A multi-omics data-driven ovarian reserve dysfunction risk assessment system

Abstract

The present application relates to the technical field of medical information processing, and in particular to a multi-omics data driven ovarian reserve function decline risk assessment system. The system determines the local resistance value of each pixel point in the region of interest, and constructs a matrix resistance map; according to the positioning edge reference anchor point of the region of interest; taking the anchor point as the starting point, the cumulative resistance value of each pixel point is determined based on the resistance map; based on the cumulative resistance value, the blood supply source point of each pixel point is screened; according to the historical hormone detection data corresponding to the anchor point, in the order of cumulative resistance value from small to large, the local mapping signal value of each pixel point is determined in turn; and according to the signal value, cumulative resistance value and matrix interference penalty value of the blood supply source point of each pixel point, the penalty value of each pixel point is determined; according to the signal value and the penalty value, the follicle response index of each pixel point is determined, and a follicle response map is constructed; according to the follicle response map, the function risk grade of the target object is evaluated, and the function impaired area is accurately identified.

Description

Technical Field

[0001] This invention relates to the field of medical information processing technology, specifically to a multi-omics data-driven risk assessment system for ovarian reserve decline. Background Technology

[0002] In auxiliary clinical practice, diminished ovarian reserve seriously affects women's health. Current clinical assessment systems mainly rely on static indicators such as serum anti-Müllerian hormone, basal follicle-stimulating hormone, and antral follicle count under ultrasound. These indicators only reflect the number of follicles in reserve, and this method assumes that the ovarian stroma is a homogeneous medium, thus ignoring the dynamic attenuation of hormone signals during transmission within the stroma environment.

[0003] In clinical practice, age or pathological factors can easily lead to ovarian stromal fibrosis, which hinders the penetration of hormones into deeper tissues. At the same time, early decline of the hypothalamus-pituitary-ovarian axis can generate background noise. The combination of these two factors leads to a decrease in the signal-to-noise ratio of hormone signals, resulting in a hidden functional abnormality of "normal hormone levels but low follicular response". This is also the main reason why some individuals have normal static hormone levels but low response during ovulation induction treatment. Summary of the Invention

[0004] To address the technical problem that existing technologies rely solely on static indicators to assess ovarian reserve function, failing to quantify the effectiveness of hormonal signal transmission in the fibrotic stroma and thus resulting in insufficient risk assessment of diminished ovarian reserve, this invention provides a multi-omics data-driven system for assessing the risk of diminished ovarian reserve. The specific technical solution adopted is as follows: This invention proposes a multi-omics data-driven risk assessment system for ovarian reserve decline, the system comprising: The acquisition module is used to acquire historical hormone testing data and ovarian ultrasound image data of the target object; perform image segmentation on the ovarian tissue in the ultrasound image data, and extract the region of interest of the ovarian stroma. The matrix resistance module is used to determine the local resistance value of each pixel in the region of interest (ROI) based on the texture features of the RIO and to construct a matrix resistance map; it also locates edge reference anchor points based on the anatomical structure of the RIO. The hormone and noise module is used to determine the cumulative resistance value of each pixel based on the resistance map, starting from the edge reference anchor point; to screen the blood supply source point of each pixel based on the cumulative resistance value; to determine the local mapping signal value of each pixel in order of increasing cumulative resistance value according to the historical hormone detection data corresponding to the edge reference anchor point; and to determine the matrix interference penalty value of each pixel based on the local mapping signal value, cumulative resistance value and matrix interference penalty value of its blood supply source point. The synthesis module is used to determine the follicle response index of each pixel based on the local mapping signal value and the matrix interference penalty value, and to construct the follicle response map; The assessment module is used to evaluate the functional risk level of the target subject based on the follicle response map; identify the dominant factor type leading to functional abnormalities based on the matrix interference penalty value; and generate individualized strategies based on the risk level and dominant factor type.

[0005] Furthermore, the matrix resistance map construction process includes: For each pixel within the region of interest of the ovarian stroma, a sliding window is constructed with the pixel as the center; the gray-level co-occurrence matrix corresponding to the gray values ​​of the pixels within the sliding window is statistically analyzed and calculated. The contrast ratio of a pixel is calculated based on the contrast characteristics of the gray-level co-occurrence matrix; the homogeneity index of a pixel is calculated based on the homogeneity characteristics of the gray-level co-occurrence matrix. The sum of the homogeneity index and the preset minimum constant is calculated as the first sum value; the ratio of the contrast to the first sum value is calculated as the local resistance value of the pixel in the region of interest of the ovarian stroma. A two-dimensional matrix corresponding to the region of interest in the ovarian stroma is constructed from the local resistance values ​​of all pixels, serving as the stroma resistance map.

[0006] Furthermore, the step of locating edge reference anchor points based on the anatomical structure of the region of interest includes: Determine the geometric center and edge points of the region of interest in the ovarian stroma; and acquire Doppler blood flow images that match the ovarian ultrasound imaging data; Calculate the Euclidean distance from each edge point to the geometric center, and use it as the geometric characteristic value of each edge point; Based on the Doppler blood flow image, points containing blood flow signals among the edge points are selected as candidate edge reference anchor points; Select the edge point with the largest geometric feature value from the candidate edge reference anchor points as the edge reference anchor point; if there are no candidate edge reference anchor points, select the edge point with the largest geometric feature value from all edge points as the edge reference anchor point.

[0007] Furthermore, the process of determining the cumulative resistance value includes: Starting from the edge reference anchor point, set its cumulative resistance value to zero; Starting from the edge reference anchor point, proceed outward point by point. For the pixel to be calculated, extract its neighboring pixels within a preset neighborhood; from the neighboring pixels, select the target neighboring pixels with a determined cumulative resistance value. For each target neighboring pixel, calculate the Euclidean distance between the target neighboring pixel and the pixel to be calculated; calculate the product of the distance and the local resistance value of the pixel to be calculated as the process resistance value; calculate the sum of the cumulative resistance value of the target neighboring pixels and the process resistance value as the neighbor cumulative resistance value. Take the minimum value among the cumulative resistance values ​​of all neighboring pixels of the target as the cumulative resistance value of the pixel to be calculated.

[0008] Furthermore, the step of filtering the blood supply source point for each pixel based on the cumulative resistance value includes: For each pixel in the region of interest of the ovarian stroma, the cumulative resistance value of each target neighboring pixel is compared, and the target neighboring pixel with the smallest cumulative resistance value is determined as the blood supply source point of the pixel. If multiple adjacent target pixels have the same minimum cumulative resistance value, one of them is selected as the blood supply source point of the pixel according to the preset priority rule; A blood supply path network with clear upstream and downstream relationships is formed by the blood supply source points of all pixels.

[0009] Furthermore, the historical hormone detection data consists of at least two basal follicle-stimulating hormone (FSH) measurements collected during the target subject's preset physiological period; the step of determining the local mapping signal value of each pixel based on the historical hormone detection data corresponding to the edge reference anchor point, in ascending order of cumulative resistance value, includes: Calculate the arithmetic mean of all baseline follicle-stimulating hormone (FSH) measurements in historical hormone testing data as the first indicator; assign the first indicator to the local mapping signal value of the edge reference anchor point. All pixels within the region of interest of the ovarian stroma are sorted in ascending order of cumulative resistance value. Starting from the edge reference anchor point with the smallest cumulative resistance value, all other pixels except the edge reference anchor point are processed sequentially. For each other pixel being processed, calculate the Euclidean distance between the other pixel and its blood supply source point; Based on the local resistance values ​​of other pixels and the Euclidean distance, the feature signal attenuation rate during the process of moving from the blood supply source point to other pixels is calculated; wherein, the feature signal attenuation rate is negatively correlated with the product of the local resistance value and the Euclidean distance; Multiply the local mapping signal value of the blood supply source point of other pixels by the feature signal attenuation rate to obtain the local mapping signal value of other pixels; until all other pixels have obtained the corresponding local mapping signal value.

[0010] Furthermore, the process for determining the matrix interference penalty value for each pixel includes: Calculate the variance of all baseline follicle-stimulating hormone (FSH) measurements in historical hormone testing data as a second indicator; assign the second indicator the matrix interference penalty value of the marginal reference anchor point. Starting from the edge reference anchor point, other pixels are processed sequentially in order of increasing cumulative resistance value. For each pixel being processed, the square of the feature signal attenuation rate during the process from its blood supply source point to other pixels is calculated as the signal attenuation index. The matrix interference penalty value of the blood supply source point of other pixels is multiplied by the signal attenuation index to obtain the first component. The product of the local mapping signal value of other pixels, the local resistance value of other pixels, and the preset matrix interference coefficient is calculated as the second component; The sum of the first component and the second component is calculated as the matrix interference penalty value for other pixels; this continues until all other pixels have obtained the corresponding matrix interference penalty value.

[0011] Furthermore, the follicle response map construction process includes: For each pixel within the region of interest of the ovarian stroma, the sum of the stroma interference penalty value of the pixel and the preset minimum constant is calculated as the second sum value; the square root operation of the second sum value is performed to obtain the background interference level value of the pixel. The ratio of the local mapping signal value of a pixel to the background interference value is calculated and used as a follicle response index for pixels within the region of interest of the ovarian stroma. A two-dimensional matrix corresponding to the region of interest in the ovarian stroma is constructed from the follicle response indices of all pixels, serving as the follicle response map.

[0012] Furthermore, the assessment of the functional risk level of the target subject based on the follicle response map includes: The recruitment threshold is determined based on the statistical distribution characteristics of the follicle response index of all pixels in the follicle response map. Pixels in the follicle response map whose follicle response index is not less than the recruitment threshold are marked as valid response points. The first number of valid response points is counted, and the second number of all pixels in the ovarian stroma region of interest is counted. Calculate the ratio of the first quantity to the second quantity, and use this as the percentage of the effective area; The effective area percentage is compared with a preset first threshold and a second threshold; wherein the first threshold is greater than the second threshold. If the proportion of effective areas is not less than the preset first threshold, a low-risk level is generated; if the proportion of effective areas is between the second threshold and the first threshold, a medium-risk level is generated; if the proportion of effective areas is not greater than the second threshold, a high-risk level is generated.

[0013] Furthermore, the identification of the dominant factor type leading to functional abnormalities based on matrix interference penalty values ​​includes: Low response regions below the recruitment threshold are identified based on the follicle response map. Low response regions are areas composed of pixels whose follicle response index is less than the recruitment threshold. For each pixel in the low response region, obtain the first and second components of the pixel matrix interference penalty value; Calculate the arithmetic mean of the first component of all pixels in the low response region as the first energy representative value; calculate the arithmetic mean of the second component of all pixels in the low response region as the second energy representative value. If the first energy representative value is greater than the second energy representative value, the dominant factor type is determined to be centrally dominant; if the first energy representative value is less than the second energy representative value, the dominant factor type is determined to be locally dominant.

[0014] The present invention has the following beneficial effects: This invention segments ovarian tissue from ultrasound images, extracts the region of interest (ROI) of the ovarian stroma, and determines the local resistance value of each pixel based on the texture features of the ROI, constructing a stroma resistance map reflecting the fibrosis characteristics of the ovarian stroma. Simultaneously, the system locates edge reference anchors based on the anatomical structure of the ovary, providing a physiologically accurate starting point for subsequent analysis of hormone signal propagation. This allows for the transformation of tissue texture information, which is difficult to quantify in conventional ultrasound images, into quantifiable indicators, providing an objective basis for assessing hormone signal transmission conditions in the stroma. Starting from the edge reference anchors, the cumulative resistance value of each pixel is determined based on the stroma resistance map, and the blood supply source points of each pixel are selected according to the cumulative resistance values ​​to construct a hormone signal propagation path conforming to the principle of least resistance path. Then, combined with historical hormone detection data corresponding to the edge reference anchors, the local mapping signal value and stroma interference penalty value of each pixel are determined sequentially, where the local mapping signal value reflects the fibrosis characteristics of the ovarian stroma. The effective intensity of the hormone signal when it reaches a pixel is calculated, and the matrix interference penalty value reflects the degree of background interference at the pixel. This calculation process enables the simulation of the attenuation process of the hormone signal in non-uniform tissue (i.e., ovarian matrix), thereby transforming the static matrix resistance map into dynamic signal propagation parameters, providing further evidence for analyzing the propagation of hormone signals in the ovarian matrix. By integrating the local mapping signal value of each pixel with the matrix interference penalty value, the follicle response index of the pixel is determined, and a follicle response map covering the entire region of interest of the ovarian matrix is ​​constructed. The follicle response index comprehensively reflects the relative relationship between signal intensity and background interference, and its value directly characterizes the clarity with which the follicle recognizes the hormone instruction from the background interference at the pixel. The entire process transforms the signal propagation parameters into biologically meaningful evaluation indicators, thereby identifying functionally impaired areas where hormone signals cannot be effectively activated by conventional methods due to excessive attenuation or excessive background interference. Attached Figure Description

[0015] 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.

[0016] Figure 1 This is a schematic diagram of the structure of a multi-omics data-driven risk assessment system for ovarian reserve decline provided in one embodiment of the present invention; Figure 2 This is an example diagram illustrating the matrix resistance map construction process provided in one embodiment of the present invention. Detailed Implementation

[0017] 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 a multi-omics data-driven ovarian reserve decline risk assessment system 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.

[0018] 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.

[0019] The following description, in conjunction with the accompanying drawings, details the specific scheme of the multi-omics data-driven ovarian reserve decline risk assessment system provided by this invention.

[0020] Please see Figure 1 The diagram illustrates a multi-omics data-driven risk assessment system for ovarian reserve decline according to an embodiment of the present invention. The system includes: The acquisition module 101 is used to acquire historical hormone testing data and ovarian ultrasound image data of the target object; to perform image segmentation on the ovarian tissue in the ultrasound image data and to extract the region of interest of the ovarian stroma.

[0021] The target group refers to female individuals who need to undergo ovarian reserve function assessment.

[0022] It should be noted that the historical hormone testing data and ovarian ultrasound imaging data used in this system are all derived from routine examination records generated during the target subjects' visits to medical institutions. The acquisition and use of data strictly comply with relevant laws, regulations, and medical ethics. With the informed consent of the target subjects or their authorized representatives, and under the premise of ensuring personal privacy and data security, the data is anonymized and retrieved from the hospital information system or imaging archive system, solely for the analysis and evaluation purposes of this system.

[0023] Since follicle-stimulating hormone (FSH) is a gonadotropin secreted by the anterior pituitary gland, its main physiological function is to promote the growth and development of follicles in the ovary. At the beginning of menstruation, the ovary is in a relatively quiescent state, and no dominant follicle has yet secreted a large amount of estrogen. Therefore, the FSH level in the target subject during the predetermined physiological period can relatively stably reflect the basic functional state of the hypothalamus-pituitary-ovarian axis. Thus, the FSH concentration value obtained from serum samples collected during this period (i.e., the basal FSH measurement) is a routine indicator used clinically to assess ovarian reserve function and central nervous system driving capacity.

[0024] Historical hormone testing data consist of at least two baseline follicle-stimulating hormone (FSH) measurements taken during the target subject's pre-defined physiological period.

[0025] Since the early menstrual cycle is the initial stage of follicular development, the serum basal follicle-stimulating hormone (FSH) level at this time can relatively stably reflect the basic functional state of the hypothalamus-pituitary-ovarian axis, unaffected by the negative feedback after dominant follicle selection. Therefore, the target physiological period is usually day 2 to 4 of the target individual's menstrual cycle.

[0026] Ovarian ultrasound imaging data refers to two-dimensional grayscale images obtained through transvaginal ultrasound examination.

[0027] It should be noted that transvaginal ultrasound is a routine method of gynecological ultrasound. The two-dimensional grayscale images contain rich tissue texture information, which can be used to reflect the density of the ovarian stroma, providing an imaging basis for subsequent assessment of ovarian stroma fibrosis. Ultrasound image data can be obtained from the hospital's image archives, which is a standard practice in this field and will not be elaborated upon in this embodiment.

[0028] Ovarian matrix specifically refers to the remaining part of ovarian tissue after removing functional structures such as follicles and corpus luteum. It is mainly composed of extracellular matrix, stromal cells, microvessels and nerve fibers, and is the tissue medium that hormone signals need to pass through after entering the blood vessels.

[0029] As a preferred implementation, the outer contour of the ovary in a two-dimensional grayscale ultrasound image is identified and extracted based on grayscale thresholding or a deep learning-based image segmentation network; anechoic areas within the outer contour of the ovary are identified and designated as candidate follicle regions; vascular regions with blood flow signals are excluded from the candidate follicle regions by combining Doppler blood flow images; the remaining anechoic areas are identified as follicle regions and removed from the outer contour of the ovary; the remaining region after removing follicles is identified as the region of interest of the ovarian stroma.

[0030] It should be noted that both the gray-scale threshold segmentation method and the deep learning-based image segmentation network are common techniques in the field of image segmentation technology. These methods can automatically identify the ovarian boundary and extract the outer contour of the ovary based on the gray-scale difference between the ovarian tissue and the surrounding tissue. This embodiment can be directly applied and will not be elaborated further here.

[0031] After segmenting the outer contour of the ovary, the area within the contour includes not only the ovarian stroma but also follicles of varying sizes and potential blood vessels. The formation of anechoic areas is due to the presence of follicular fluid within the follicles, which reflects weak signals when ultrasound waves pass through, appearing as dark areas on the image. Therefore, the system can use grayscale thresholding (e.g., if the ultrasound image uses 8-bit grayscale, the grayscale threshold is set to 50) to identify connected regions within the outer contour of the ovary where the grayscale value is significantly lower than the surrounding stroma, marking these regions as candidate follicular areas. This is a standard method for identifying cystic structures in ultrasound image processing and will not be elaborated upon here.

[0032] It is understandable that areas with blood flow signals are identified as blood vessels and excluded from the candidate follicle areas; areas without blood flow signals are identified as follicle areas.

[0033] Doppler blood flow images are Doppler blood flow images registered with two-dimensional grayscale images. These images are acquired using Doppler ultrasound technology and can display blood flow signals. Registration between the Doppler blood flow image and the grayscale image is a standard function of ultrasound equipment; the images are naturally aligned during acquisition and require no additional processing. This embodiment can be applied directly.

[0034] The matrix resistance module 102 is used to determine the local resistance value of each pixel in the region of interest based on the texture features of the region of interest, and to construct a matrix resistance map; and to locate edge reference anchor points based on the anatomical structure of the region of interest.

[0035] The process of constructing the matrix resistance map is as follows: Figure 2 As shown, it includes: S101-1, For each pixel in the region of interest of the ovarian stroma, a sliding window is constructed with the pixel as the center; the gray-level co-occurrence matrix corresponding to the gray values ​​of the pixels in the sliding window is statistically analyzed and calculated.

[0036] It should be noted that the size of the sliding window can be set based on clinical experience, and this embodiment does not impose a specific limitation. For example, a sliding window of 5×5 pixels or 7×7 pixels can be selected.

[0037] If a pixel is located at the edge of the region of interest (ROI), the sliding window may extend beyond the ROI. To address this, the system can employ conventional image boundary processing methods, such as using only pixels within the ROI that fall within the window for calculations, or filling the boundary region through mirroring or copying, to ensure that each pixel has corresponding sliding window data.

[0038] It should be noted that, for each sliding window, the system statistically analyzes the spatial dependencies of pixel grayscale values ​​within the window to construct a grayscale co-occurrence matrix. The construction of this matrix is ​​a publicly known technique in the field and can be directly applied in this embodiment, so it will not be elaborated upon further. Specifically, for each sliding window, the system statistically analyzes the frequency of grayscale value combinations of pixel pairs that satisfy preset spacing (e.g., typically 1, i.e., adjacent pixels) and orientation (e.g., horizontal 0°, vertical 90°, diagonal 45° and 135°), constructing a two-dimensional matrix. The rows and columns of the matrix correspond to the grayscale levels of the pixels, and the matrix element values ​​represent the frequency of occurrence of the corresponding grayscale level combination.

[0039] S101-2, the contrast of a pixel is calculated based on the contrast characteristics of the gray-level co-occurrence matrix; the homogeneity index of a pixel is calculated based on the homogeneity characteristics of the gray-level co-occurrence matrix.

[0040] The contrast feature of the gray-level co-occurrence matrix (GLCM) reflects the degree of difference in gray-level values ​​of pixels within a sliding window, i.e., the depth of the texture. The contrast feature is obtained by statistically analyzing the distribution of elements far from the diagonal in the GLCM, a standard procedure in GLCM texture analysis, and will not be elaborated upon in this embodiment. Specifically, a higher contrast for a given pixel indicates a more significant difference in gray-level values ​​within the sliding window, resulting in a deeper texture.

[0041] The homogeneity characteristic of the gray-level co-occurrence matrix (GLCM) reflects the uniformity of pixel gray-level values ​​within the sliding window, i.e., the smoothness of the texture. The homogeneity characteristic is obtained by statistically analyzing the distribution of elements near the diagonal of the GLCM, a standard procedure in GLCM texture analysis, and will not be elaborated upon in this embodiment. A higher homogeneity index for a pixel indicates a smoother gray-level change and a more uniform texture within the sliding window containing that pixel.

[0042] S101-3, calculate the sum of the homogeneity index and the preset minimum constant as the first sum value; calculate the ratio of the contrast to the first sum value as the local resistance value of the pixel in the region of interest of the ovarian stroma.

[0043] It should be noted that the specific value of the preset minimum constant is set based on engineering experience, and this embodiment does not impose a specific limitation. For example, a typical value for the preset minimum constant is... .

[0044] Local resistance values ​​are used to quantify the degree of obstruction to hormone signal penetration at various locations (i.e., at each pixel) within the ovarian stroma. A higher local resistance value at a pixel indicates a denser and more disordered ovarian stroma at that location, making it more difficult for hormones to pass through. Conversely, a lower local resistance value indicates a looser and more homogeneous ovarian stroma at that location, making it easier for hormones to penetrate.

[0045] It is important to understand that since contrast reflects the degree of difference in the gray values ​​of pixels within a window, a higher contrast indicates a deeper tissue texture and a more complex structure; homogeneity reflects the uniformity of the gray values ​​of pixels within a window, a lower homogeneity indicates a more uneven tissue and a more irregular structure, in which case hormones are less likely to penetrate.

[0046] S101-4 is a two-dimensional matrix consisting of the local resistance values ​​of all pixels, corresponding to the region of interest of the ovarian stroma, and serves as a stroma resistance map.

[0047] It's important to understand that in real physiological environments, hormone molecules diffusely enter ovarian tissue through the capillary network of the ovarian hilum, rather than concentrating at a single geometric point. However, in numerical simulations, a clear source point boundary condition must be defined to subsequently solve the initial value problem of the functional equation (i.e., solving for the cumulative resistance value). Therefore, this system makes a reasonable simplification based on the following anatomical observations: the ovarian hilum is usually the pedicle connecting the ovary to the broad ligament, a region where blood vessels concentrate their entry and exit. On ultrasound images, it appears as the edge point furthest from the geometric center of the ovary and possessing abundant blood flow signals. Therefore, defining this edge reference anchor point represents the main source of hormone signals entering the ovarian tissue. This simplification, while ensuring the validity of the simulation results, allows complex physiological processes to be quantified numerically.

[0048] To accurately locate edge reference anchors, as an example, the geometric center and edge points of the region of interest in the ovarian stroma are determined; and Doppler blood flow images matching the ovarian ultrasound imaging data are acquired; the Euclidean distance from each edge point to the geometric center is calculated as the geometric feature value of each edge point; based on the Doppler blood flow images, points containing blood flow signals among the edge points are selected as candidate edge reference anchors; the edge point with the largest geometric feature value is selected from the candidate edge reference anchors as the edge reference anchor; if there are no candidate edge reference anchors, the edge point with the largest geometric feature value is selected from all edge points as the edge reference anchor.

[0049] The geometric center can be obtained by calculating the arithmetic mean of the coordinates of all pixels within the region of interest, that is, the point corresponding to the arithmetic mean of the horizontal coordinates and the arithmetic mean of the vertical coordinates.

[0050] Edge points refer to pixels located on the boundary of the region of interest. The specific method for determining these edges can be obtained using edge detection algorithms in image processing, which are publicly available techniques and will not be elaborated upon in this embodiment.

[0051] The Euclidean distance calculation formula is a conventional coordinate distance calculation method in this field, and will not be described in detail in this embodiment.

[0052] Geometric eigenvalues ​​reflect the positional relationship of edge points relative to the geometric center. Specifically, the larger the geometric eigenvalue of a given edge point, the farther it is from the geometric center, and the more likely it is to be located at the pedicle of the ovarian hilum.

[0053] It should be noted that identifying blood flow signals through Doppler blood flow images is a publicly disclosed technique in this field, and this embodiment can be directly applied. Specifically, for each edge point, it is checked whether its corresponding position in the Doppler blood flow image displays a color-coded area with blood flow signals. Edge points with blood flow signals are determined to potentially contain vascular structures and are used as candidate edge reference anchor points, while edge points without blood flow signals are excluded.

[0054] In this embodiment, there may be a situation where multiple edge points have the same maximum geometric feature value. In this case: the blood flow signal intensity of multiple edge points is obtained based on the Doppler blood flow image; the edge point with the largest blood flow signal intensity is selected as the edge reference anchor point; if the blood flow signal intensity of multiple edge points is the same, one of the edge points is selected as the edge reference anchor point according to the preset coordinate priority rule.

[0055] It should be noted that the preset coordinate priority rule can be determined based on industry experience, and this embodiment does not impose specific limitations. For example, the preset coordinate priority rule can be to prioritize edge points with the smallest horizontal or vertical coordinates.

[0056] The hormone and noise module 103 is used to determine the cumulative resistance value of each pixel based on the resistance map, starting from the edge reference anchor point; to screen the blood supply source point of each pixel based on the cumulative resistance value; to determine the local mapping signal value of each pixel in order of increasing cumulative resistance value according to the historical hormone detection data corresponding to the edge reference anchor point; and to determine the matrix interference penalty value of each pixel based on the local mapping signal value, cumulative resistance value and matrix interference penalty value of its blood supply source point.

[0057] It's important to understand that because hormone signals propagate through the non-uniform ovarian stroma, they preferentially choose paths with less resistance rather than straight lines. Therefore, by calculating the total resistance that must be overcome along the path of least resistance from the edge reference anchor point to each pixel—the cumulative resistance value—we can determine the order in which the hormone signal reaches a particular pixel. This is because positions with lower cumulative resistance values ​​are easier for the signal to reach and are therefore prioritized in subsequent calculations. Furthermore, by comparing the cumulative resistance values ​​of adjacent pixels, we can determine the direction of signal propagation; hormones always flow from points with lower cumulative resistance values ​​to points with higher cumulative resistance values. This provides an objective basis for subsequently determining the blood supply source and sequentially calculating the local mapped signal value and stroma interference penalty value.

[0058] The process of determining the cumulative resistance value includes the following steps: 1) Starting from the edge reference anchor point, set its cumulative resistance value to zero.

[0059] Since the marginal reference anchor point is the source of hormone signals entering the ovarian tissue, it has not yet experienced resistance consumption. Therefore, its cumulative resistance value is set to zero.

[0060] 2) Starting from the edge reference anchor point, proceed outward point by point. For the pixel to be calculated, extract its neighboring pixels within the preset neighborhood; from the neighboring pixels, select the target neighboring pixels with the determined cumulative resistance value.

[0061] It should be noted that the specific value of the preset neighborhood is determined based on engineering experience, and this embodiment does not impose a specific limitation. For example, an 8-neighborhood is usually used as the preset neighborhood, which includes the four orthogonal directions of the pixel (up, down, left, and right) and the four diagonal directions (upper left, upper right, lower left, and lower right).

[0062] 3) For each target neighboring pixel, calculate the Euclidean distance between the target neighboring pixel and the pixel to be calculated; calculate the product of the distance and the local resistance value of the pixel to be calculated as the process resistance value; calculate the sum of the cumulative resistance value of the target neighboring pixels and the process resistance value as the neighbor cumulative resistance value.

[0063] Since the local resistance value reflects the degree to which the local matrix represented by a pixel hinders hormone penetration, and the Euclidean distance reflects the path length from the target's neighboring pixels to the pixel to be calculated, the product of the two is the total resistance consumption of that path segment. Therefore, the process resistance value characterizes the local resistance that needs to be overcome to travel the short distance from the neighboring pixels to the pixel to be calculated.

[0064] In particular, if the process resistance value from a target's adjacent pixel to a pixel to be calculated is greater, it means that the path from the target's adjacent pixel to the pixel is more difficult to traverse, which is more likely due to the denser matrix at the pixel (i.e., the greater the local resistance value) or the greater the distance between the target's adjacent pixel and the pixel.

[0065] The cumulative resistance value of the neighbor represents the total resistance that needs to be overcome to complete the path from the edge reference anchor point, through a target neighbor pixel, and then to the pixel to be calculated.

[0066] It's important to understand that for the pixel to be calculated, different target neighboring pixels correspond to different cumulative resistance values. If the cumulative resistance value of the path from the edge reference anchor point, through a target neighboring pixel, to the pixel to be calculated is higher, the total resistance from the edge reference anchor point to the target neighboring pixel is higher, and this path is more difficult for hormone signals. Conversely, if the cumulative resistance value of the path from the edge reference anchor point, through a target neighboring pixel, to the pixel to be calculated is lower, the total resistance of the entire path is lower, and hormone signals are more likely to choose this path.

[0067] 4) Take the minimum value among the cumulative resistance values ​​of all neighboring pixels of the target as the cumulative resistance value of the pixel to be calculated.

[0068] Because hormone signals actually choose the path of least total resistance during propagation, rather than any arbitrary path, the system takes the minimum cumulative resistance value among all the target's neighboring pixels as the cumulative resistance value.

[0069] In this embodiment, for each pixel within the region of interest of the ovarian stroma, the cumulative resistance values ​​of each target neighboring pixel are compared, and the target neighboring pixel with the smallest cumulative resistance value is determined as the blood supply source point of the pixel; if multiple target neighboring pixels have the same minimum cumulative resistance value, one of them is selected as the blood supply source point of the pixel according to a preset priority rule; the blood supply source points of all pixels constitute a blood supply path network with a clear upstream and downstream relationship.

[0070] The blood supply source point is the previous pixel in the path of least resistance from the edge reference anchor point to a certain pixel.

[0071] It should be noted that the specific content of the preset priority rules can be set based on experience, and this embodiment does not impose specific limitations. For example, the priority rules can be set as follows: prioritize selecting pixels that are adjacent in the horizontal or vertical direction, and then select pixels that are adjacent in the diagonal direction.

[0072] It's important to understand that after all pixels have their respective blood supply sources determined, the system organizes the connections between these points to form a blood supply path network covering the entire ovarian stroma region of interest. Therefore, the blood supply path network has a clear upstream and downstream relationship: the edge reference anchor point is located at the very upstream of the network, with the lowest cumulative resistance value and no blood supply source point; all other pixels have a unique blood supply source point pointing upstream; the direction of hormone signal propagation is consistent with the direction of the blood supply path network, that is, it diffuses downstream step by step from the edge reference anchor point.

[0073] It is important to understand that, as hormone signals enter the ovarian tissue from the edge reference anchor point, they are attenuated during the process of passing through the fibrotic matrix due to tissue absorption and path loss, resulting in significant differences in hormone concentration at different locations. Therefore, it is necessary to determine the local mapping signal value of each pixel to quantify the effective intensity of the hormone signal when it actually reaches each location in the ovary, providing basic data for subsequent assessment of whether the follicle can receive instructions of sufficient strength.

[0074] As one possible implementation, the arithmetic mean of all baseline follicle-stimulating hormone (FSH) measurements in historical hormone testing data is calculated as a first indicator. This first indicator is then assigned the local mapping signal value of the edge reference anchor point. All pixels within the region of interest of the ovarian stroma are sorted in ascending order of cumulative resistance value. Starting from the edge reference anchor point with the smallest cumulative resistance value, all other pixels except the edge reference anchor point are processed sequentially. For each pixel being processed, the Euclidean distance between the pixel and its blood supply source is calculated. Based on the local resistance value and Euclidean distance of the pixel, the feature signal attenuation rate during the journey from the blood supply source point to the pixel is calculated. The feature signal attenuation rate is negatively correlated with the product of the local resistance value and the Euclidean distance. The local mapping signal value of the blood supply source point of each pixel is multiplied by the feature signal attenuation rate to obtain the local mapping signal value of the pixel. This process continues until all other pixels have obtained their corresponding local mapping signal values.

[0075] Since basal follicle-stimulating hormone (FSH) is a hormone signal secreted by the pituitary gland and transported to the ovaries via the bloodstream, its serum concentration directly reflects the strength of the central nervous system's commands. Therefore, using the arithmetic mean of historical test data as the local mapping signal value of the marginal reference anchor point can represent the initial effective concentration of the hormone signal when it first enters the ovarian tissue.

[0076] The local mapping signal value characterizes the remaining effective concentration of hormone signals after propagation from the edge reference anchor point to a certain pixel. A larger local mapping signal value indicates a stronger hormone signal reaching that pixel, increasing the likelihood that the follicle receives sufficient signals. Conversely, a smaller local mapping signal value indicates more severe attenuation of the hormone signal during propagation, making it more difficult for the follicle to be effectively activated due to insufficient hormones.

[0077] Since the cumulative resistance value represents the total resistance that needs to be overcome to reach each pixel from the edge reference anchor point, a smaller cumulative resistance value indicates that the hormone signal is easier to reach. Therefore, in the calculation order, points with small cumulative resistance values ​​should be processed first, followed by points with large cumulative resistance values. Among them, the edge reference anchor point itself has the smallest cumulative resistance value, so it is processed first as the starting point, and then other pixels are processed in ascending order.

[0078] The characteristic signal attenuation rate characterizes the proportion of the remaining hormone signal intensity after the hormone signal travels from the blood supply point to a certain pixel, relative to the initial hormone signal intensity. A higher characteristic signal attenuation rate indicates less hormone signal loss during its journey, higher transmission efficiency, and the pixel receiving a signal intensity closer to the upstream point. Conversely, a lower characteristic signal attenuation rate indicates more hormone signal loss during its journey, lower transmission efficiency, and a greater signal attenuation compared to the upstream point.

[0079] The attenuation rate of the feature signal from the blood supply point to a specific pixel is determined by both the local resistance value of that pixel and the Euclidean distance between the two points. A larger local resistance value indicates a denser matrix at that pixel, while a larger Euclidean distance indicates a longer path. The product of these two values ​​reflects the total resistance overcome to travel this short distance from the blood supply point to the pixel. A smaller product of local resistance and Euclidean distance indicates an easier path to traverse, with the feature signal attenuation rate closer to 1. Conversely, a larger product indicates a more difficult path to traverse, resulting in greater hormone consumption and a smaller feature signal attenuation rate. Therefore, the feature signal attenuation rate can be expressed by the following formula: Feature signal attenuation rate = exp(-local resistance value × Euclidean distance × preset tissue attenuation coefficient).

[0080] Here, exp() represents an exponential function with the natural constant as its base.

[0081] It should be noted that the preset tissue attenuation coefficient is used to control the attenuation rate of hormone signals when passing through a unit resistance over a unit distance. Its specific value can be determined based on the physiological characteristics of ovarian tissue and clinical experience; this embodiment does not impose specific limitations. An example of the determination process is as follows: ovarian tissue samples with different degrees of fibrosis are collected, and the diffusion rate of hormone molecules through tissue of known thickness is measured under laboratory conditions. The corresponding attenuation coefficient is then calculated. For example, the preset tissue attenuation coefficient can be set to 0.1 per pixel per unit of resistance, meaning that if the local resistance value is 1, the signal intensity attenuates to 90% of its original value for every pixel traveled.

[0082] The process of determining the matrix interference penalty value for each pixel includes the following steps: 1) Calculate the variance of all baseline follicle-stimulating hormone (FSH) measurements in the historical hormone testing data as the second indicator; assign the second indicator the matrix interference penalty value of the marginal reference anchor point.

[0083] The second indicator reflects the degree of fluctuation in the central driving signal of the target subject (i.e., the hormonal signal generated by the neuroendocrine regulatory system of the hypothalamus-pituitary-ovarian axis). A larger variance indicates greater differences between basal follicle-stimulating hormone (FSH) measurements, and more unstable central driving signals; a smaller variance indicates closer similarities between basal FSH measurements, and more stable central driving signals.

[0084] Since basal follicle-stimulating hormone (FSH) is a hormone signal secreted by the pituitary gland, fluctuations in its measurement directly reflect the stability of the hypothalamic-pituitary-ovarian axis regulation. This instability itself is background noise carried by the hormone signal at its source. Therefore, the variance of all basal FSH measurements is used as the matrix interference penalty value of the marginal reference anchor point, representing the initial interference level when the hormone signal first enters the ovarian tissue.

[0085] 2) Process other pixels sequentially from the edge reference anchor point in order of increasing cumulative resistance value. For each other pixel being processed, calculate the square of the feature signal attenuation rate during the process of moving from its blood supply source point to other pixels, and use it as the signal attenuation index. Multiply the matrix interference penalty value of the blood supply source point of other pixels by the signal attenuation index to obtain the first component.

[0086] Since noise intensity is typically measured by variance, and variance is proportional to the square of the signal amplitude, if the signal undergoes linear attenuation, with the amplitude decreasing to k times its original value, the corresponding variance will decrease to a fraction of its original value. Therefore, based on the above principle, the local mapped signal value of the hormone signal in this scheme reflects the signal amplitude, while its matrix interference penalty value reflects the variance of the signal fluctuation. If the hormone signal propagates from the blood supply source point to a certain pixel, the signal amplitude attenuates according to the characteristic signal attenuation rate, and the corresponding noise intensity should attenuate according to the square of the characteristic signal attenuation rate. Therefore, the square of the characteristic signal attenuation rate is used as the signal attenuation index to calculate the residual amount of upstream noise after propagation to the pixel.

[0087] The first component is the value obtained by multiplying the matrix interference penalty value at the blood supply source point by the signal attenuation index. It represents the residual noise transmitted from upstream, that is, the part remaining after the noise at the edge reference anchor point has been attenuated during propagation. Specifically, a larger first component for a pixel indicates more residual upstream noise, and the background interference at that pixel is more primarily caused by the instability of the central driving signal; conversely, a smaller first component for a pixel indicates less residual upstream noise, and the instability of the central driving signal has been significantly attenuated during propagation.

[0088] 3) Calculate the product of the local mapping signal value of other pixels, the local resistance value of other pixels, and the preset matrix interference coefficient, and use it as the second component.

[0089] It should be noted that the preset matrix interference coefficient is used to control the conversion efficiency of noise generated when hormone signals interact with fibrotic tissue. A larger preset matrix interference coefficient indicates that the signal is more likely to generate additional scattering noise when it impacts the dense matrix; conversely, a smaller preset matrix interference coefficient indicates a weaker scattering effect of the matrix on the signal.

[0090] The specific value of the preset matrix interference coefficient can be determined based on the physiological characteristics of ovarian tissue and clinical experience; this embodiment does not impose a specific limitation. An example of the determination process is as follows: ovarian tissue samples with different degrees of fibrosis are collected, and the degree of waveform disorder after the signal passes through the tissue is measured under laboratory conditions. The corresponding matrix interference coefficient is then calculated. For example, the preset matrix interference coefficient can be set to 0.05.

[0091] The second component characterizes the matrix noise generated by the local environment of a pixel. Specifically, the larger the second component of a pixel, the stronger the signal of that pixel, the denser the matrix, or the larger the matrix interference coefficient, and the stronger the locally generated noise. Conversely, the smaller the second component of a pixel, the weaker the signal of that pixel, the looser the matrix, or the smaller the matrix interference coefficient, and the weaker the locally generated noise.

[0092] 4) Calculate the sum of the first component and the second component as the matrix interference penalty value for other pixels; until all other pixels have obtained the corresponding matrix interference penalty value.

[0093] The integration module 104 is used to determine the follicle response index of each pixel based on the local mapping signal value and the matrix interference penalty value, and to construct the follicle response map.

[0094] The process of constructing a follicle response map includes the following steps: 1) For each pixel in the region of interest of the ovarian stroma, calculate the sum of the stroma interference penalty value of the pixel and the preset minimum constant, and use it as the second sum value; perform a square root operation on the second sum value to obtain the background interference level value of the pixel.

[0095] Since the matrix interference penalty value is represented by variance, reflecting the intensity of background interference at a certain pixel, it is necessary to convert the variance into the standard deviation with the same dimension as the local mapped signal value for easier subsequent calculations. Therefore, the background interference level value is obtained by taking the square root of the sum of the matrix interference penalty value and a preset minimum constant.

[0096] 2) Calculate the ratio of the local mapping signal value of a pixel to the background interference value, and use it as the follicle response index of the pixel in the region of interest of the ovarian stroma.

[0097] Since the local mapping signal value reflects the effective intensity of the signal at a certain pixel, and the background interference value reflects the amplitude of the background interference at a certain pixel, the ratio of the two, i.e., the signal-to-noise ratio (SNR), is a classic indicator for measuring whether a hormone signal can be clearly identified from background interference. Specifically, a larger follicle response index at a pixel indicates that the hormone signal is more prominent relative to background interference, and the follicle at that pixel is more likely to recognize hormone signals from the background interference; conversely, a smaller follicle response index at a pixel indicates that the hormone signal is more easily submerged by background interference, and the follicle at that pixel is less likely to respond effectively.

[0098] 3) A two-dimensional matrix corresponding to the region of interest of the ovarian stroma is constructed from the follicle response indices of all pixels, which serves as the follicle response map.

[0099] The assessment module 105 is used to assess the functional risk level of the target subject based on the follicle response map; identify the dominant factor type leading to functional abnormalities based on the matrix interference penalty value; and generate individualized strategies based on the risk level and dominant factor type.

[0100] To accurately assess the functional risk level of the target object, as an example, a recruitment threshold is determined based on the statistical distribution characteristics of the follicle response index of all pixels in the follicle response map. Pixels in the follicle response map whose follicle response index is not less than the recruitment threshold are marked as valid response points. The first number of valid response points is counted, and the second number of all pixels in the region of interest of the ovarian stroma is counted. The ratio of the first number to the second number is calculated as the effective region percentage. The effective region percentage is compared with the preset first threshold and the second threshold, wherein the first threshold is greater than the second threshold. If the effective region percentage is not less than the preset first threshold, a low-risk level is generated. If the effective region percentage is between the second threshold and the first threshold, a medium-risk level is generated. If the effective region percentage is not greater than the second threshold, a high-risk level is generated.

[0101] The recruitment threshold is the boundary value that distinguishes between the effective region and the low-response region. This embodiment does not limit the specific method of determining the recruitment threshold. For example, the median or a specific percentile (such as the 60th percentile) of the overall follicle response index can be used as the recruitment threshold.

[0102] An effective response point represents a pixel where the follicle can clearly identify hormone commands from background interference and has the functional potential to be recruited normally.

[0103] The effective area percentage reflects the proportion of the ovarian stroma that can effectively respond to hormonal commands. A larger effective area percentage indicates more functionally intact areas; a smaller effective area percentage indicates more functionally silent areas.

[0104] It should be noted that the specific values ​​of the preset first and second thresholds can be obtained based on statistical analysis of clinical sample data, and this embodiment does not impose specific limitations. The specific determination process is as follows: collect a large amount of follicle response map data and actual ovulation induction response outcomes from patients undergoing ovulation induction treatment. Using methods such as characteristic curve analysis, find the optimal cutoff values ​​that can distinguish different response levels, i.e., the preset first and second thresholds. For example, the first threshold can be set to 60%, and the second threshold can be set to 30%.

[0105] It is important to understand that if the effective area percentage is not less than the preset first threshold, it indicates that most of the ovarian stromal area is functionally intact, and the remaining follicles can still be effectively recruited, resulting in a low-risk level. If the effective area percentage is between the second and first thresholds, it indicates that the functional silencing zone is beginning to expand, and conventional doses of hormones may not be sufficient to induce enough follicle development, resulting in a medium-risk level. If the effective area percentage is not greater than the preset second threshold, it indicates that most of the ovarian stromal area is in a functional silencing state, indicating a very poor ovulation induction response, resulting in a high-risk level.

[0106] To fully identify the dominant factor type leading to functional abnormalities, as an example, low-response regions below the recruitment threshold are identified based on the follicle response map. Low-response regions are areas comprised of pixels whose follicle response index is less than the recruitment threshold. For each pixel within a low-response region, the first and second components of the pixel's matrix interference penalty value are obtained. The arithmetic mean of the first components of all pixels within the low-response region is calculated as the first energy representative value. The arithmetic mean of the second components of all pixels within the low-response region is calculated as the second energy representative value. If the first energy representative value is greater than the second energy representative value, the dominant factor type is determined to be centrally dominant; if the first energy representative value is less than the second energy representative value, the dominant factor type is determined to be locally dominant.

[0107] Since the first component characterizes residual noise originating from the instability of the central drive signal, the first energy representative value reflects the average intensity of the central drive signal noise in the low-response region.

[0108] It should be noted that before calculating the first and second energy representative values, Z-score normalization processing needs to be performed on the first and second components of all pixels in the low-response region. Z-score normalization is a common technique in this field and will not be elaborated upon in this embodiment.

[0109] Since the second component characterizes matrix noise originating from local matrix scattering, i.e., newly generated local interference when hormone signals interact with fibrotic tissue, the second energy representative value reflects the average intensity of matrix noise in low-response regions.

[0110] It is important to understand that if the first energy representative value is greater than the second energy representative value, it indicates that the background interference in the low response area mainly originates from the instability of the central drive, and the dominant factor type is determined to be centrally dominant; if the first energy representative value is less than the second energy representative value, it indicates that the background interference in the low response area mainly originates from the scattering effect of the local matrix, and the dominant factor type is determined to be locally dominant.

[0111] For example, for low-risk cases, a standard ovulation induction protocol can be recommended; for medium- to high-risk cases where the dominant factor is centrally dominant, a central downregulation pretreatment protocol can be recommended; and for medium- to high-risk cases where the dominant factor is locally dominant, an ovarian stromal microenvironment conditioning protocol can be recommended.

[0112] It should be noted that the above strategies are merely illustrative suggestions. Actual clinical decisions should be made by physicians based on a comprehensive assessment of the patient's specific circumstances. The individualized strategies output by this system are for clinical reference only and do not constitute mandatory treatment plans.

[0113] 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.

[0114] 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 multi-omics data-driven risk assessment system for diminished ovarian reserve, characterized in that, The system includes: The acquisition module is used to acquire historical hormone testing data and ovarian ultrasound image data of the target object; perform image segmentation on the ovarian tissue in the ultrasound image data, and extract the region of interest of the ovarian stroma. The matrix resistance module is used to determine the local resistance value of each pixel in the region of interest (ROI) based on the texture features of the RIO and to construct a matrix resistance map; it also locates edge reference anchor points based on the anatomical structure of the RIO. The hormone and noise module is used to determine the cumulative resistance value of each pixel based on the resistance map, starting from the edge reference anchor point; to screen the blood supply source point of each pixel based on the cumulative resistance value; to determine the local mapping signal value of each pixel in order of increasing cumulative resistance value according to the historical hormone detection data corresponding to the edge reference anchor point; and to determine the matrix interference penalty value of each pixel based on the local mapping signal value, cumulative resistance value and matrix interference penalty value of its blood supply source point. The synthesis module is used to determine the follicle response index of each pixel based on the local mapping signal value and the matrix interference penalty value, and to construct the follicle response map. The assessment module is used to evaluate the functional risk level of the target subject based on the follicle response map; identify the dominant factor type leading to functional abnormalities based on the matrix interference penalty value; and generate individualized strategies based on the risk level and dominant factor type.

2. The multi-omics data-driven risk assessment system for diminished ovarian reserve according to claim 1, characterized in that, The matrix resistance map construction process includes: For each pixel within the region of interest of the ovarian stroma, a sliding window is constructed with the pixel as the center; the gray-level co-occurrence matrix corresponding to the gray values ​​of the pixels within the sliding window is statistically analyzed and calculated. The contrast of a pixel is calculated based on the contrast characteristics of the gray-level co-occurrence matrix; the homogeneity index of a pixel is calculated based on the homogeneity characteristics of the gray-level co-occurrence matrix. The sum of the homogeneity index and the preset minimum constant is calculated as the first sum value; the ratio of the contrast to the first sum value is calculated as the local resistance value of the pixel in the region of interest of the ovarian stroma. A two-dimensional matrix corresponding to the region of interest in the ovarian stroma is constructed from the local resistance values ​​of all pixels, serving as the stroma resistance map.

3. The multi-omics data-driven risk assessment system for diminished ovarian reserve according to claim 1, characterized in that, The step of locating edge reference anchor points based on the anatomical structure of the region of interest includes: Determine the geometric center and edge points of the region of interest in the ovarian stroma; and acquire Doppler blood flow images that match the ovarian ultrasound imaging data; Calculate the Euclidean distance from each edge point to the geometric center, and use it as the geometric characteristic value of each edge point; Based on the Doppler blood flow image, points containing blood flow signals among the edge points are selected as candidate edge reference anchor points; Select the edge point with the largest geometric feature value from the candidate edge reference anchor points as the edge reference anchor point; if there are no candidate edge reference anchor points, select the edge point with the largest geometric feature value from all edge points as the edge reference anchor point.

4. The multi-omics data-driven risk assessment system for diminished ovarian reserve according to claim 1, characterized in that, The process of determining the cumulative resistance value includes: Starting from the edge reference anchor point, set its cumulative resistance value to zero; Starting from the edge reference anchor point, proceed outward point by point. For the pixel to be calculated, extract its neighboring pixels within a preset neighborhood; from the neighboring pixels, select the target neighboring pixels with a determined cumulative resistance value. For each target neighboring pixel, calculate the Euclidean distance between the target neighboring pixel and the pixel to be calculated; calculate the product of the distance and the local resistance value of the pixel to be calculated as the process resistance value; calculate the sum of the cumulative resistance value of the target neighboring pixels and the process resistance value as the neighbor cumulative resistance value. Take the minimum value among the cumulative resistance values ​​of all neighboring pixels of the target, and use it as the cumulative resistance value of the pixel to be calculated.

5. The multi-omics data-driven risk assessment system for diminished ovarian reserve according to claim 4, characterized in that, The method of filtering the blood supply source points for each pixel based on cumulative resistance values ​​includes: For each pixel in the region of interest of the ovarian stroma, the cumulative resistance value of each target neighboring pixel is compared, and the target neighboring pixel with the smallest cumulative resistance value is determined as the blood supply source point of the pixel. If multiple adjacent target pixels have the same minimum cumulative resistance value, one of them is selected as the blood supply source point of the pixel according to the preset priority rule; A blood supply path network with clear upstream and downstream relationships is formed by the blood supply source points of all pixels.

6. The multi-omics data-driven risk assessment system for diminished ovarian reserve according to claim 1, characterized in that, The historical hormone detection data consists of at least two baseline follicle-stimulating hormone (FSH) measurements collected during the target subject's preset physiological period; the determination of the local mapping signal value of each pixel based on the historical hormone detection data corresponding to the edge reference anchor point, in ascending order of cumulative resistance value, includes: Calculate the arithmetic mean of all baseline follicle-stimulating hormone (FSH) measurements in historical hormone testing data as the first indicator; assign the first indicator to the local mapping signal value of the edge reference anchor point. All pixels within the region of interest of the ovarian stroma are sorted in ascending order of cumulative resistance value. Starting from the edge reference anchor point with the smallest cumulative resistance value, all other pixels except the edge reference anchor point are processed sequentially. For each other pixel being processed, calculate the Euclidean distance between that pixel and its blood supply source point; Based on the local resistance values ​​of other pixels and the Euclidean distance, the feature signal attenuation rate during the process of moving from the blood supply source point to other pixels is calculated; wherein, the feature signal attenuation rate is negatively correlated with the product of the local resistance value and the Euclidean distance; Multiply the local mapping signal value of the blood supply source point of other pixels by the feature signal attenuation rate to obtain the local mapping signal value of other pixels; until all other pixels have obtained the corresponding local mapping signal value.

7. The multi-omics data-driven risk assessment system for diminished ovarian reserve according to claim 6, characterized in that, The process for determining the matrix interference penalty value for each pixel includes: Calculate the variance of all baseline follicle-stimulating hormone (FSH) measurements in historical hormone testing data as a second indicator; assign the second indicator the matrix interference penalty value of the marginal reference anchor point. Starting from the edge reference anchor point, other pixels are processed sequentially in order of increasing cumulative resistance value. For each pixel being processed, the square of the feature signal attenuation rate during the process from its blood supply source point to other pixels is calculated as the signal attenuation index. The matrix interference penalty value of the blood supply source point of other pixels is multiplied by the signal attenuation index to obtain the first component. The product of the local mapping signal value of other pixels, the local resistance value of other pixels, and the preset matrix interference coefficient is calculated as the second component; The sum of the first component and the second component is calculated as the matrix interference penalty value for other pixels; this continues until all other pixels have obtained the corresponding matrix interference penalty value.

8. The multi-omics data-driven risk assessment system for diminished ovarian reserve according to claim 1, characterized in that, The process of constructing the follicle response map includes: For each pixel within the region of interest of the ovarian stroma, the sum of the stroma interference penalty value of the pixel and the preset minimum constant is calculated as the second sum value; the square root operation of the second sum value is performed to obtain the background interference level value of the pixel. The ratio of the local mapping signal value of a pixel to the background interference value is calculated and used as a follicle response index for pixels within the region of interest of the ovarian stroma. A two-dimensional matrix corresponding to the region of interest in the ovarian stroma is constructed from the follicle response indices of all pixels, serving as the follicle response map.

9. A multi-omics data-driven risk assessment system for diminished ovarian reserve according to claim 7, characterized in that, The assessment of the functional risk level of the target subject based on the follicle response map includes: The recruitment threshold is determined based on the statistical distribution characteristics of the follicle response index of all pixels in the follicle response map. Pixels in the follicle response map whose follicle response index is not less than the recruitment threshold are marked as valid response points. The first number of valid response points is counted, and the second number of all pixels in the ovarian stroma region of interest is counted. Calculate the ratio of the first quantity to the second quantity, and use this as the percentage of the effective area; The effective area percentage is compared with a preset first threshold and a second threshold; wherein the first threshold is greater than the second threshold. If the proportion of effective areas is not less than the preset first threshold, a low-risk level is generated; if the proportion of effective areas is between the second threshold and the first threshold, a medium-risk level is generated; if the proportion of effective areas is not greater than the second threshold, a high-risk level is generated.

10. A multi-omics data-driven risk assessment system for diminished ovarian reserve according to claim 9, characterized in that, The identification of dominant factors leading to functional abnormalities based on matrix interference penalty values ​​includes: Low response regions below the recruitment threshold are identified based on the follicle response map. Low response regions are areas composed of pixels whose follicle response index is less than the recruitment threshold. For each pixel in the low response region, obtain the first and second components of the pixel matrix interference penalty value; Calculate the arithmetic mean of the first component of all pixels in the low response region as the first energy representative value; calculate the arithmetic mean of the second component of all pixels in the low response region as the second energy representative value. If the first energy representative value is greater than the second energy representative value, the dominant factor type is determined to be centrally dominant; if the first energy representative value is less than the second energy representative value, the dominant factor type is determined to be locally dominant.

Images