Endometrial receptivity assessment system based on transvaginal high frequency ultrasound imaging

Abstract

The present application relates to the technical field of medical data processing, in particular to a kind of endometrial receptivity evaluation system based on vaginal high-frequency ultrasound imaging.The system extracts blood flow pulsation rhythmicity coefficient through time series ultrasound echo image, generates local microvessel perfusion resistance in combination with brightness characteristics;Based on endometrial anatomical profile, construct vertical growth direction field, introduce direction difference penalty when calculating adjacent position transmission cost, construct anisotropic resistance network;Based on the principle of minimum energy consumption, construct supply and demand bipartite graph model, obtain effective blood perfusion path by global optimal matching solution, and output overall perfusion resistance and endometrial non-perfusion degree accordingly.The present application can effectively distinguish pathological chaotic blood flow from physiological effective perfusion, improve the specificity and accuracy of endometrial receptivity evaluation.

Description

Technical Field

[0001] This invention relates to the field of medical data processing technology, specifically to an endometrial receptivity assessment system based on vaginal high-frequency ultrasound imaging. Background Technology

[0002] Endometrial receptivity refers to the endometrium's ability to accept an embryo, and its microcirculatory state directly determines the success rate of embryo implantation. Current clinical assessments primarily rely on Doppler ultrasound blood flow imaging, calculating the area or intensity integral of the blood flow signal to infer endometrial function. However, this intensity-based assessment method has significant limitations in assisted reproductive clinical practice. Pathological inflammatory congestion or arteriovenous fistulas also present with high-intensity blood flow signals, but these vessels are often structurally disordered and lack regular pulsation, failing to effectively support embryo implantation and leading to false positives in the assessment results. Furthermore, functional spiral arteries of the endometrium have specific anatomical characteristics, namely, they must grow vertically upwards from the endometrial-myolipin junction to the uterine cavity surface. Existing whole-domain integration methods ignore this crucial anatomical constraint of vessel growth direction, failing to distinguish between effective vertical perfusion and ineffective lateral turbulence. Summary of the Invention

[0003] To address the problem of high false positives in endometrial receptivity assessment caused by the inability of existing ultrasound assessments to distinguish between high-intensity, chaotic pathological blood flow and effective physiological perfusion, this invention aims to provide an endometrial receptivity assessment system based on high-frequency vaginal ultrasound imaging. The specific technical solution adopted is as follows:

[0004] This invention proposes an endometrial receptivity assessment system based on vaginal high-frequency ultrasound imaging, the system comprising:

[0005] The time-domain ultrasound image feature extraction module is used to acquire continuous frames of endometrial ultrasound echo images; based on the brightness change characteristics of each location in the image over time, the blood flow pulsation rhythm coefficient of each location is obtained; based on the blood flow pulsation rhythm coefficient and the brightness characteristics of each location, the local microvascular perfusion resistance of each location is obtained.

[0006] The blood flow cost quantification module is used to take the direction between the location and the set of required locations as the anatomical vertical growth direction for any location; for two adjacent locations, the difference between the adjacent locations and the anatomical vertical growth direction, and the local microvascular perfusion resistance at the corresponding locations are used to obtain the single-step perfusion resistance cost between the two adjacent locations and form a cost matrix.

[0007] The blood flow path matching analysis module is used to filter out blood perfusion paths based on the path cost between the supply and demand locations of blood perfusion in the cost matrix. The blood perfusion paths form a bipartite graph. The module performs matching and solving on the bipartite graph to obtain the optimal perfusion matching set.

[0008] The evaluation index output module is used to output the overall endometrial perfusion resistance based on the path cost corresponding to the optimal perfusion matching set; and to output the degree of endometrial non-perfusion based on the number of unmatched demand locations.

[0009] Furthermore, the method for obtaining the blood flow pulsation rhythm coefficient includes:

[0010] The cardiac cycle frequency is obtained by measuring the temporal variation of the average brightness of the endometrial ultrasound echo image; the brightness sequence at each location is filtered based on the cardiac cycle frequency to obtain a filtered sequence; the blood flow pulsation rhythm coefficient is obtained by comparing the filtered sequence with the brightness sequence.

[0011] Furthermore, the method for obtaining the cardiac cycle frequency includes:

[0012] The average brightness of each frame of endometrial ultrasound echo image is arranged in chronological order to form an average brightness sequence; the average brightness sequence is subjected to a fast Fourier transform, and the frequency with the largest amplitude in the preset conventional heart rate frequency band is searched as the cardiac cycle frequency.

[0013] Further, the step of comparing the filtered sequence with the luminance sequence to obtain the blood flow pulsatility coefficient includes:

[0014] The ratio of the total energy of the filtered sequence as the numerator and the total energy of the brightness sequence as the denominator is the blood flow pulsation rhythm coefficient.

[0015] Furthermore, the method for obtaining the local microvascular perfusion resistance includes:

[0016] The overall blood perfusion volume is obtained based on the brightness characteristics of each location in consecutive frames of endometrial ultrasound echo images. The local microvascular perfusion resistance is obtained by multiplying the blood flow pulsation rhythm coefficient by the overall blood perfusion volume and performing negative correlation mapping.

[0017] Furthermore, the method for obtaining the total blood perfusion volume includes:

[0018] The average value of the elements in the brightness sequence is taken as the overall blood perfusion volume.

[0019] Furthermore, the method for obtaining the single-step infusion resistance cost includes:

[0020] Obtain a direction vector pointing from one location to an adjacent location, and the cosine value of the deviation angle between the vector and the anatomical perpendicular growth direction; obtain the single-step perfusion resistance cost based on the local microvascular perfusion resistance at the pointed location, the cosine value of the deviation angle, and the Euclidean distance between the two locations.

[0021] Furthermore, the method for screening the blood perfusion pathway includes:

[0022] For any combination of supply and demand locations, the optimal path of the combination in the cost matrix is ​​obtained, and the cumulative value of the single-step injection resistance cost corresponding to the optimal path is normalized to obtain the path cost of the optimal path.

[0023] Among all combinations, the optimal path corresponding to the combination whose path cost is less than a preset cost threshold is taken as the blood perfusion path.

[0024] Furthermore, the method for obtaining the overall endometrial perfusion resistance includes:

[0025] The average path cost of the optimal perfusion matching set is used as the overall perfusion resistance of the endometrium.

[0026] Furthermore, the method for obtaining the degree of endometrial nonperfusion includes:

[0027] If the number of required locations is 1, then the degree of incomplete perfusion of the endometrium will not be output;

[0028] If the number of required locations is greater than 1, the proportion of unmatched required locations in the total number of required locations is taken as the degree of incomplete perfusion of the endometrium.

[0029] The present invention has the following beneficial effects:

[0030] This invention first extracts the rhythmic coefficient of blood flow pulsation and utilizes the physical characteristic of normal microcirculation being modulated by the cardiac cycle to identify and suppress venous congestion or noise signals lacking regular pulsation, thus preliminarily quantifying local microvascular perfusion resistance in the time domain. Furthermore, this application innovatively introduces the anatomical vertical growth direction as a spatial constraint. When calculating the cost of single-step perfusion resistance between adjacent locations, the difference in direction from the anatomical direction is used to penalize the transmission path, causing pathological vessels with disordered orientations and not conforming to the vertical growth characteristics of spiral arteries (such as arteriovenous fistulas) to appear as high-resistance pathways in the cost matrix. Finally, a bipartite graph is constructed based on this cost matrix, and the optimal perfusion matching set is solved, simulating the optimal flow process of blood under the principle of minimum energy consumption. This automatically filters out the high-resistance ineffective paths marked in the aforementioned steps, retaining only the functionally intact microcirculation network. The resulting overall endometrial perfusion resistance and endometrial non-perfusion degree can accurately quantify the patency and coverage integrity of effective blood supply, providing a new, highly specific, and pathologically resistant indicator for endometrial function assessment in clinical practice. This invention can effectively distinguish between pathological states with high-intensity blood flow signals and true physiological effective perfusion, thereby enabling effective assessment of endometrial receptivity. Attached Figure Description

[0031] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a structural block diagram of an endometrial receptivity assessment system based on vaginal high-frequency ultrasound imaging, provided as an embodiment of the present invention. Detailed Implementation

[0033] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an endometrial receptivity assessment system based on vaginal high-frequency ultrasound imaging proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0035] The following description, in conjunction with the accompanying drawings, details the specific scheme of the endometrial receptivity assessment system based on vaginal high-frequency ultrasound imaging provided by the present invention.

[0036] Please see Figure 1 The diagram illustrates a structural block diagram of an endometrial receptivity assessment system based on vaginal high-frequency ultrasound imaging according to an embodiment of the present invention. The system includes: a time-domain ultrasound image feature extraction module 101, a blood flow cost quantification module 102, a blood flow path matching analysis module 103, and an assessment index output module 104.

[0037] The time-domain ultrasound image feature extraction module 101 is used to extract the local microvascular perfusion resistance at each location from multiple consecutive frames of endometrial ultrasound echo images in the time domain. This means that the time-varying three-dimensional image signal is collapsed into a two-dimensional static resistance distribution, thereby quantifying the basic patency of microcirculation at various points in the endometrium.

[0038] It should be noted that the ultrasound images analyzed in this embodiment of the invention are acquired using vaginal high-frequency ultrasound imaging technology. The information acquired by the probe should be a time-series ultrasound echo data volume containing multiple complete cardiac cycles of the subject. This data volume is a three-dimensional matrix with dimensions defined as H×W×T, where H and W represent the spatial height and width of the ultrasound section, respectively, and T represents the number of frames continuously acquired along the time axis. During transvaginal ultrasound examination, minute probe displacement or the subject's respiratory movements can cause the same spatial coordinates to correspond to different anatomical tissue points in different time frames. This spatial misalignment introduces motion artifacts, severely interfering with the subsequent extraction of blood flow pulsation features. Therefore, a unified spatial reference system must be established using tissue texture features. Furthermore, the original data volume contains other non-interested regions outside the endometrium, requiring further segmentation and preprocessing to obtain the endometrial ultrasound echo image. The time-domain ultrasound image feature extraction module 101 also includes an ultrasound data preprocessing unit, which specifically performs the following steps:

[0039] (1) The original data volume acquired by the probe is separated into a grayscale channel data volume and a color Doppler blood flow channel data volume. The intermediate time frame is selected as the reference frame for anatomical dissection. It should be noted that in other implementations of this invention, the image similarity between each frame and other frames can also be calculated in the grayscale channel data volume, and the frame with the highest overall image similarity can be selected as the reference frame for anatomical dissection. The specific operation is a well-known technique known to those skilled in the art, and will not be described or limited here.

[0040] (2) For each frame in the time sequence, the current frame and the reference frame are placed in the grayscale channel data volume, and the two-dimensional plane translation displacement between the corresponding images of the two frames is calculated using the phase correlation algorithm. The phase correlation algorithm can also be replaced by the well-known rigid registration technique, which will not be elaborated or limited in detail.

[0041] (3) Apply the inverse vector of the translation displacement corresponding to the current frame to the color Doppler blood flow channel data volume of the current frame, and resample and correct the pixel coordinates contained therein. Obtain a corrected data volume with unified geometric coordinates.

[0042] (4) The endometrial boundary is identified using a known image segmentation algorithm to generate a mask for the functional layer region of the endometrium. This mask is then used to segment the corrected data volume to obtain an endometrial ultrasound echo image for each frame. It should be noted that the image segmentation methods used to design the endometrial functional layer region mask are all techniques known to those skilled in the art. In other implementations of this invention, models such as neural networks can also be used for identification, or a preset mask can be used directly for identification. These are all techniques well-known to those skilled in the art and will not be elaborated upon here.

[0043] After obtaining consecutive frames of endometrial ultrasound echo images, the temporal ultrasound image feature extraction module 101 can reflect the blood flow intensity characteristics at each location in each frame. Effective endometrial microcirculation perfusion is characterized by regular pulsations driven by cardiac pumping, while background noise or venous congestion usually manifests as chaotic or constant signals. Therefore, for each location, statistical analysis of the temporal brightness variation characteristics yields the blood flow pulsation rhythm coefficient for each location. This coefficient quantifies the degree to which the blood flow signal is modulated by heart rate at the corresponding location; a larger coefficient indicates a greater modulation depth at that location, suggesting significant effective blood flow perfusion. Conversely, for arteriovenous fistulas or inflammatory congestion, the obtained blood flow pulsation rhythm coefficient is significantly smaller, indicating inefficient perfusion and greater blood flow resistance at that location. Based on this principle, the local microvascular perfusion resistance at each location can be obtained by further combining the blood flow pulsation rhythm coefficient with the brightness variation characteristics at each location. A greater resistance indicates an inefficient perfusion area at that location.

[0044] Preferably, in this embodiment of the invention, in order to effectively analyze the blood flow pulsation rhythm coefficient, considering that blood flow signals will generate significant and highly referential information at specific physiological frequencies of the subject, this embodiment first obtains the cardiac cycle frequency based on the temporal variation of the average brightness of the endometrial ultrasound echo image. The information at this cardiac cycle frequency is the analyzable and highly referential blood flow signal intensity information. Therefore, the brightness sequence at each location is filtered based on the cardiac cycle frequency. The larger the amplitude of the filtered sequence, the more significant the subject's pulsation amplitude at the cardiac cycle frequency, indicating a more regular blood flow pulsation. Therefore, the blood flow pulsation rhythm coefficient can be obtained by comparing the filtered sequence with the brightness sequence. That is, the larger the proportion of the filtered sequence's capability, the higher the modulation depth at that location, and the more regular the blood flow perfusion. Conversely, in cases such as arteriovenous fistulas or inflammatory congestion, because the blood flow is chaotic, the signal energy is not concentrated near the cardiac cycle frequency, so the obtained blood flow pulsation rhythm coefficient is significantly smaller. In other words, the blood flow pulsation rhythm coefficient can effectively quantify the degree to which blood flow signals are modulated by heart rate.

[0045] In this embodiment of the invention, the cardiac cycle frequency is used as the center frequency, and the bandwidth is set to 0.2 Hz to construct a bandpass filter to filter the luminance sequence. The specific filtering method is a technique well known to those skilled in the art and will not be described in detail here.

[0046] Furthermore, methods for obtaining cardiac cycle frequency include:

[0047] The average brightness of each frame of endometrial ultrasound echo image is arranged in chronological order to form an average brightness sequence; a fast Fourier transform is performed on the average brightness sequence, and the frequency with the largest amplitude in a preset conventional heart rate frequency band is searched as the cardiac cycle frequency. The preset conventional heart rate frequency band is prior knowledge and can be set to 0.5 to 3.0 Hz.

[0048] Furthermore, in a specific implementation of this invention, the specific quantification method of the blood flow pulsation rhythm coefficient includes: using the total energy of the filtered sequence as the numerator and the total energy of the brightness sequence as the denominator, the ratio obtained is the blood flow pulsation rhythm coefficient. That is, the larger the ratio, the closer the overall energy is to the distribution near the cardiac cycle frequency, indicating that the subject's endothelial blood flow pulsation rhythm is stronger and the blood perfusion is more regular.

[0049] Preferably, in one embodiment of the present invention, the method for obtaining local microvascular perfusion resistance includes:

[0050] The overall blood perfusion volume was obtained based on the brightness characteristics of each location in consecutive frames of endometrial ultrasound echo images. Regarding the characteristic of local microvascular perfusion resistance, a larger overall blood perfusion volume indicates lower resistance at the corresponding location, suggesting smoother perfusion.

[0051] Furthermore, considering that a larger blood flow pulsation rhythm coefficient indicates a significant rhythmicity in blood flow at that location, consistent with normal physiological blood flow patterns, it should correspond to lower perfusion resistance. Therefore, overall blood perfusion volume should be negatively correlated with local microvascular perfusion resistance, and the blood flow pulsation rhythm coefficient should also be negatively correlated with this resistance.

[0052] Therefore, by multiplying the blood flow pulsation rhythm coefficient by the overall blood perfusion volume and performing a negative correlation mapping, the local microvascular perfusion resistance can be obtained.

[0053] As an example, in a specific implementation of this invention, a preset resistance scaling constant is used as the numerator, and the product of the blood flow pulsation rhythm coefficient and the total blood perfusion volume, plus a preset zero-prevention coefficient, is used as the denominator to obtain the local microvascular perfusion resistance. The purpose of the resistance scaling constant is to balance the final value range, ensuring that the obtained resistance value has a significant numerical dominance in subsequent cost calculations. This resistance scaling constant can be set between 255 and 2000 in 8-bit depth image data; in this embodiment, it is specifically set to 1000. The preset zero-prevention coefficient aims to avoid a denominator of 0 and can be set to 0.1% of the maximum image bit depth, i.e., 0.255 in 8-bit depth image data.

[0054] Furthermore, the average value of the elements in the brightness sequence at each location is taken as the overall blood perfusion volume. That is, this process is based on the properties of the ultrasound image itself; the stronger the brightness of a location in the ultrasound image, the stronger the blood perfusion component at that location, and therefore it can be quantified by averaging.

[0055] Finally, the temporal ultrasound image feature extraction module 101 can extract features from multiple consecutive frames of endometrial ultrasound echo images to obtain a two-dimensional local microvascular perfusion resistance map. This resistance map records the local microvascular perfusion resistance at each location. The resistance map is stripped of the time variable and only retains the biophysical transmission properties of each point in space.

[0056] The blood flow cost quantification module 102 further analyzes the anatomical relationships at various locations on the endometrium. Supply and demand locations, as well as other locations, can be extracted from the endometrial functional layer region mask. Demand locations represent the physiological endpoint of blood perfusion, supply locations represent the physiological starting point of spiral arteries penetrating the endometrium, and other locations may be process locations during blood perfusion. Therefore, the direction pointing from each location to the set of demand locations can be taken as the anatomical vertical growth direction, which is the directional constraint for each location. In the subsequent analysis of blood perfusion path costs, a penalty for vascular growth direction deviation needs to be introduced. Therefore, for two adjacent locations, assuming blood perfusion has occurred, the single-step perfusion resistance cost between the two adjacent locations can be obtained based on the directional difference between the adjacent locations and the anatomical vertical growth direction, as well as the local microvascular perfusion resistance at the corresponding location. This cost, based on the local microvascular perfusion resistance, introduces directional differences, which can further ensure the accuracy of cost calculation and the effectiveness of subsequent path analysis from a physiological and anatomical perspective. This allows pathological vessels with disordered pathways and that do not conform to the vertical growth characteristics of spiral arteries (such as arteriovenous fistulas) to appear as high-resistance pathways in the cost matrix. Finally, the combination of all adjacent positions forms a cost matrix. The horizontal axis of the cost matrix represents the number of all positions, which can be regarded as the starting point of perfusion; the vertical axis also represents the number of all positions, which can be regarded as the ending point of perfusion. It should be noted that since blood perfusion only occurs at adjacent positions, the corresponding elements in the matrix for non-adjacent positions should be set to infinity. Similarly, the elements pointing from a position to a perfusion position should also be infinity, and the elements pointing from a position to itself should also be infinity, thus obtaining the final cost matrix.

[0057] In this embodiment of the invention, considering that ultrasound echo images have both transverse and longitudinal sections, for the transverse section, the endometrial functional layer region is a concentric circle, with the endometrium enveloping the uterine cavity in a ring shape. Therefore, the demand end of the transverse section is the center point, and there is one and only one demand position. The longitudinal section is different; its demand positions are a set, which is the topological centerline of the endometrial functional layer region mask. Therefore, this embodiment of the invention uses the position corresponding to the outer contour of the mask as the supply position. If the current endometrial ultrasound echo image set is a transverse data volume, then the center point of the mask is used as the demand position, that is, the other position with the largest average distance from all supply positions is selected as the demand position. If it is a longitudinal data volume, a morphological thinning algorithm is used to use the set of positions on the topological centerline of the mask as the demand position set, and the intersection of the centerline and the outer contour is removed. This embodiment of the invention considers this intersection to be a non-blood source area such as the cervix, and therefore it needs to be removed and not used as a supply position. It should be noted that the morphological thinning algorithm in this embodiment of the invention specifically adopts the Zhang-Suen skeletonization algorithm.

[0058] In this embodiment of the invention, the method for obtaining the anatomical vertical growth direction includes: calculating the distance between the target location and the set of required locations for all locations except the required location. It should be noted that if the set of required locations is the centerline formed by the longitudinal section, the distance can be directly calculated using the distance formula between points and lines; if the set of required locations is a single required location formed by the transverse section, the Euclidean distance between the two points is directly calculated. Finally, a distance can be obtained for each location except the required location, thus forming a distance scalar field. This distance scalar field is a two-dimensional matrix with the same size as the image, where each element represents the minimum distance between the corresponding pixel and the set of required locations. The gradient of this distance field is then calculated, and normalized by taking the inverse direction of the gradient to obtain a vector field. Each vector in the vector field represents the direction from the corresponding pixel to the set of required locations, which is the anatomical vertical growth direction. It should be noted that the gradient calculation can be performed using existing gradient analysis algorithms such as the Sobel operator; the specific operations are well-known techniques to those skilled in the art and will not be elaborated upon here.

[0059] Preferably, in this embodiment of the invention, the method for obtaining the cost of single-step infusion resistance includes:

[0060] Obtain the cosine of the deviation angle between a direction vector pointing from one location to an adjacent location and the anatomical growth direction perpendicular to it. This cosine deviation angle can be calculated by the dot product of the two vectors. The closer the cosine deviation angle is to 1, the more likely the local propagation path conforms to the anatomical growth direction. If the cosine deviation angle is less than or equal to 0, it indicates a laterally disordered or retrograde pathological structure. In other words, the larger the cosine deviation angle, the lower the corresponding cost.

[0061] The single-step perfusion resistance cost is obtained based on the local microvascular perfusion resistance at the adjacent supply locations, the cosine value of the deviation angle, and the Euclidean distance between the two locations. That is, the greater the local microvascular perfusion resistance, the smaller the cosine value of the deviation angle, and the larger the Euclidean distance, the greater the single-step perfusion resistance cost of this local propagation path between the two locations.

[0062] As an example, in a specific implementation of this invention, the cost of single-step infusion resistance is expressed by the formula:

[0063] ;in For a local propagation path, the location point to The cost of single-step infusion resistance, The location being pointed to The local microvascular perfusion resistance, exp() is an exponential function with the natural constant as its base. As an anisotropic penalty factor, The cosine of the deviation angle between the two positions. This represents the Euclidean distance between the two locations.

[0064] In the above formula, the anisotropy penalty factor determines the model's sensitivity to directional deviation. A larger value indicates stronger inhibition of lateral blood vessels, meaning a higher cost. The value can be set between 2 and 5; in this embodiment, it is set to 3.5. It should be noted that in this embodiment, adjacent positions refer to the eight-neighborhood of a given position. Therefore, there are two cases: linearly adjacent and diagonally adjacent. Thus, the Euclidean distance for linearly connected positions is defined as 1, and the Euclidean distance for diagonally adjacent positions is defined as... This formula establishes a positive-negative correlation, while simultaneously demonstrating that the transmission cost of any path deviating from the vertical direction increases exponentially, thus mathematically simulating the high fluid resistance characteristics of ineffective blood vessels.

[0065] The blood flow path matching analysis module 103 further analyzes the cost matrix. Since the cost matrix constructs the blood perfusion cost between each pair of adjacent positions, the supply position can be defined as the starting point and the demand position as the ending point. Effective blood perfusion paths are selected based on the path cost between the supply and demand positions. A bipartite graph is then constructed to simulate the optimal blood flow process under the principle of minimum energy consumption, automatically filtering out high-resistance invalid paths marked in the previous steps and retaining only functional microcirculation networks. Solving the matching problem of this bipartite graph allows for the optimization of the cost of all matching edges while satisfying as many demand positions as possible, outputting the optimal perfusion matching set. This optimal perfusion matching set contains multiple confirmed effective microcirculation blood supply pathways that conform to anatomical growth characteristics.

[0066] Preferably, in this embodiment of the invention, the method for screening blood perfusion pathways includes:

[0067] For any combination of supply and demand locations, the optimal path of the combination in the cost matrix is ​​obtained. In this embodiment of the invention, since the cost matrix can be regarded as a directed graph, for any combination of supply and demand locations, an optimal path from one supply location to one demand location can be found in the directed graph by using the multi-source Dijkstra algorithm or the heap-optimized fast traversal algorithm, that is, the path with the minimum sum of single-step injection resistance costs is obtained, thus obtaining the optimal path for each supply-demand location combination.

[0068] The cumulative value of the single-step perfusion resistance cost corresponding to the optimal path is normalized to obtain the path cost of the optimal path. In this embodiment of the invention, the purpose of the normalization process is to perform geometric normalization, which aims to eliminate the influence of differences in endometrial thickness among different subjects on the dimension of the resistance value. In this embodiment of the invention, the cumulative value of the single-step perfusion resistance cost is used as the numerator, and the linear Euclidean distance between the supply and demand positions is used as the denominator to obtain the final path cost. It should be noted that because the attack position and the demand position are not the same location, the denominator of the above ratio will not be 0.

[0069] It should be noted that since some elements in the cost matrix are infinite, representing "unreachable" situations, if the final optimal path contains this information, it means that the supply-demand location is an unreachable disconnect, and therefore the final path cost also needs to be set to infinite.

[0070] The path cost of all potential connecting paths from the junctional zone to the uterine cavity centerline is included in all combinations. Some paths have extremely high computational costs due to significant deviations from the anatomical growth direction or passage through high-resistance regions, indicating that they cannot physiologically form effective blood perfusion. To simulate the vascular pruning mechanism of organisms, a threshold needs to be set to eliminate these invalid paths. Therefore, the optimal path corresponding to the combination whose path cost is less than the preset cost threshold is taken as the blood perfusion path.

[0071] In this embodiment of the invention, to accommodate individual differences among different subjects, an adaptive statistical method is used to define the maximum effective perfusion resistance threshold. Non-infinite elements from all combinations are extracted to form a cost set, and the median and interquartile range of this cost set are obtained. A cost threshold is then set. for: ;in The median, Interquartile range, The leniency coefficient is set, and in this embodiment of the invention, the leniency coefficient is set to a range of 1.5 to 3.0, and in this embodiment, it is set to 2. The final threshold dynamically defines the upper limit of the physiologically acceptable microcirculation resistance for the current subject. Any resistance value exceeding this threshold will result in a threshold being set for the subject to be removed. All paths are considered functionally disconnected.

[0072] In this embodiment of the invention, the bipartite graph can be solved using a minimum-cost maximum matching algorithm (such as the Hungarian algorithm or the continuous shortest path algorithm). The goal is to minimize the sum of the weights of all matching edges while satisfying as many demand nodes as possible. It should be noted that in the microcirculation, a thick spiral artery (a supply location) may branch, simultaneously supplying blood to multiple demand locations in the uterine cavity center. This presents a one-to-many relationship, and bipartite graph matching in this one-to-many situation can produce significant false negatives because the supply locations are fewer than the demand locations, leading to false alarms. For example, if there are 50 supply locations and 100 demand locations, a conventional algorithm can only obtain 50 matching pairs, resulting in false alarms. Therefore, before solving the bipartite graph, the matrix corresponding to the bipartite graph is expanded. For example, if the original matrix is ​​30×100, meaning there are 30 supply positions and 100 demand positions, then each row is copied 4 times to obtain a new matrix of 120×100. Then, matching is performed. After matching, these 120 virtual points are mapped back to the original 30 supply positions to realize the physical representation of a blood vessel supplying blood to multiple points.

[0073] The optimal perfusion matching set stores multiple confirmed and effective microcirculatory blood supply pathways that conform to anatomical growth characteristics. Simultaneously, the path cost represents the resistance characteristics during blood perfusion. Therefore, the evaluation index output module 104 statistically analyzes the path costs of these pathways to output the overall endometrial perfusion resistance. This index reflects the average patency of blood transport within the area where microcirculation has been successfully established. A lower value indicates that the vascular network structure is more anatomically aligned, and the pulsation is more regular, facilitating the delivery of nutrients to the uterine cavity surface.

[0074] Further analysis of the number of unmatched demand locations yields the degree of intimal non-perfusion, a quantifiable indicator of the extent of blood supply blind spots in the functional layer of the endometrium. Higher values ​​indicate larger areas lacking effective blood supply due to vascular structural disorder (high-resistance vessels being eliminated) or vascular absence, suggesting a higher risk of embryo implantation failure.

[0075] The system proposed in this invention ultimately outputs the overall endometrial perfusion resistance and the degree of endometrial non-perfusion. This indicator can be displayed to the physician through printing or visualization, making it easier for the physician to judge the microenvironmental state of the patient's endometrium.

[0076] Preferably, in an embodiment of the present invention, the average path cost of the optimal perfusion matching set is used as the overall perfusion resistance of the endometrium.

[0077] Preferably, in this embodiment of the invention, considering the case that the data volume has a cross-sectional data volume and a longitudinal data volume, wherein there is only one required position in the cross-sectional data volume, the degree of incomplete perfusion of the endometrium cannot be obtained. Therefore, if the number of required positions is 1, the degree of incomplete perfusion of the endometrium is not output. That is, although the degree of incomplete perfusion of the endometrium can be calculated to be 0 at this time, it is not used as the system output.

[0078] If the number of required locations is greater than 1, the proportion of unmatched required locations in the total number of required locations is taken as the degree of incomplete perfusion of the endometrium.

[0079] In summary, this invention extracts the rhythmic coefficient of blood flow pulsation from time-series ultrasound echo images and generates local microvascular perfusion resistance by combining it with brightness features. Based on the endometrial anatomical contour, a vertical growth direction field is constructed, and a directional difference penalty is introduced when calculating the transmission cost at adjacent locations to build an anisotropic resistance network. A supply-demand bipartite graph model is constructed based on the principle of minimum energy consumption, and an effective blood perfusion path is obtained through global optimal matching. Based on this, the overall perfusion resistance and the degree of endometrial non-perfusion are quantitatively output. This invention can effectively distinguish between pathologically disordered blood flow and physiologically effective perfusion, improving the specificity and accuracy of endometrial receptivity assessment.

[0080] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0081] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A system for assessing endometrial receptivity based on vaginal high-frequency ultrasound imaging, characterized in that, The system includes: The time-domain ultrasound image feature extraction module is used to acquire continuous frames of endometrial ultrasound echo images; based on the brightness change characteristics of each location in the image over time, the blood flow pulsation rhythm coefficient of each location is obtained; based on the blood flow pulsation rhythm coefficient and the brightness characteristics of each location, the local microvascular perfusion resistance of each location is obtained. The blood flow cost quantification module is used to take the direction between the location and the set of required locations as the anatomical vertical growth direction for any location; for two adjacent locations, the difference between the adjacent locations and the anatomical vertical growth direction, and the local microvascular perfusion resistance at the corresponding locations are used to obtain the single-step perfusion resistance cost between the two adjacent locations and form a cost matrix. The blood flow path matching analysis module is used to filter out blood perfusion paths based on the path cost between the supply and demand locations of blood perfusion in the cost matrix. The blood perfusion paths form a bipartite graph. The module performs matching and solving on the bipartite graph to obtain the optimal perfusion matching set. The evaluation index output module is used to output the overall endometrial perfusion resistance based on the path cost corresponding to the optimal perfusion matching set; and to output the degree of endometrial non-perfusion based on the number of unmatched demand locations. The method for obtaining the blood flow pulsation rhythm coefficient includes: The cardiac cycle frequency is obtained by analyzing the temporal variation of the average brightness of the endometrial ultrasound echo image; the brightness sequence at each location is filtered based on the cardiac cycle frequency to obtain a filtered sequence; the blood flow pulsation rhythm coefficient is obtained by comparing the filtered sequence with the brightness sequence. The method for obtaining local microvascular perfusion resistance includes: The overall blood perfusion volume is obtained based on the brightness characteristics of each location in consecutive frames of endometrial ultrasound echo images. The local microvascular perfusion resistance is obtained by multiplying the blood flow pulsation rhythm coefficient by the overall blood perfusion volume and performing negative correlation mapping. The method for obtaining the cost of single-step infusion resistance includes: Obtain a direction vector pointing from one location to an adjacent location, and the cosine value of the deviation angle between the vector and the anatomical perpendicular growth direction; obtain the single-step perfusion resistance cost based on the local microvascular perfusion resistance at the pointed location, the cosine value of the deviation angle, and the Euclidean distance between the two locations.

2. The endometrial receptivity assessment system based on vaginal high-frequency ultrasound imaging according to claim 1, characterized in that, The method for obtaining the cardiac cycle frequency includes: The average brightness of each frame of endometrial ultrasound echo image is arranged in chronological order to form an average brightness sequence; the average brightness sequence is subjected to a fast Fourier transform, and the frequency with the largest amplitude in the preset conventional heart rate frequency band is searched as the cardiac cycle frequency.

3. The endometrial receptivity assessment system based on vaginal high-frequency ultrasound imaging according to claim 1, characterized in that, The step of comparing the filtered sequence with the luminance sequence to obtain the blood flow pulsatility coefficient includes: The ratio of the total energy of the filtered sequence as the numerator and the total energy of the brightness sequence as the denominator is the blood flow pulsation rhythm coefficient.

4. The endometrial receptivity assessment system based on vaginal high-frequency ultrasound imaging according to claim 1, characterized in that, The method for obtaining the total blood perfusion volume includes: The average value of the elements in the brightness sequence is taken as the overall blood perfusion volume.

5. The endometrial receptivity assessment system based on vaginal high-frequency ultrasound imaging according to claim 1, characterized in that, The method for selecting the blood perfusion pathway includes: For any combination of supply and demand locations, the optimal path of the combination in the cost matrix is ​​obtained, and the cumulative value of the single-step injection resistance cost corresponding to the optimal path is normalized to obtain the path cost of the optimal path. Among all combinations, the optimal path corresponding to the combination whose path cost is less than a preset cost threshold is taken as the blood perfusion path.

6. The endometrial receptivity assessment system based on vaginal high-frequency ultrasound imaging according to claim 1, characterized in that, The method for obtaining the overall endometrial perfusion resistance includes: The average path cost of the optimal perfusion matching set is used as the overall perfusion resistance of the endometrium.

7. The endometrial receptivity assessment system based on vaginal high-frequency ultrasound imaging according to claim 1, characterized in that, The method for obtaining the degree of endometrial nonperfusion includes: If the number of required locations is 1, then the degree of incomplete perfusion of the endometrium will not be output; If the number of required locations is greater than 1, the proportion of unmatched required locations in the total number of required locations is taken as the degree of incomplete perfusion of the endometrium.

Images