the target subject, which is then sent to the medical analysis terminal for clinical doctors to assist in diagnosis.

[0047] It should be noted that the clinical diagnostic threshold in this application refers to a pre-set critical value used to determine whether each assessment dimension is abnormal. It is used to provide an objective abnormality judgment standard for the scores in the multi-dimensional assessment matrix, realizing the automatic conversion from quantitative assessment to clinical abnormality identification. As a preferred embodiment, the clinical diagnostic threshold can be based on obstetric and gynecological ultrasound diagnostic guidelines and large-scale clinical sample data to set corresponding thresholds for the percentage scores of each assessment dimension in the multi-dimensional assessment matrix. For example, the endometrial receptivity dimension can be set with a score below 60 as an abnormal threshold, the follicular development synchronicity dimension can be set with a score below 55 as a developmental abnormality threshold, the ovarian reserve function dimension can be set with a score above 90 as a high-risk threshold for polycystic changes, and the myometrial health status dimension can be set with a score below 60 as an abnormal threshold. The clinical diagnostic threshold can be adjusted differently according to different age stages such as reproductive age and perimenopause, and a threshold mapping table corresponding to the percentage scoring rules can be established. In other embodiments, the optimal diagnostic threshold can also be determined by receiver operating characteristic curve analysis, or the threshold can be set using the dynamic percentile method. No specific limitation is made here.

[0048] In specific implementation, the comparison and analysis between the multi-dimensional assessment matrix and the preset clinical diagnostic thresholds can be achieved in the following way: First, the scores of each dimension in the multi-dimensional assessment matrix can be compared with the preset clinical diagnostic threshold database to identify abnormal dimensions that exceed the clinical diagnostic thresholds; when at least two dimensions, such as the endometrial receptivity dimension and the follicular development synchronicity dimension, are detected to exceed their corresponding thresholds simultaneously, the report generation engine is triggered; the engine selects the corresponding text module from the predefined diagnostic template library according to the combination pattern of abnormal dimensions, automatically fills in the specific abnormal values ​​and clinical prompts, and generates a structured report containing the target subject's basic information, ultrasound feature description, abnormal dimension analysis, and diagnostic suggestions, i.e., the target subject's comprehensive diagnostic report; finally, the comprehensive diagnostic report is sent to the medical analysis terminal through a medical data exchange protocol for clinicians to assist in diagnosis; other methods can also be used in other embodiments, which are not limited here.

[0049] In another aspect, in some embodiments, this application provides a gynecological ultrasound image analysis system, with reference to... Figure 4 The figure is a schematic diagram of the structure of a gynecological ultrasound image analysis system according to some embodiments of this application. The gynecological ultrasound image analysis system 400 includes: an acquisition module 401, a processing module 402, and an execution module 403, which are described below: The acquisition module 401 in this application is mainly used to acquire the original ultrasound image data of the target subject during the gynecological ultrasound examination, and to preprocess the original ultrasound image data to obtain a standardized ultrasound image. Processing module 402, in this application, is mainly used to perform multi-scale feature extraction on the standardized ultrasound image to obtain an anatomical structure feature set including endometrial boundary features, ovarian follicle distribution features and uterine myometrial texture features, and generate a structured segmentation map of the key anatomical structure of the target subject based on the anatomical structure feature set. The processing module 402 described in this application is also used to perform dynamic parameter quantization on the structured segmentation map to obtain the physiological dynamic parameters of the target subject, and to perform trend fitting on the physiological dynamic parameters at multiple consecutive time points to obtain the functional state evolution curve of the reproductive organs of the target subject. The processing module 402 described in this application is further used to perform feature-level fusion of the structured segmentation map and the functional state evolution curve of the reproductive organs, and then determine the multi-dimensional assessment matrix of the reproductive health status of the target subject based on the obtained fusion result. The execution module 403 in this application is mainly used to compare and analyze the multi-dimensional evaluation matrix with the preset clinical diagnostic threshold, generate a comprehensive diagnostic report of the target subject, and send it to the medical analysis terminal for clinical doctors to assist in diagnosis.

[0050] The foregoing has detailed examples of the gynecological ultrasound image analysis method and system provided in the embodiments of this application. It is understood that the corresponding apparatus, in order to achieve the above functions, includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0051] In some embodiments, this application also provides a computer device, the computer device including a memory and a processor, the memory for storing a computer program, and the processor for calling and running the computer program from the memory, so that the computer device performs the above-described gynecological ultrasound image analysis method.

[0052] In some embodiments, reference Figure 5 The dashed lines in the figure indicate that the unit or module is optional. This figure is a schematic diagram of the structure of a computer device implementing the gynecological ultrasound image analysis method of this application. The gynecological ultrasound image analysis method in the above embodiments can be achieved through… Figure 5 The computer device 500 shown is used to implement this, and the computer device 500 includes at least one processor 501, a memory 502 and at least one communication unit 505. The computer device 500 may be a terminal device, a server or a chip.

[0053] The processor 501 can be a general-purpose processor or a special-purpose processor. For example, the processor 501 can be a central processing unit (CPU). The CPU can be used to control the computer device 500, execute software programs, and process data from the software programs. The computer device 500 may also include a communication unit 505 for inputting (receiving) and outputting (transmitting) signals.

[0054] For example, computer device 500 may be a chip, communication unit 505 may be the input and / or output circuit of the chip, or communication unit 505 may be the communication interface of the chip, and the chip may be a component of terminal device, network device or other device.

[0055] For example, computer device 500 may be a terminal device or a server, and communication unit 505 may be a transceiver of the terminal device or the server, or communication unit 505 may be a transceiver circuit of the terminal device or the server.

[0056] The computer device 500 may include one or more memories 502 storing a program 504. The program 504 can be executed by a processor 501 to generate instructions 503, causing the processor 501 to perform the methods described in the above method embodiments according to the instructions 503. Optionally, the memory 502 may also store data (such as a target audit model). Optionally, the processor 501 may also read data stored in the memory 502, which may be stored at the same storage address as the program 504, or the data may be stored at a different storage address than the program 504.

[0057] The processor 501 and memory 502 can be configured separately or integrated together, for example, integrated on the system-on-chip (SOC) of the terminal device.

[0058] It should be understood that each step of the above method embodiment can be completed by hardware logic circuits or software instructions in the processor 501. The processor 501 can be a CPU, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, such as discrete gates, transistor logic devices, or discrete hardware components.

[0059] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0060] For example, in some embodiments, this application also provides a computer-readable storage medium storing instructions or code that, when executed on a computer, cause the computer to perform the above-described gynecological ultrasound image analysis method.

[0061] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0062] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for analyzing gynecological ultrasound images, characterized in that, Includes the following steps: The raw ultrasound image data of the target subject during the gynecological ultrasound examination is acquired, and the raw ultrasound image data is preprocessed to obtain standardized ultrasound images. Multi-scale feature extraction is performed on the standardized ultrasound images to obtain an anatomical feature set including endometrial boundary features, ovarian follicle distribution features, and uterine myometrial texture features. Based on the anatomical feature set, a structured segmentation atlas of key anatomical structures of the target subject is generated. The structured segmentation map is dynamically parameterized to obtain the physiological dynamic parameters of the target subject. The physiological dynamic parameters are then fitted with trends at multiple consecutive time points to obtain the functional state evolution curve of the target subject's reproductive organs. The structured segmentation map is fused with the functional state evolution curve of the reproductive organs at the feature level, and then a multi-dimensional assessment matrix of the reproductive health status of the target subject is determined based on the fusion result. The multi-dimensional assessment matrix is ​​compared and analyzed with the preset clinical diagnostic thresholds to generate a comprehensive diagnostic report for the target subject, which is then sent to a medical analysis terminal for clinical doctors to assist in diagnosis.

2. The method as described in claim 1, characterized in that, Multi-scale feature extraction was performed on the standardized ultrasound images to obtain a set of anatomical features, including endometrial boundary features, ovarian follicle distribution features, and myometrial texture features. Specifically, this set includes: Extract multi-scale image features from the standardized ultrasound images; Based on the multi-scale image features, the endometrial boundary was identified, the ovarian follicle region was located, and the texture of the myometrium was analyzed. The obtained identification results, localization results, and analysis results are integrated into an anatomical structure feature set.

3. The method as described in claim 1, characterized in that, Generating a structured segmentation atlas of key anatomical structures of the target subject based on the aforementioned anatomical feature set specifically includes: Pixel-level classification feature vectors are constructed based on the anatomical structure feature set; Based on the pixel-level classification feature vector, tissue category classification is performed on each pixel in the standardized ultrasound image; The classification results generate a structured segmentation map of the key anatomical structures of the target subject.

4. The method as described in claim 1, characterized in that, The structured segmentation map is dynamically quantified to obtain the target subject's physiological dynamic parameters, specifically including: Quantitative indicators of key anatomical structures are extracted from the structured segmentation map; The physiological dynamic parameters of the target subject are calculated based on the quantitative indicators.

5. The method as described in claim 1, characterized in that, The physiological dynamic parameters are fitted to trends over multiple consecutive time points to obtain the functional state evolution curve of the target subject's reproductive organs, specifically including: Organize physiological dynamic parameters at multiple consecutive time points into time series data; The time series data is subjected to curve fitting processing; Based on the fitting results, the functional state evolution curve of the reproductive organs of the target subject is generated.

6. The method as described in claim 1, characterized in that, The feature-level fusion of the structured segmentation map and the functional state evolution curve of the reproductive organ specifically includes: The spatial morphological feature vector of the target subject's reproductive organs is extracted from the structured segmentation map; The temporal dynamic feature vector of the reproductive organ functional state of the target subject is determined based on the aforementioned functional state evolution curve; The spatial morphological feature vector and the temporal dynamic feature vector are fused to obtain the fusion result.

7. The method as described in claim 1, characterized in that, Based on the obtained fusion results, the multi-dimensional assessment matrix for the reproductive health status of the target subjects is determined, which specifically includes: A multi-dimensional assessment matrix was determined based on clinical diagnostic criteria. The fusion results are mapped to the corresponding dimensions of the multi-dimensional evaluation matrix; The mapped evaluation indicators are standardized and scored to generate a multi-dimensional evaluation matrix of the reproductive health status of the target examinee.

8. A gynecological ultrasound image analysis system, characterized in that, include: The acquisition module is used to acquire the original ultrasound image data of the target subject during the gynecological ultrasound examination, and to preprocess the original ultrasound image data to obtain standardized ultrasound images. The processing module is used to perform multi-scale feature extraction on the standardized ultrasound image to obtain an anatomical structure feature set including endometrial boundary features, ovarian follicle distribution features and uterine myometrial texture features, and to generate a structured segmentation atlas of key anatomical structures of the target subject based on the anatomical structure feature set. The processing module is also used to perform dynamic parameter quantization on the structured segmentation map to obtain the physiological dynamic parameters of the target subject, and to perform trend fitting on the physiological dynamic parameters over multiple consecutive time points to obtain the functional state evolution curve of the target subject's reproductive organs. The processing module is also used to perform feature-level fusion of the structured segmentation map and the functional state evolution curve of the reproductive organs, and then determine the multi-dimensional assessment matrix of the reproductive health status of the target subject based on the obtained fusion result. The execution module is used to compare and analyze the multi-dimensional evaluation matrix with the preset clinical diagnostic threshold, generate a comprehensive diagnostic report for the target subject, and send it to the medical analysis terminal for clinical doctors to assist in diagnosis.

9. A computer device, characterized in that, The computer device includes a memory and a processor, the memory being used to store computer programs, and the processor being used to call and run the computer programs from the memory, causing the computer device to perform the gynecological ultrasound image analysis method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions or code that, when executed on a computer, cause the computer to perform the gynecological ultrasound image analysis method as described in any one of claims 1 to 7.

Images