AI-based Endometriosis Management System

Abstract

This invention relates to the field of medical information management technology, including an AI-based endometriosis management system. The system comprises a thickness dynamic monitoring module, an inflammation level assessment module, a lesion spread analysis module, a periodic lesion tracking module, and a disease progression analysis module. In this invention, AI is used to segment ultrasound image data, enabling dynamic monitoring of lesion areas. Lesion boundaries are extracted, and changes in local inflammatory marker concentrations are calculated, making the screening of abnormal inflammatory areas more accurate and enabling early identification of lesion development. The tracking of periodic lesions, combined with analysis of lesion contour changes and area increases / decreases, enhances the quantitative control of disease progression. The future development trend of lesions, combined with the calculation of periodic inflammatory pathways, makes the prediction of disease evolution more consistent with physiological changes. Multi-dimensional lesion analysis combined with AI intelligent calculation expands the diagnosis and treatment of endometriosis from static assessment to dynamic trend prediction, improving the adaptability of personalized treatment plans.

Description

Technical Field

[0001] This invention relates to the field of medical information management technology, and in particular to an AI-based endometriosis management system. Background Technology

[0002] The field of medical information management technology encompasses computer-based information processing, storage, and analysis technologies, aiming to support the efficient management and application of medical data. Core components include electronic medical record storage, medical image processing, clinical decision support systems, and disease diagnosis and predictive analysis. Medical information management technology relies on the collaborative work of computer hardware and software to achieve precise management of medical data and support clinical decision-making through data collection, processing, transmission, and computational analysis. In recent years, the application of artificial intelligence, big data analytics, and deep learning algorithms has driven the intelligent development of this technology field, significantly improving the efficiency and accuracy of medical data processing.

[0003] The AI-based endometriosis management system refers to a system that uses artificial intelligence technology to collect, analyze, and manage the diagnostic and treatment data of endometriosis patients. This includes standardized modeling of endometriosis-related pathological data, natural language processing-based medical record text parsing, deep learning-based medical image recognition, and correlation analysis between clinical symptoms and biomarkers. The system trains a characteristic data model of endometriosis using machine learning methods and combines it with a medical knowledge graph to analyze the disease's progression. Through data cleaning, feature extraction, and classification modeling, structured information that can be used to assist in diagnosis and treatment is formed.

[0004] Traditional endometriosis management relies on single-image imaging for lesion identification, lacking dynamic monitoring and limiting early detection. Inflammation level assessment depends on limited image resolution, failing to accurately measure local inflammation and making it difficult to identify active lesion areas. The lack of rate calculation and direction analysis for lesion spread makes it difficult to accurately control the trend of inflammation transmission, affecting the timeliness of early intervention. Analysis of cyclical lesions relies on historical medical records, lacking quantitative standards for lesion morphological changes, leading to significant uncertainty in assessing disease progression. Disease prediction is based on static data retrospection, failing to establish a cyclical trend calculation model, resulting in a lack of accurate basis for predicting future lesion development trends. Because treatment decisions do not fully incorporate dynamic changes in lesions, targeted adjustments to clinical interventions are difficult, leaving significant room for improvement in the long-term management of endometriosis. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing an AI-based endometriosis management system.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: An AI-based endometriosis management system includes: The thickness dynamic monitoring module acquires continuous ultrasound image data, uses AI to segment ultrasound images, calculates the thickness of the functional layer and basal layer and the thickness change rate of adjacent frames, compares the current thickness growth rate, filters out areas that exceed the baseline threshold, and obtains endometrial functional lesion areas. The inflammation level assessment module, based on the functional lesion area of ​​the endometrium, uses AI to extract the boundary contour of the lesion, collects the concentration of local inflammatory markers in the lesion, calculates the rate of change of inflammatory concentration at different levels, and obtains areas with abnormal inflammation levels. The lesion spread analysis module calculates the inflammation spread rate around the lesion based on the abnormal inflammation level area, establishes the inflammation spread path, compares the spread rate at the edge of the lesion with the inflammation level gradient, and obtains the early spread area of ​​the lesion. The periodic lesion tracking module uses AI to extract the lesion outline change area and calculate the lesion area change rate based on the early lesion spread area, and obtains the periodic lesion change trend analysis results. Based on the analysis results of the periodic lesion change trend, the disease development analysis module calculates the differences in the inflammation spread path of the previous cycle, uses AI to calculate the inflammation occurrence area of ​​the next cycle, and obtains the analysis results of the future trend of the lesion.

[0007] As a further aspect of the present invention, the functional lesion region of the endometrium includes regions exceeding a baseline threshold, regions with abnormal thickness growth rates, and regions with abrupt changes in the thickness of the functional layer and the basal layer; the abnormal inflammation level region includes regions where the inflammation concentration changes exceed a set threshold, regions with abnormal concentrations of local inflammatory markers in the lesion, and regions with abnormal spatial distribution gradients of IL-6 and TNF-α; the early lesion spread region includes regions with low spread rates at the lesion edge, regions with high inflammation levels around the lesion, and regions where the initial formation of the inflammation spread path is observed; the periodic lesion change trend analysis results include the lesion area change rate, lesion growth region, and lesion-affected structural region; the future trend analysis results of the lesion include the region where inflammation occurs in the next cycle, the inflammation spread path in the current cycle, and comparative data on the inflammation spread path in the previous cycle.

[0008] As a further aspect of the present invention, the thickness dynamic monitoring module includes: The ultrasound image acquisition submodule acquires continuous ultrasound image data, extracts the pixel matrix of each frame, establishes an image time series, detects the signal continuity between image frames, removes abnormal frames, and generates a stable ultrasound image sequence. The functional layer-basal layer segmentation submodule, based on the stable ultrasound image sequence, uses AI to segment the image pixel matrix, extracts the boundary between the endometrial functional layer and the basal layer, normalizes the coordinates of the boundary point set, establishes the spatial contour relationship between the functional layer and the basal layer, obtains the contour change features between adjacent frames, and generates functional layer-basal layer boundary data. The thickness change rate calculation submodule calculates the thickness of the functional layer and the base layer based on the boundary data of the functional layer and the base layer, extracts the thickness data of adjacent frames for difference calculation, and obtains the thickness change rate using the formula: ; Calculate the anomaly of the rate of change of thickness By combining the cyclical thickness variation range of healthy status, areas exceeding the baseline threshold are screened to obtain functional endometrial lesions, where represents . Representing the Functional layer thickness of frame image This represents the thickness of the functional layer in the previous frame. Represents the number of image frames. This represents the average thickness of all frames.

[0009] As a further aspect of the present invention, the inflammation level assessment module includes: The lesion area extraction submodule extracts ultrasound image data of the lesion area based on the endometrial functional lesion area, identifies the lesion boundary contour, establishes a set of lesion boundary point coordinates, calculates the curvature change of the lesion boundary, removes abnormal boundary points, and obtains lesion area boundary data. The inflammatory marker concentration calculation submodule, based on the lesion region boundary data, collects the concentration data of local inflammatory markers IL-6 and TNF-α, stratifies the lesion region, groups the concentration data according to spatial location, and uses the following formula: ; Calculate the spatial gradient distribution values ​​of IL-6 and TNF-α. The inflammation distribution gradient data were obtained, where, Representing the Concentration of inflammatory markers in the layer This represents the concentration of inflammatory markers in the previous layer. Representing the The coordinates of the layer within the lesion region. , Representing the The coordinates of the layer within the lesion region. Indicates the number of floors; The inflammation concentration change screening submodule calculates the rate of change of inflammation concentration at different levels based on the inflammation distribution gradient data, filters out areas where the inflammation concentration change exceeds a set threshold, and obtains areas with abnormal inflammation levels.

[0010] As a further aspect of the present invention, the lesion diffusion analysis module includes: The lesion edge extraction submodule extracts the pixel distribution of the lesion edge based on the abnormal inflammation level region, performs edge detection on the ultrasound image of the abnormal inflammation level region, selects the region where the pixel gradient change is greater than a set threshold as the initial point set of the lesion edge, calculates the curvature change for the extracted edge point set, removes discrete points with abnormal curvature, fits the edge, calculates the lesion edge change rate, and obtains the lesion edge pixel data. The inflammation spread rate calculation submodule calculates the difference in inflammation level between adjacent pixels based on the lesion edge pixel data, and combines this with time series data using the following formula: ; Calculate the rate of change of pixels at the edge of the lesion Data on the rate of inflammation spread were obtained, among which, Representing the The inflammation level value per pixel, Representing the -1 pixel's inflammation level value, This represents the distance from the pixel to the center of the lesion. Represents the spatial attenuation coefficient. This represents the time interval corresponding to that pixel. Represents the number of pixels; The lesion spread screening submodule, based on the inflammation spread rate data, compares the lesion edge spread rate with the inflammation level gradient, screens areas with high inflammation level and low spread rate, and obtains early lesion spread areas.

[0011] As a further aspect of the present invention, the periodic lesion tracking module includes: The lesion contour change extraction submodule acquires multi-cycle endometrial image data based on the early lesion spread area, aligns the ultrasound images of each cycle, matches the lesion area, performs pixel matching on the lesion contour between image frames, calculates the contour displacement, removes abnormal contour changes caused by image noise, corrects the lesion edge through boundary point curvature analysis, and obtains lesion contour change data. The lesion growth rate calculation submodule calculates the pixel area of ​​the lesion region for each cycle based on the lesion contour change data, and compares the lesion areas of adjacent cycles using the following formula: ; Calculate the lesion growth rate per unit time. Data on lesion growth rate were obtained, among which, Represents the growth rate of the lesion. Representing the The area of ​​lesions in the cycle, Representing the The area of ​​lesions in the cycle, This represents the time interval corresponding to this cycle. This represents the thickness of the functional layer in that cycle. Represents the thickness influencing factor. Indicates the number of cycles; The lesion influence area determination submodule, based on the lesion growth rate data, calls the dynamic data of functional layer thickness and the dynamic data of basal layer thickness, compares the lesion growth area with the thickness change area, calculates the correlation between the lesion growth area and the thickness change, filters the area where the lesion growth and thickness change are consistent, and obtains the analysis results of the periodic lesion change trend.

[0012] As a further aspect of the present invention, the disease progression analysis module includes: The current period inflammation spread extraction submodule obtains the inflammation spread path of the current period based on the analysis results of the periodic lesion change trend, extracts the lesion area in the ultrasound image data, measures the spatial distribution of the concentration of inflammatory markers of pixels around the lesion, calculates the concentration gradient and establishes the inflammation spread direction vector, selects pixels with the same gradient direction to form the inflammation spread path, and obtains the inflammation spread path data of the current period. The inflammation spread path comparison submodule calculates the difference in inflammation spread paths from previous periods based on the current period's inflammation spread path data. For the inflammation spread vector fields of the preceding and following periods, the following formula is used: ; Calculate the degree of change in the inflammatory spread pathway during the differential period. Data on differences in the path of inflammation spread were obtained, among which, For the first The pixel coordinates of the inflammation spread path during the period. For the first The pixel coordinates of the inflammation spread path during the period. For the first Inflammation concentration value at the point For the first Inflammation concentration value at the point This represents the total number of pixels along the path. The lesion future trend prediction submodule calculates the areas where inflammation may occur in the next cycle based on the difference data of the inflammation spread path. Based on the inflammation spread trend of multiple cycles, it predicts the future lesion spread area, selects the area with the most stable spread trend as the high-risk lesion spread area, and obtains the lesion future trend analysis results.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, AI is used to segment ultrasound image data, enabling dynamic monitoring of lesion areas and precise quantification of thickness changes in the functional and basal layers of the endometrium. AI is also used to extract lesion boundaries and calculate changes in local inflammatory marker concentrations, making the screening of abnormal inflammatory areas more accurate. Analysis of diffusion trends, by calculating the rate of inflammatory spread and changes in lesion edges, enables early identification of lesion development, clearly presenting the extent of inflammatory influence. Tracking of periodic lesions, combined with analysis of lesion contour changes and area increases / decreases, enhances quantitative control over the disease process, expanding disease trend prediction from single-point analysis to periodic comparison. The future development trend of lesions, combined with the calculation of periodic inflammatory pathways, makes the prediction of disease evolution more consistent with physiological changes. Precise identification of lesion changes, inflammatory spread, and periodic characteristics allows treatment plans to be built on more refined data support, enhancing the proactive intervention capability of disease management. Multi-dimensional lesion analysis combined with AI intelligent calculation expands the diagnosis and treatment of endometriosis from static assessment to dynamic trend prediction, improving the adaptability of personalized treatment plans. Attached Figure Description

[0014] Figure 1 This is a system flowchart of the present invention; Figure 2 This is a flowchart of the thickness dynamic monitoring module of the present invention; Figure 3 This is a flowchart of the inflammation level assessment module of the present invention; Figure 4 This is a flowchart of the lesion diffusion analysis module of the present invention; Figure 5 This is a flowchart of the periodic lesion tracking module of the present invention; Figure 6 This is a flowchart of the disease progression analysis module of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0016] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0017] Please see Figure 1 The AI-based endometriosis management system includes: The thickness dynamic monitoring module acquires continuous ultrasound image data, uses AI to segment ultrasound images, extracts the boundary between the functional layer and the basal layer of the endometrium, calculates the thickness of the functional layer and the basal layer, calculates the thickness change rate of adjacent frames, combines the periodic thickness change range of the healthy state, compares the current thickness growth rate, filters and marks areas that exceed the baseline threshold, and obtains the functional lesion areas of the endometrium. The inflammation level assessment module is based on the functional lesion area of ​​the endometrium. It extracts ultrasound image data of the lesion area, uses AI to extract the boundary contour of the lesion, collects the concentration of local inflammatory markers in the lesion, calculates the spatial distribution gradient of IL-6 and TNF-α, calculates the rate of change of inflammatory concentration at different levels, and screens areas where the change of inflammatory concentration exceeds the set threshold to obtain areas with abnormal inflammation levels. The lesion spread analysis module extracts the pixel distribution at the edge of the lesion based on the abnormal area of ​​inflammation level, calculates the inflammation spread rate around the lesion, establishes the inflammation spread path, compares the spread rate at the edge of the lesion with the gradient of inflammation level, and filters out areas with high inflammation level and low spread rate to obtain the early area of ​​lesion spread. The periodic lesion tracking module acquires multi-cycle endometrial imaging data based on the early stage of lesion spread, extracts the lesion contour change area through AI, calculates the lesion area change rate, calls dynamic data of functional layer thickness and dynamic data of basal layer thickness, compares the lesion growth area with the thickness change area, determines the structural area affected by the lesion, and obtains the periodic lesion change trend analysis results. The disease progression analysis module obtains the current cycle's inflammation spread path based on the analysis results of the periodic lesion change trend, calculates the difference in inflammation spread path between previous cycles, uses AI to calculate the inflammation occurrence area in the next cycle, and obtains the future trend analysis results of the lesion.

[0018] Endometrial functional lesion areas include areas exceeding the baseline threshold, areas with abnormal thickness growth rates, and areas with abrupt changes in the thickness of the functional layer and basal layer; areas with abnormal inflammation levels include areas where inflammation concentration changes exceed the set threshold, areas with abnormal concentrations of local inflammatory markers, and areas with abnormal spatial distribution gradients of IL-6 and TNF-α; early lesion spread areas include areas with low spread rates at the lesion edge, areas with high inflammation levels around the lesion, and areas where the inflammatory spread path is initially formed; the results of periodic lesion change trend analysis include the rate of change of lesion area, lesion growth areas, and areas of lesion-affected structures; the results of future lesion trend analysis include the area of ​​inflammation occurrence in the next cycle, the inflammatory spread path in the current cycle, and comparative data on the inflammatory spread path in the previous cycle.

[0019] Please see Figure 2 The thickness dynamic monitoring module includes: The ultrasound image acquisition submodule acquires continuous ultrasound image data, extracts the pixel matrix of each frame, establishes an image time series, detects the signal continuity between image frames, removes abnormal frames, and generates a stable ultrasound image sequence. In the management of endometriosis, the first step is to acquire continuous ultrasound imaging data. Patients undergo transvaginal ultrasound (TVUS). After the ultrasound probe is inserted into the vagina, it collects ultrasound reflection signals from different angles to obtain high-resolution images of the endometrium. 30 frames are acquired per second and stored as a continuous time-series data matrix. The size of each frame is set to [specific dimensions would be inserted here]. Each pixel has a grayscale value ranging from 0 to 255. Image data is stored in a buffer. During the acquisition of images for endometriosis, there may be interfering factors such as abnormal uterine position, ectopic ovarian cysts, and deep invasive lesions. Therefore, image data screening is necessary. First, the grayscale value changes of image frames are detected, and the average grayscale value of each frame is calculated. The normal range is set to 90-140. This range is based on the typical grayscale value statistics of the functional layer of the endometrium in the middle (proliferative phase) of the menstrual cycle in healthy women. Under normal circumstances, the grayscale value of the endometrium in this stage is concentrated in the range of 100±15. Considering individual differences, the range is expanded to 90-140 as a screening range. A standard is selected. If the grayscale value of a frame exceeds this range, the frame is considered to be affected by artifacts and is removed. At the same time, the timestamps between frames are compared. The standard inter-frame interval is set to 0.033 seconds. This value comes from the frame rate of the ultrasound equipment, which samples 30 frames per second. The interval between each frame is 1 / 30 of a second, or 0.033 seconds. The allowable error is set to ±0.005 seconds. This error is based on the synchronization error range of the ultrasound equipment. If the timestamp of a frame exceeds the standard interval of ±0.005 seconds, it is considered an abnormal frame and is removed. Finally, the ultrasound image frames with qualified quality are retained, and a stable ultrasound image sequence is established for subsequent dynamic monitoring of endometrial thickness, resulting in a stable ultrasound image sequence.

[0020] The functional layer-basal layer segmentation submodule is based on a stable ultrasound image sequence. It uses AI to segment the image pixel matrix, extracts the boundary between the endometrial functional layer and basal layer, normalizes the coordinates of the boundary point set, establishes the spatial contour relationship between the functional layer and basal layer, obtains the contour change features between adjacent frames, and generates functional layer-basal layer boundary data. When extracting the boundary between the functional and basal layers of the endometrium based on stable ultrasound image sequences, factors such as uneven thickening of the myometrium or endometriotic cysts in patients with endometriosis need to be considered. Therefore, during segmentation, image enhancement techniques are first used to improve boundary contrast. Then, the grayscale histogram of each frame is analyzed to set grayscale threshold ranges for the functional and basal layers. For example, the functional layer is mainly between 80-160, and the basal layer is mainly between 40-100. Pixels that fit within this grayscale range are selected, and their boundary point coordinates are calculated. To ensure the stability of the boundary points, the Euclidean distance change between boundary points in adjacent frames is calculated using the following formula: ; in, Represents the inter-frame displacement of boundary points, if If the number of pixels is not specified, the frame may have abnormal boundaries that need to be corrected. In addition, considering that patients with endometriosis may have irregular endometrial hyperplasia, the boundary curvature is calculated separately, and the curvature threshold range is set to 0.15-0.45. This range is calculated based on the curvature of normal endometrium in the image. The boundary of the functional layer of healthy endometrium usually shows a smooth structure with a local curvature between 0.15 and 0.3, while the curvature of the lesion area increases, reaching more than 0.45. Therefore, this threshold is set to screen for abnormal boundaries. If the boundary curvature exceeds the normal range, morphological correction is required to obtain the boundary data of the functional layer basal layer.

[0021] The thickness change rate calculation submodule calculates the thickness of the functional layer and the base layer based on the boundary data of the functional layer and the base layer. It extracts the thickness data of adjacent frames, performs difference calculation, and obtains the thickness change rate using the following formula: ; Calculate the anomaly of the rate of change of thickness By combining the cyclical thickness variation range of healthy status, areas exceeding the baseline threshold are screened to obtain functional endometrial lesions, where represents . Representing the Functional layer thickness of frame image This represents the thickness of the functional layer in the previous frame. Represents the number of image frames. This represents the average thickness of all frames. Based on the boundary data of the functional layer and basal layer, the rate of change of endometrial functional layer thickness was calculated to monitor the thickness change pattern in endometriosis patients during different menstrual cycles. Endometrial thickness was defined as the vertical distance between the upper boundary of the functional layer and the lower boundary of the basal layer, and the calculation formula is as follows: ; in, Indicates the first The functional layer thickness of the frame image may exhibit periodic abnormal thickness changes in patients with endometriosis; therefore, the thickness change between adjacent frames is calculated. ; To further analyze the abnormal growth rate of the lesion's thickness, it is necessary to calculate the degree of abnormality in the rate of thickness change. .

[0022] Table 1.1 Endometrial thickness monitoring data

[0023] Calculate the change: ; ; ; ; Calculate the average thickness: ; Calculate the standard deviation: ; calculate : ; An abnormal threshold of endometrial thickness change rate was set at 0.5. This threshold is based on statistical data of endometrial thickness growth rate in healthy women at different stages of the menstrual cycle. Under normal circumstances, the thickness change rate during the proliferative phase ranges from 0.3 to 0.5 mm / day. If the rate exceeds 0.5 mm / day, it may indicate pathological hyperplasia or ectopic tissue growth; therefore, this threshold was set as a judgment criterion. If the calculated... If the result is within the normal range, it indicates that there are no endometriosis lesions in the current situation. This result shows that the increase in endometrial thickness in a short period of time is normal. If it exceeds the normal range, it indicates that the increase in endometrial thickness in a short period of time is abnormal, which may be endometriosis-related lesions. Further examination with ultrasound imaging and pathological analysis is needed to ultimately obtain the functional lesion area of ​​the endometrium.

[0024] Please see Figure 3 The inflammation level assessment module includes: The lesion area extraction submodule extracts ultrasound image data of the lesion area based on the endometrial functional lesion area, identifies the lesion boundary contour, establishes a set of lesion boundary point coordinates, calculates the curvature change of the lesion boundary, removes abnormal boundary points, and obtains lesion area boundary data. Based on the functional lesion region of the endometrium, ultrasound image data of the lesion region is extracted. First, the image data needs to be preprocessed. To address noise interference in the ultrasound images, a fixed threshold denoising method is used, setting pixels with noise intensity below 20 grayscale values ​​as background points and removing them to obtain a data matrix with optimized image clarity. Based on this data, the lesion boundary contour is identified, and the boundary region is calculated using the pixel gradient change rate. Pixels with gradient changes exceeding 50 are selected as the initial boundary point set. Further filtering of boundary points is performed through curvature calculation, calculating the angle change rate of adjacent pixels. If the angle change exceeds 90°, the point is considered an abnormal boundary point and is removed, ultimately forming a smooth boundary. The lesion boundary contour data is used to set the 90° angle change threshold. This threshold is based on the fact that normal lesion boundary angle changes are typically within the range of 45° to 75°, and changes exceeding 90° are usually caused by abnormal points. Furthermore, this value is affected by the complexity of the lesion morphology; the more complex the lesion contour, the smaller the fluctuation range of this value, thus enhancing boundary stability. In this process, for common lesion shapes, such as regular circles, ellipses, or irregular protrusions, methods such as ellipse fitting and morphological closure operations are used to optimize the boundary to ensure the stability and continuity of the contour. The closure parameter of the lesion contour is calculated; if the closure is below 0.85, further boundary compensation is performed. The closure calculation is based on the perimeter of the contour. and area The formula is In the calculation of multiple lesion contour samples, if the closure degree is lower than 0.85, the integrity of the lesion contour is insufficient, and there are broken or discontinuous areas. Therefore, this value is set as the compensation trigger threshold. This threshold changes with the regularity of the lesion. For lesions with strong regularity, this value can be relaxed to 0.9, while for lesions with weak regularity, it should be set to 0.8 to ensure morphological continuity and finally generate lesion region boundary data.

[0025] The inflammatory marker concentration calculation submodule collects local inflammatory marker concentration data of IL-6 and TNF-α based on lesion region boundary data, stratifies the lesion interior region, groups the concentration data according to spatial location, and uses the following formula: ; Calculate the spatial gradient distribution values ​​of IL-6 and TNF-α. The inflammation distribution gradient data were obtained, where, Representing the Concentration of inflammatory markers in the layer This represents the concentration of inflammatory markers in the previous layer. Representing the The coordinates of the layer within the lesion region. , Representing the The coordinates of the layer within the lesion region. Indicates the number of floors; Based on the boundary data of the lesion area, the concentration data of local inflammatory markers IL-6 and TNF-α were collected. First, several detection points were selected within the lesion area, and samples were uniformly collected according to spatial distribution, with the spacing between each detection point controlled within 0.5 mm to improve spatial resolution. For different detection points, the concentration data of IL-6 and TNF-α were collected, and a concentration matrix was established. Each row of the matrix corresponds to one detection point, and each column stores the concentration values ​​of IL-6 and TNF-α, respectively. Based on this data matrix, the inflammatory distribution gradient at different spatial locations was calculated. For this purpose, the coordinates of the detection points were set as follows: The corresponding IL-6 concentration is TNF-α concentration was For adjacent detection points, the inflammation distribution gradient is calculated. By calculating the gradient of multiple detection points within the lesion, the spatial distribution of inflammatory markers in the entire lesion area is obtained. If the gradient value of a certain area is higher than 1.5 times the average gradient value of the lesion, it is considered that there is an abnormal distribution of inflammation in that area, and finally the inflammation distribution gradient data is obtained.

[0026] There are four detection points within the lesion area, and their IL-6 concentrations and coordinates are shown in Table 2.1:

[0027] According to Table 2.1, calculate the Euclidean distance from point 2 to point 3: ; Calculate the IL-6 gradient: ; If this value is higher than 1.5 times the average gradient value of the lesion, there may be an abnormal distribution of inflammation in the area.

[0028] The inflammation concentration change screening submodule calculates the rate of change of inflammation concentration at different levels based on inflammation distribution gradient data, filters out areas where the change of inflammation concentration exceeds a set threshold, and obtains areas with abnormal inflammation levels. Based on the inflammatory distribution gradient data, the rate of change of inflammatory concentration at different levels was calculated. To this end, the lesion area was first divided into several layers along the depth direction, with a layer spacing of 0.2 mm. The average value of the data points within each layer was taken as the representative concentration value for that layer. Let the concentrations of IL-6 and TNF-α in a certain layer be... and For adjacent layers, calculate the concentration change rate, assuming the first layer... The concentration of the layer is , No. The concentration of the layer is The formula for calculating the concentration change rate is as follows: ; in, Represents the rate of concentration change between levels, if If the value exceeds a set threshold of 0.3, the inflammation concentration in that area is considered to be abnormal. The threshold of 0.3 is based on the fact that in healthy tissue, the concentration changes of IL-6 and TNF-α are usually stable within the range of 0.1 to 0.2. When the rate of change exceeds 0.3, it usually corresponds to a sharp change in the inflammatory area. If the value further exceeds 0.5, it represents a trend of lesion deterioration. This threshold varies with different tissue types. It can be set to 0.25 in areas with mild inflammation, while it can be appropriately increased to 0.35 in high-risk lesion areas to enhance sensitivity and ultimately screen out areas with abnormal changes in inflammation concentration.

[0029] Please see Figure 4 The lesion spread analysis module includes: The lesion edge extraction submodule extracts the pixel distribution of the lesion edge based on the abnormal region of inflammation level, performs edge detection on the ultrasound image of the abnormal region of inflammation level, selects the region where the pixel gradient change is greater than the set threshold as the initial point set of the lesion edge, calculates the curvature change for the extracted edge point set, removes discrete points with abnormal curvature, fits the edge, calculates the lesion edge change rate, and obtains the lesion edge pixel data. Based on regions with abnormal inflammation levels, the pixel distribution at the lesion edges is extracted. First, ultrasound image data of the lesion region is acquired and grayscale normalized to adjust pixel values ​​to the range of 0-255, improving edge feature contrast. Then, a gradient detection method is used to identify the lesion edge region. A pixel gradient change threshold of 40 is set, and pixels with gradient changes greater than this threshold are selected as candidate edge points. Based on this, the curvature change of edge points is calculated. Curvature calculation is based on a three-point fitting method, calculating the rate of change of the angle between three adjacent pixels. If the angle change of a pixel is greater than 85°, then... These points are then marked as potential outliers, and their gradient differences with neighboring points are further evaluated. If the gradient difference between neighboring points exceeds a set threshold of 20, they are identified as true edge points, and isolated high-angle points are removed to reduce false edge identification. In this process, contour closure is calculated for the edge features of circular, elliptical, and irregular lesions. If the closure is less than 0.9, boundary interpolation is used to compensate for the edge, optimize the lesion edge morphology, calculate the coordinate difference of pixels, and calculate the rate of change of the lesion edge based on the inter-frame time change to generate lesion edge pixel data.

[0030] The inflammation spread rate calculation submodule calculates the difference in inflammation level between adjacent pixels based on lesion edge pixel data, and combines this with time series data using the following formula: ; Calculate the rate of change of pixels at the edge of the lesion Data on the rate of inflammation spread were obtained, among which, Representing the The inflammation level value per pixel, Representing the -1 pixel's inflammation level value, This represents the distance from the pixel to the center of the lesion. Represents the spatial attenuation coefficient. This represents the time interval corresponding to that pixel. Represents the number of pixels; Based on pixel data at the lesion edge, the rate of inflammation spread around the lesion is calculated. For each pixel at the lesion edge, the pixel distribution around it is extracted, the difference in inflammation level between adjacent pixels is calculated, and combined with time series data, the rate of change of pixels at the lesion edge is calculated, with a set detection time interval. For each edge pixel, calculate its inflammation level value. The difference from the previous time step, taking into account the attenuation due to the distance from the lesion center to the pixel, is used to calculate the inflammation spread rate using a formula, with an exponential decay factor introduced in the molecule. To take into account the influence of distance during the diffusion process, the area closer to the core of the lesion contributes more to the diffusion rate, while the area farther away from the lesion has a smaller impact, thus better reflecting the actual physiological characteristics of inflammation diffusion.

[0031] The lesion has 4 pixels at its edge, and the changes in its inflammation level are shown in the table below: Table 3.1 Calculation of Inflammation Spread Rate

[0032] Calculate the diffusion rate contribution of the second pixel: ; Similarly, the contribution values ​​of the remaining pixels are calculated and summed to finally obtain the inflammation spread rate data.

[0033] The lesion spread screening submodule is based on inflammation spread rate data. It compares the spread rate at the edge of the lesion with the gradient of the inflammation level, and screens out areas with high inflammation level and low spread rate to obtain the early lesion spread area. Based on inflammation spread rate data, by comparing the spread rate at the lesion edge with the inflammation level gradient, regions with high inflammation levels and low spread rates are selected. First, the inflammation level gradient of each pixel at the lesion edge is calculated, and the gradient value of a certain pixel is set as... Calculate its average diffusion rate in spatial distribution. Set the filtering threshold and ,in This value represents an area with a high inflammatory gradient. It is derived from the statistical characteristics of changes in inflammation levels within the lesion area. Typically, in normal tissue areas, the gradient value is distributed between 3 and 7, while the inflammatory gradient in lesion areas is usually above 8. Therefore, it is set as follows: As a benchmark for judging the risk zone of lesion spread, this value is affected by the lesion morphology. If the boundary of the lesion area is irregular and the gradient value fluctuates greatly, it can be appropriately increased to 12 to reduce misjudgment; similarly, A value of 2.0 indicates a region with a low diffusion rate. This value is based on time-series data of pixels at the lesion edge. Diffusion rates typically range from 2.5 to 4.5. When the diffusion rate is below 2.0, it means the inflammation has not effectively spread outwards and may be in a state of blocked diffusion or local stagnation. Therefore, this value can be used to identify areas in the early stages of lesion spread. This value can also be adjusted with the time step. For example, when the time step is shortened to 1 second, the value can be relaxed to 2.5 to ensure dynamic adaptability in diffusion assessment. If a pixel meets the criteria... and If so, it is marked as an early lesion spread area, and eventually an early lesion spread area is generated.

[0034] Inflammation level gradient at a certain pixel Its diffusion rate is 12.5. The calculated value is 1.46, therefore it satisfies the condition. and This pixel was identified as an early diffusion region.

[0035] Table 3.2 Calculation Table for Lesion Spread Screening

[0036] As shown in Table 3.2, pixels 2 and 4 have higher inflammatory gradients and lower diffusion rates, and are therefore identified as early diffusion regions, thus obtaining the early diffusion regions of the lesions.

[0037] Please see Figure 5 The periodic lesion tracking module includes: The lesion contour change extraction submodule acquires multi-cycle endometrial image data based on the early stage of lesion spread, aligns the ultrasound images of each cycle, matches the lesion area, performs pixel matching on the lesion contour between image frames, calculates the contour displacement, removes abnormal contour changes caused by image noise, corrects the lesion edge through boundary point curvature analysis, and obtains lesion contour change data. Based on the early stage of lesion spread, multi-cycle endometrial imaging data was acquired. First, the ultrasound images from each cycle were preprocessed. Image registration methods were used to spatially align images from adjacent cycles, reducing lesion region offset errors caused by changes in scanning angle. The pixel offset between different cycles was calculated, and points with offsets exceeding twice the standard deviation of the average were removed to ensure stable boundaries of the matched lesion regions. Pixel matching was performed on the lesion contours between image frames, using pixel grayscale gradients as features to search for lesion boundary points. The gradient difference between adjacent points was calculated, and a threshold was set to filter boundary points with significant changes. For matched boundary points, the displacement vector of the lesion contour was calculated to further determine the continuity of the boundary. If the displacement direction of a point deviates from the direction of its neighboring points by more than 30°, the boundary is considered closed. If a point is deemed an abnormal boundary point, it is considered to be excluded. The 30° threshold is set based on the spatial continuity analysis of the lesion boundary. During the matching of images in different cycles, the normal growth offset angle of the lesion edge is usually distributed in the range of 10° to 25°. When the direction change of a point exceeds 30°, it is usually due to a sudden change caused by image noise or mismatch. Therefore, 30° is set as the abnormal exclusion threshold. This value fluctuates with the stability of the lesion growth direction. If the lesion shape is relatively regular, the value can be relaxed to 35°. For lesions with more complex shapes, the value needs to be reduced to 25° to enhance the contour matching accuracy. By calculating the curvature of the boundary point, the area with prominent curvature changes is smoothed. Quadratic spline interpolation is used to compensate for the broken boundary points, and finally the lesion contour change data is obtained.

[0038] The lesion growth rate calculation submodule calculates the pixel area of ​​the lesion region for each cycle based on lesion contour change data, and compares the lesion areas of adjacent cycles using the following formula: ; Calculate the lesion growth rate per unit time. Data on lesion growth rate were obtained, among which, Represents the growth rate of the lesion. Representing the The area of ​​lesions in the cycle, Representing the The area of ​​lesions in the cycle, This represents the time interval corresponding to this cycle. This represents the thickness of the functional layer in that cycle. Represents the thickness influencing factor. Indicates the number of cycles; Based on lesion contour change data, the rate of change of lesion area is calculated. For the lesion region in each cycle, the lesion contour is first closed, and the boundary points are constructed into a continuous contour curve. The pixel area of ​​the closed region is calculated, and a pixel size conversion coefficient is set to convert the pixel area into an actual area value. The lesion area of ​​adjacent cycles is compared to calculate the lesion growth rate per unit time. The lesion growth rate is calculated using a formula. Based on the correlation analysis between changes in functional layer thickness and lesion growth rate, the trend of lesion growth rate variation and functional layer thickness are positively correlated in multiple lesion data samples, with correlation coefficients ranging from 0.65 to 0.85. Setting 0.1 as an influencing factor allows for adjustment of the weight of lesion growth rate calculation in different periods, so that it can reflect the influence of thickness on lesion growth without over-amplifying the influence of thickness fluctuations. If the lesion is in the proliferative phase, it can be appropriately increased to 0.15, while when the lesion tends to stabilize, it can be reduced to 0.05.

[0039] Assume the lesion area measurements over three cycles are as follows: Table 4.1 Changes in the area of ​​lesions over time

[0040] Calculate the lesion growth rate for cycles 1 to 2: ; Calculate the lesion growth rate in cycles 2 and 3: ; As shown in Table 4.1, the growth rate of the lesion fluctuated slightly with the increase of thickness in different cycles, and the lesion growth rate data were finally obtained.

[0041] The lesion impact area determination submodule uses lesion growth rate data, calls dynamic data of functional layer thickness and dynamic data of basal layer thickness, compares the lesion growth area with the thickness change area, calculates the correlation between the lesion growth area and the thickness change, filters the area where lesion growth and thickness change are consistent, and obtains the analysis results of periodic lesion change trend. Based on lesion growth rate data, dynamic data on functional layer thickness and basal layer thickness are retrieved. First, data on the thickness of the functional layer and basal layer for each cycle are collected and a time series is established. For the lesion growth region and thickness change region, a region matching method is used to calculate the thickness change of the lesion growth region in different cycles. For the matched region, the angle between the lesion expansion direction and the thickness change direction is calculated. If the angle is less than 15°, the lesion growth direction is considered to be consistent with the thickness change direction. This 15° threshold is set based on lesion expansion trajectory analysis and the lesion growth trend... The angle usually aligns with the direction of thickness change in the surrounding tissue. When the angle is less than 15°, it means the two directions are highly consistent. However, when the angle is greater than 15°, the directional deviation increases, potentially indicating that lesion growth is no longer affected by thickness. Adjusting this value depends on the lesion's spread trend. If the lesion's spread path is relatively stable, it can be relaxed to 20°. If the lesion's spread path is unstable, this value needs to be reduced to 10°. Further calculation of the correlation coefficient between the lesion growth rate and the thickness change rate is then performed. If the correlation coefficient is greater than 0.8, it is considered that lesion growth is significantly affected by thickness changes. The 0.8 threshold is based on correlation analysis results. Clinical data shows that when the correlation coefficient between lesion growth rate and thickness change rate exceeds 0.8, the linear relationship between the two is significant, indicating that lesion growth is directly affected by thickness changes. When this value is less than 0.6, the correlation decreases. Therefore, 0.8 is set as the standard for judging whether lesions are affected by thickness. This value can be adjusted according to specific circumstances. If the lesion spreads rapidly, it can be increased to 0.85, while if the lesion changes slowly, it can be decreased to 0.75. Regions that meet this condition are screened, and their proportion is calculated. If this region accounts for... If the percentage of the lesion affected by thickness exceeds 50%, it is considered that the lesion is mainly affected by the thickness of the functional layer or the basal layer. This 50% standard is based on the spatial distribution of the lesion area. When the proportion of the part affected by thickness in the lesion expansion area exceeds 50%, it indicates that the growth pattern of the lesion is dominated by the thickness change. If the value is less than 40%, it indicates that the thickness change has a weaker impact on the lesion growth. This value can be adjusted according to the lesion morphology. If the lesion morphology is relatively regular, it can be relaxed to 55%, while if the morphology is more complex, it can be reduced to 45%. Finally, the results of the periodic lesion change trend analysis are obtained.

[0042] Please see Figure 6 The disease progression analysis module includes: The current period inflammation spread extraction submodule obtains the inflammation spread path in the current period based on the analysis results of the periodic lesion change trend. It extracts the lesion area in the ultrasound image data, measures the spatial distribution of the concentration of inflammatory markers in the pixels around the lesion, calculates the concentration gradient and establishes the inflammation spread direction vector, selects pixels with the same gradient direction to form the inflammation spread path, and obtains the inflammation spread path data in the current period. Based on the analysis of periodic lesion change trends, the inflammation spread path of the current period is obtained. First, the lesion area is extracted from the ultrasound image data, and the pixel grayscale range of the lesion area is set. Pixels with grayscale values ​​higher than a set threshold are selected as lesion boundary points. The spatial distribution of the lesion boundary point set is calculated. Further, the concentration of inflammatory markers in pixels around the lesion is extracted, and the concentration of all pixels is measured. The change in inflammation level at the lesion edge is calculated using the concentration gradient change rate. Pixels with a concentration change exceeding 15% are selected as the initial point set of the inflammation spread area. The 15% threshold is set based on the fact that the inflammation level of healthy tissue typically fluctuates within a certain range. Within 5%, the inflammatory concentration change in the lesion area often exceeds 10%. By analyzing multiple lesion diffusion samples, it was found that when the concentration change exceeds 15%, the diffusion trend in this area is most stable in future cycles. Moreover, this value is affected by the lesion type. For example, in the inflammatory area with slower proliferation, this value can be reduced to 12%, while in the lesion area with rapid progression, this value can be adjusted to 18%. Based on this point set, an inflammatory diffusion direction vector is established, the vector directivity is calculated, and pixels with consistent gradient directions are selected to determine the coherence of the diffusion direction. Based on the directionality of the diffusion vector, the inflammatory diffusion path is constructed, and finally, the inflammatory diffusion path data of the current cycle is obtained.

[0043] Table 5.1 Current Period Inflammation Spread Pixel Data

[0044] As shown in Table 5.1, the inflammatory diffusion path is composed of regions where the concentration gradient changes exceed a set threshold. The changes in the inflammatory concentration of pixels around the lesion can reflect the diffusion process. If the gradient change exceeds the set threshold, it indicates that the region is in an active inflammatory state, which meets the construction conditions of the inflammatory diffusion path.

[0045] The inflammation spread path comparison submodule calculates the difference in inflammation spread paths from previous periods based on the current period's inflammation spread path data. For the inflammation spread vector fields of the preceding and following periods, the following formula is used: ; Calculate the degree of change in the inflammatory spread pathway during the differential period. Data on differences in the path of inflammation spread were obtained, among which, For the first The pixel coordinates of the inflammation spread path during the period. For the first The pixel coordinates of the inflammation spread path during the period. For the first Inflammation concentration value at the point For the first Inflammation concentration value at the point This represents the total number of pixels along the path. Based on the inflammation spread path data of the current cycle, the comparison difference of the inflammation spread path of the previous cycle is calculated. First, the inflammation spread path data of the previous cycle is called. For the inflammation spread vector fields of the current cycle and the previous cycle, the pixel data of the same coordinate points are extracted, the pixel concentration change of adjacent cycles is calculated, and the displacement vector of the corresponding pixel is calculated.

[0046] The degree of change in the inflammatory spread path during different cycles was obtained to obtain data on the differences in inflammatory spread paths. If the Q value is large, it indicates that the direction and intensity of inflammatory spread have changed significantly compared with the previous cycle. The judgment criterion of the Q value is based on the stability analysis of lesion spread. Generally, a Q value in the range of 0.3 to 0.5 represents a relatively stable change in the spread path, while a value exceeding 0.6 indicates that the inflammatory spread pattern has been significantly adjusted, usually accompanied by lesion growth or enhanced reaction of surrounding tissues. This threshold is affected by lesion type and inflammatory factors. For example, for chronic inflammatory lesions, a Q value below 0.4 can be considered normal, while for acute lesions, a Q value above 0.7 can indicate accelerated inflammatory spread.

[0047] Suppose the inflammatory spread pathways of two adjacent cycles are as follows: Table 5.2:

[0048] Calculate the spatial displacement of this point: ; Calculate changes in inflammatory concentrations: ; Calculate the differences in inflammation spread pathways: ; As calculated above, the Q value is 0.612, which exceeds the threshold of 0.6, indicating that the pattern of inflammation spread has changed, which may suggest accelerated lesion progression or enhanced inflammatory response.

[0049] The lesion future trend prediction submodule calculates the areas where inflammation may occur in the next cycle based on the difference data of inflammation spread path. Based on the inflammation spread trend of multiple cycles, it predicts the future lesion spread area, selects the area with the most stable spread trend as the high-risk lesion spread area, and obtains the lesion future trend analysis results. Based on the differential data of inflammatory spread paths, a time series model is used to predict the regions where inflammation may occur in the next cycle. First, for the inflammatory spread path data of multiple cycles, the inflammatory spread trend of each cycle is calculated, and regions with stable spread directions are screened out. A prediction time window is set, and the trend of future cycles is fitted. A lesion trend prediction model is constructed to calculate the probability of lesion spread areas in future cycles. Finally, high-risk lesion areas are screened out, and the future trend analysis results of lesions are obtained. Based on the data of the previous cycle and the current cycle, the inflammatory spread path of future cycles can be predicted. This prediction is based on a comprehensive calculation of the spread rate, changes in inflammatory concentration, and spatial spread trend. Generally, if the inflammatory spread direction is consistent for three consecutive cycles and the path offset is less than 0.2 mm, the region is more likely to spread in the future cycle. If the increase in inflammatory concentration exceeds 20%, the spread range may further expand. If the concentration fluctuates greatly and the path offset exceeds 0.5 mm, the spread path of future cycles may be adjusted.

[0050] Table 5.3 Prediction data of inflammation spread pathways

[0051] As shown in Table 5.3, based on data from previous and current cycles, the path of inflammation spread in future cycles can be predicted up to [date missing]. The location of the diffusion trend score increased to 0.85, indicating that the diffusion probability in this area is relatively high.

[0052] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An AI-based endometriosis management system, characterized in that, The system includes: The thickness dynamic monitoring module acquires continuous ultrasound image data, uses AI to segment ultrasound images, calculates the thickness of the functional layer and basal layer and the thickness change rate of adjacent frames, compares the current thickness growth rate, filters out areas that exceed the baseline threshold, and obtains endometrial functional lesion areas. The inflammation level assessment module, based on the functional lesion area of ​​the endometrium, uses AI to extract the boundary contour of the lesion, collects the concentration of local inflammatory markers in the lesion, calculates the rate of change of inflammatory concentration at different levels, and obtains areas with abnormal inflammation levels. The lesion spread analysis module calculates the inflammation spread rate around the lesion based on the abnormal inflammation level area, establishes the inflammation spread path, compares the spread rate at the edge of the lesion with the inflammation level gradient, and obtains the early spread area of ​​the lesion. The periodic lesion tracking module uses AI to extract the lesion outline change area and calculate the lesion area change rate based on the early lesion spread area, and obtains the periodic lesion change trend analysis results. Based on the analysis results of the periodic lesion change trend, the disease development analysis module calculates the differences in the inflammation spread path of the previous cycle, uses AI to calculate the inflammation occurrence area of ​​the next cycle, and obtains the analysis results of the future trend of the lesion.

2. The AI-based endometriosis management system according to claim 1, characterized in that, The functional lesion areas of the endometrium include areas exceeding the baseline threshold, areas with abnormal thickness growth rates, and areas with abrupt changes in the thickness of the functional layer and basal layer; the areas with abnormal inflammation levels include areas where inflammation concentration changes exceed a set threshold, areas with abnormal concentrations of local inflammatory markers, and areas with abnormal spatial distribution gradients of IL-6 and TNF-α; the early lesion spread areas include areas with low spread rates at the lesion edge, areas with high inflammation levels around the lesion, and areas where the inflammatory spread path initially forms; the periodic lesion change trend analysis results include the lesion area change rate, lesion growth area, and lesion-affected structural area; the future trend analysis results of the lesion include the area where inflammation occurs in the next cycle, the current cycle's inflammatory spread path, and comparative data on the inflammatory spread path of the previous cycle.

3. The AI-based endometriosis management system according to claim 1, characterized in that, The thickness dynamic monitoring module includes: The ultrasound image acquisition submodule acquires continuous ultrasound image data, extracts the pixel matrix of each frame, establishes an image time series, detects the signal continuity between image frames, removes abnormal frames, and generates a stable ultrasound image sequence. The functional layer-basal layer segmentation submodule, based on the stable ultrasound image sequence, uses AI to segment the image pixel matrix, extracts the boundary between the endometrial functional layer and the basal layer, normalizes the coordinates of the boundary point set, establishes the spatial contour relationship between the functional layer and the basal layer, obtains the contour change features between adjacent frames, and generates functional layer-basal layer boundary data. The thickness change rate calculation submodule calculates the thickness of the functional layer and the base layer based on the boundary data of the functional layer and the base layer, extracts the thickness data of adjacent frames for difference calculation, and obtains the thickness change rate using the formula: ； Calculate the anomaly of the rate of change of thickness By combining the cyclical thickness variation range of healthy status, areas exceeding the baseline threshold are screened to obtain functional endometrial lesions, where represents . Representing the Functional layer thickness of frame image This represents the thickness of the functional layer in the previous frame. Represents the number of image frames. This represents the average thickness of all frames.

4. The AI-based endometriosis management system according to claim 1, characterized in that, The inflammation level assessment module includes: The lesion area extraction submodule extracts ultrasound image data of the lesion area based on the endometrial functional lesion area, identifies the lesion boundary contour, establishes a set of lesion boundary point coordinates, calculates the curvature change of the lesion boundary, removes abnormal boundary points, and obtains lesion area boundary data. The inflammatory marker concentration calculation submodule, based on the lesion region boundary data, collects the concentration data of local inflammatory markers IL-6 and TNF-α, stratifies the lesion region, groups the concentration data according to spatial location, and uses the following formula: ； Calculate the spatial gradient distribution values ​​of IL-6 and TNF-α. The inflammation distribution gradient data were obtained, where, Representing the Concentration of inflammatory markers in the layer This represents the concentration of inflammatory markers in the previous layer. Representing the The coordinates of the layer within the lesion region. , Representing the The coordinates of the layer within the lesion region. Indicates the number of floors; The inflammation concentration change screening submodule calculates the rate of change of inflammation concentration at different levels based on the inflammation distribution gradient data, filters out areas where the inflammation concentration change exceeds a set threshold, and obtains areas with abnormal inflammation levels.

5. The AI-based endometriosis management system according to claim 1, characterized in that, The lesion spread analysis module includes: The lesion edge extraction submodule extracts the pixel distribution of the lesion edge based on the abnormal inflammation level region, performs edge detection on the ultrasound image of the abnormal inflammation level region, selects the region where the pixel gradient change is greater than a set threshold as the initial point set of the lesion edge, calculates the curvature change for the extracted edge point set, removes discrete points with abnormal curvature, fits the edge, calculates the lesion edge change rate, and obtains the lesion edge pixel data. The inflammation spread rate calculation submodule calculates the difference in inflammation level between adjacent pixels based on the lesion edge pixel data, and combines this with time series data using the following formula: ； Calculate the rate of change of pixels at the edge of the lesion Data on the rate of inflammation spread were obtained, among which, Representing the The inflammation level value per pixel, Representing the -1 pixel's inflammation level value, This represents the distance from the pixel to the center of the lesion. Represents the spatial attenuation coefficient. This represents the time interval corresponding to a pixel. Represents the number of pixels; The lesion spread screening submodule, based on the inflammation spread rate data, compares the lesion edge spread rate with the inflammation level gradient, screens areas with high inflammation level and low spread rate, and obtains early lesion spread areas.

6. The AI-based endometriosis management system according to claim 1, characterized in that, The periodic lesion tracking module includes: The lesion contour change extraction submodule acquires multi-cycle endometrial image data based on the early lesion spread area, aligns the ultrasound images of each cycle, matches the lesion area, performs pixel matching on the lesion contour between image frames, calculates the contour displacement, removes abnormal contour changes caused by image noise, corrects the lesion edge through boundary point curvature analysis, and obtains lesion contour change data. The lesion growth rate calculation submodule calculates the pixel area of ​​the lesion region for each cycle based on the lesion contour change data, and compares the lesion areas of adjacent cycles using the following formula: ； Calculate the lesion growth rate per unit time. Data on lesion growth rate were obtained, among which, Represents the growth rate of the lesion. Representing the The area of ​​lesions in the cycle, Representing the The area of ​​lesions in the cycle, This represents the time interval corresponding to the cycle. The thickness of the functional layer representing the cycle. Represents the thickness influencing factor. Indicates the number of cycles; The lesion influence area determination submodule, based on the lesion growth rate data, calls the dynamic data of functional layer thickness and the dynamic data of basal layer thickness, compares the lesion growth area with the thickness change area, calculates the correlation between the lesion growth area and the thickness change, filters the area where the lesion growth and thickness change are consistent, and obtains the analysis results of the periodic lesion change trend.

7. The AI-based endometriosis management system according to claim 1, characterized in that, The disease progression analysis module includes: The current period inflammation spread extraction submodule obtains the inflammation spread path of the current period based on the analysis results of the periodic lesion change trend, extracts the lesion area in the ultrasound image data, measures the spatial distribution of the concentration of inflammatory markers of pixels around the lesion, calculates the concentration gradient and establishes the inflammation spread direction vector, selects pixels with the same gradient direction to form the inflammation spread path, and obtains the inflammation spread path data of the current period. The inflammation spread path comparison submodule calculates the difference in inflammation spread paths from previous periods based on the current period's inflammation spread path data. For the inflammation spread vector fields of the preceding and following periods, the following formula is used: ； Calculate the degree of change in the inflammatory spread pathway during the differential period. Data on differences in the path of inflammation spread were obtained, among which, For the first The pixel coordinates of the inflammation spread path during the period. For the first The pixel coordinates of the inflammation spread path during the period. For the first Inflammation concentration value at the point For the first Inflammation concentration value at the point This represents the total number of pixels along the path. The lesion future trend prediction submodule calculates the potential inflammation areas in the next cycle based on the inflammation spread path difference data. Based on the inflammation spread trend of multiple cycles, it predicts the future lesion spread areas, selects the areas with the most stable spread trends as high-risk lesion spread areas, and obtains the lesion future trend analysis results.

Images