A multi-factor risk assessment method for ovarian hyperstimulation syndrome

Abstract

The application discloses a kind of multi-factor risk assessment methods for ovarian hyperstimulation, it is related to intelligent medical technical field, including, based on multi-source baseline data package, the trend of follicle volume change with time is calculated by ovarian response simulation, obtain follicle response curve, the probability of vascular injury is calculated by puncture path hemorrhage simulation, obtain bleeding probability, and the follicle response curve and bleeding probability are fused, form patient characteristic risk vector;According to the intervention suggestion in the structured strategy list is associated with ovarian hyperstimulation risk score and bleeding risk score to generate risk change description, and form risk assessment report according to risk change description;The application constructs the patient characteristic risk vector by fusing follicle dynamic response and puncture bleeding risk, effectively enhance the safety and individualization level of assisted reproductive treatment.

Description

Technical Field

[0001] This invention relates to the field of intelligent medical technology, and in particular to a multifactor risk assessment method for ovarian hyperstimulation syndrome. Background Technology

[0002] Ovarian hyperstimulation syndrome (OHSS) is a common iatrogenic complication during ovulation induction treatment and has long been a focus of clinical attention. With the popularization and optimization of in vitro fertilization-embryo transfer (IVF-ET) technology, individualized ovulation induction protocols have gradually become an important strategy to reduce the incidence of OHSS. Existing risk assessment methods are mostly based on single or limited-dimensional clinical indicators, such as age, body mass index (BMI), basal antral follicle count (AFC), and anti-Müllerian hormone (AMH) levels, combined with previous OHSS history for empirical judgment. Some studies have attempted to introduce machine learning models to integrate multi-source clinical data to improve predictive performance.

[0003] Existing technologies have improved the sensitivity of OHSS risk identification to some extent. However, when integrating individual patient differences (such as the influence of gene polymorphism on drug metabolism and pharmacokinetics) with operation-related variables (such as the potential damage to vascular structures caused by the transvaginal oocyte retrieval route), existing methods are difficult to achieve dynamic quantification and interactive analysis of risk factors. Traditional scoring systems usually treat bleeding risk and OHSS risk separately, ignoring the possible synergistic effect between the two in the ovulation induction cycle, which in turn affects the overall coordination of intervention strategies. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a multifactor risk assessment method for ovarian hyperstimulation syndrome to address the problem of insufficient accuracy in comprehensive risk assessment caused by the lack of joint simulation of dynamic follicular response and bleeding risk during puncture.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a multifactor risk assessment method for ovarian hyperstimulation syndrome (OHSS), comprising: collecting the patient's past medical history, baseline hormone levels, ovarian reserve parameters, gene polymorphism information, and transvaginal ultrasound imaging data, and performing standardized processing to obtain a multi-source baseline data package; based on the multi-source baseline data package, calculating the trend of follicular volume change over time through ovarian response simulation to obtain a follicular response curve, calculating the probability of vascular injury through puncture path bleeding simulation to obtain a bleeding probability, and fusing the follicular response curve and bleeding probability to form a patient-characterized risk vector; fusing the patient-characterized risk vector with the multi-source baseline data package in a multimodal manner to obtain a comprehensive risk score and confidence level, and forming a multifactor comprehensive risk table; and judging the risk based on the confidence level of the multifactor comprehensive risk table. To assess the stability of the comprehensive risk score, simulation parameters for ovarian response simulation and puncture path bleeding simulation were adjusted to generate a corrected risk vector. This corrected risk vector was then recalculated to output the ovarian hyperstimulation risk score and bleeding risk score. The ovarian hyperstimulation risk score and bleeding risk score were compared with risk grading thresholds to obtain risk intervals. A two-dimensional risk matrix was established through multi-objective hierarchical calculation. This matrix was used to calculate recommendations for adjusting ovulation induction dosage, triggering method selection, puncture path adjustment, and post-ovarian retrieval monitoring, which were then integrated into a strategy list. Based on the intervention recommendations in the structured strategy list, the ovarian hyperstimulation risk score and bleeding risk score were correlated to generate a risk change description, and a risk assessment report was generated based on this description.

[0008] As a preferred embodiment of the multifactor risk assessment method for ovarian hyperstimulation syndrome described in this invention, the specific steps for obtaining the multi-source baseline data package are as follows:

[0009] Collect patients' past medical history data and extract medication history to obtain medical history information; collect basal hormone level data and extract follicle-stimulating hormone, luteinizing hormone and estradiol indicators, and calculate the average value and volatility to generate hormone characteristics; collect ovarian reserve parameters and extract antral follicle count and ovarian volume to generate ovarian reserve characteristics.

[0010] Gene polymorphism information was collected and gene loci related to follicle development and vascular permeability were extracted and sequenced to generate gene features;

[0011] Transvaginal ultrasound imaging data was acquired and follicular boundaries, follicular diameter, follicular number, and ovarian blood flow velocity were extracted to generate imaging features. Medical history information, hormonal characteristics, ovarian reserve characteristics, genetic characteristics, and imaging features were normalized and aligned with a time reference to generate a multi-source baseline data package.

[0012] As a preferred embodiment of the multifactorial risk assessment method for ovarian hyperstimulation syndrome described in this invention, the specific steps for obtaining the follicular response curve are as follows:

[0013] Time point extraction, time base mapping, and missing feature interpolation completion were performed on multi-source baseline data packets to obtain time series data related to follicular growth.

[0014] Based on time series data related to follicular growth, we analyzed hormone changes and combined them with the time changes in follicle number and average follicle diameter to obtain the temporal coupling relationship between hormone levels and follicular growth.

[0015] By combining the temporal coupling relationship and the patient's ovarian sensitivity parameters, the change rate of follicle number and the change rate of average follicle diameter were simultaneously converted to form an expression for the change rate of follicle volume, and the calculation relationship of follicle volume change over time was obtained.

[0016] Based on the calculation relationship of follicle volume change over time, a time step recursive algorithm is used to calculate the follicle volume change sequence, and interpolation and smoothing are performed to form a continuous time curve, thus obtaining the follicle response curve.

[0017] As a preferred embodiment of the multifactor risk assessment method for ovarian hyperstimulation syndrome described in this invention, the specific steps for forming a patient-characteristic risk vector are as follows:

[0018] The three-dimensional spatial coordinates of the ovary, the direction of adjacent blood vessels, the spatial distribution of follicles, the feasible puncture angle range and depth are extracted from the multi-source baseline data package and unified into the anatomical coordinate system to obtain the puncture simulation input set.

[0019] Based on the puncture simulation input set, the spatial representation of ovarian blood vessels is reconstructed and the ovarian blood vessel space volume is generated. Random collision detection is performed in the ovarian blood vessel space volume according to the puncture angle and depth to obtain the probability of blood vessel damage.

[0020] Physiological corrections are applied to the probability of vascular injury to obtain the probability of bleeding. The probability of bleeding is then aligned with the follicular response curve on the oocyte retrieval time axis, and the slope and curvature of the follicular response curve are extracted. By regularizing the combination of the aligned probability of bleeding and the features of the follicular response curve, a patient-characterized risk vector is obtained.

[0021] As a preferred embodiment of the multifactor risk assessment method for ovarian hyperstimulation syndrome described in this invention, the specific steps for forming the multifactor comprehensive risk table are as follows:

[0022] Hormone time series, medical record text features, and gene features are extracted from multi-source baseline data packages and aligned with patient-characterized risk vectors on the same time reference to obtain a joint fusion input set;

[0023] The joint fusion input set is numerically normalized, image encoded and compressed, text semantic vectorized and gene locus vectorized, and the temporal difference, slope and volatility are calculated to obtain the temporal feature matrix;

[0024] Numerical synthesis of feature correlation between the time-series feature matrix and the patient-characterized risk vector is performed to obtain a comprehensive risk score and confidence level. The comprehensive risk score and confidence level are then integrated into a multi-factor comprehensive risk table.

[0025] As a preferred embodiment of the multifactor risk assessment method for ovarian hyperstimulation syndrome described in this invention, the specific steps for generating the modified risk vector are as follows:

[0026] Based on the comprehensive risk score and confidence level in the multi-factor comprehensive risk table, the score stability index is calculated. Based on the score stability index and the feature contribution degree in the multi-factor comprehensive risk table, the parameters in the ovarian response simulation and puncture path bleeding simulation are adjusted to obtain the adjusted simulation parameters.

[0027] The ovarian response simulation and puncture path bleeding simulation were re-executed using the adjusted simulation parameters. The corrected follicular response curve and corrected bleeding probability were obtained and fused to output the corrected risk vector.

[0028] As a preferred embodiment of the multifactor risk assessment method for ovarian hyperstimulation syndrome described in this invention, the specific steps for outputting the ovarian hyperstimulation risk score and bleeding risk score are as follows:

[0029] The corrected risk vector is aligned with the hormone time series, ovarian reserve features and image features in the multi-source baseline data package on the same time reference to obtain the fusion input set; the features in the fusion input set are weighted and summarized to obtain the updated initial value of the comprehensive risk score and uncertainty measure.

[0030] Based on the uncertainty measure, the corresponding confidence level is calculated and the initial value of the updated comprehensive risk score is subjected to stability correction. Based on the correspondence between the stability-corrected comprehensive risk score and the hormone time series, ovarian reserve characteristics and imaging characteristics characterizing the risk of ovarian hyperstimulation in the multi-source baseline data package, the ovarian hyperstimulation risk score is obtained.

[0031] The bleeding risk score is obtained based on the correspondence between the stability-corrected comprehensive risk score and the image features representing bleeding risk in the multi-source baseline data package.

[0032] As a preferred embodiment of the multifactor risk assessment method for ovarian hyperstimulation syndrome described in this invention, the specific steps for establishing a two-dimensional risk matrix through multi-objective hierarchical calculation are as follows:

[0033] The ovarian hyperstimulation risk score and bleeding risk score were compared with the risk grading threshold to obtain the ovarian hyperstimulation risk interval and bleeding risk interval respectively, and then combined into a risk interval.

[0034] Based on the risk intervals, corresponding stratification levels were obtained on the ovarian hyperstimulation risk dimension and the bleeding risk dimension, and then matched according to the coordinate method to obtain two-dimensional stratification coordinate points;

[0035] A two-dimensional risk matrix is ​​established in a two-dimensional coordinate plane with multiple risk level areas based on two-dimensional layered coordinate points for positioning.

[0036] As a preferred embodiment of the multifactorial risk assessment method for ovarian hyperstimulation syndrome described in this invention, the specific steps for integrating the strategy list are as follows:

[0037] Based on the risk level regions corresponding to the ovarian hyperstimulation risk level and the bleeding risk level in the two-dimensional risk matrix, the risk control parameter set is obtained;

[0038] Based on the risk control parameter group, the adjustment range of the stimulation dose, the selection result of the triggering method, the adjustment amount of the angle and depth of the puncture path, and the time interval and number of monitoring after oocyte retrieval are calculated to obtain the stimulation dose adjustment suggestions, triggering method selection suggestions, puncture path adjustment suggestions and oocyte retrieval monitoring plan, and arrange and structure them according to the risk control logic to output a strategy list.

[0039] As a preferred embodiment of the multifactor risk assessment method for ovarian hyperstimulation syndrome described in this invention, the steps for generating a risk assessment report based on risk change descriptions are as follows:

[0040] Based on the intervention recommendations in the structured strategy checklist, the risk scores for ovarian hyperstimulation and bleeding were matched with the oocyte retrieval timeline to obtain phased risk information.

[0041] By combining the phased risk information with the ovarian hyperstimulation risk score and bleeding risk score, a description of risk changes is obtained, and a risk assessment report is formed by combining the intervention logic in the structured strategy checklist.

[0042] The beneficial effects of this invention are as follows: by constructing a patient-characterized risk vector that integrates dynamic follicular response and puncture bleeding risk, and introducing a simulation parameter adaptive correction mechanism based on confidence feedback, physiological response and surgical operation risk are unified into the same quantitative framework, providing precise, interpretable and executable intervention strategies for ovulation induction protocol optimization, puncture path planning and postoperative monitoring, effectively enhancing the safety and personalization of assisted reproductive treatment. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. 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.

[0044] Figure 1 This is a flowchart of a multifactor risk assessment method for ovarian hyperstimulation syndrome.

[0045] Figure 2 A flowchart for generating a patient-specific risk vector.

[0046] Figure 3 A flowchart for generating a multi-factor comprehensive risk table.

[0047] Figure 4 A flowchart for outputting ovarian hyperstimulation risk score and bleeding risk score. Detailed Implementation

[0048] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0049] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0050] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0051] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a multifactor risk assessment method for ovarian hyperstimulation syndrome, comprising the following steps:

[0052] S1. Collect the patient's past medical history, baseline hormone levels, ovarian reserve parameters, gene polymorphism information, and transvaginal ultrasound imaging data, and perform standardized processing to obtain a multi-source baseline data package.

[0053] S1.1. Collect the patient's past medical history data and extract the medication history to obtain medical history information; collect basal hormone level data and extract follicle-stimulating hormone, luteinizing hormone and estradiol indicators, and calculate the average value and volatility to generate hormone characteristics; collect ovarian reserve parameters and extract antral follicle count and ovarian volume to generate ovarian reserve characteristics.

[0054] Specifically, the patient's past medical history data is collected, and medical records and medication records related to reproductive endocrinology are screened out from the patient's past medical history data. The drug name, medication time sequence and medication duration are separated from the medication records. The sorted medication history is included in the medical history information as a component of the medical history information.

[0055] After collecting baseline hormone level data, the baseline hormone level data were arranged in chronological order of detection. Follicle-stimulating hormone (FSH), luteinizing hormone (LH), and estradiol indicators were extracted from the baseline hormone level data item by item. The average value and volatility of FSH, LH, and estradiol indicators were calculated for each indicator during the observation period. The average value and volatility of FSH, LH, and estradiol indicators were then combined to form a hormone characteristic.

[0056] When collecting ovarian reserve parameters, antral follicle count and ovarian volume are extracted from the ovarian reserve parameters. The antral follicle count is obtained by counting the number of visible antral follicles in the ovary through transvaginal ultrasound imaging data. The ovarian volume is obtained by measuring the long diameter, wide diameter, and thick diameter of the ovary in transvaginal ultrasound imaging data. The antral follicle count and ovarian volume are sorted and statistically unified, and the sorted antral follicle count and ovarian volume are used as ovarian reserve characteristics.

[0057] S1.2. Collect gene polymorphism information and extract gene loci related to follicle development and vascular permeability for sequence encoding to generate gene features; collect transvaginal ultrasound image data and extract follicle boundaries, follicle diameter, follicle number and ovarian blood flow velocity to generate image features.

[0058] Specifically, gene polymorphism information is collected, organized, and gene loci related to follicle development and vascular permeability are screened from the gene polymorphism information. The gene loci related to follicle development and vascular permeability are arranged and sequence-encoded to obtain gene characteristics.

[0059] Transvaginal ultrasound imaging data was acquired, and the images were preprocessed to highlight the ovarian region. Follicular boundaries were identified and delineated in the transvaginal ultrasound imaging data, and follicular diameters were measured within the follicular boundary range. The number of follicles in the transvaginal ultrasound imaging data at the same time was counted, and ovarian blood flow velocity was calculated by combining the Doppler information in the transvaginal ultrasound imaging data. Follicular boundaries, follicular diameters, follicular numbers, and ovarian blood flow velocity were combined into imaging features.

[0060] S1.3. Normalize and align the medical history information, hormonal characteristics, ovarian reserve characteristics, genetic characteristics, and imaging characteristics with the time reference to generate a multi-source baseline data package.

[0061] Specifically, the format of consultation and medication times in medical history information is standardized, and all time fields are converted into consistent time stamps to form a medical history time series that can be used for time alignment. The hormonal characteristics of follicle-stimulating hormone (FSH), luteinizing hormone (LH), and estradiol (ESH) values, the ovarian reserve characteristics of antral follicle count values, the genetic characteristics of sequence-coded values, and the imaging characteristics of follicle diameter, follicle number, and ovarian blood flow velocity values ​​are linearly normalized. The timeline related to ovulation induction treatment is used as a unified time benchmark. The time stamps of medical history information, the detection time points of hormonal characteristics, the assessment time points of ovarian reserve characteristics, the detection time points of genetic characteristics, and the acquisition time points of imaging characteristics are all mapped to the unified time benchmark. When any time point lacks hormonal characteristic values, ovarian reserve characteristic values, genetic characteristic values, or imaging characteristic values, linear interpolation is used to estimate and complete the data between adjacent time points, and the data is integrated into a multi-source baseline data package. S2. Based on multi-source baseline data, the trend of follicle volume change over time is calculated through ovarian response simulation to obtain the follicle response curve. The probability of vascular injury is calculated through puncture path bleeding simulation to obtain the bleeding probability. The follicle response curve and the bleeding probability are then fused to form a patient-characterized risk vector.

[0062] S2.1. Extract time points, map time references, and interpolate missing features from multi-source baseline data packets to obtain time series data related to follicle growth.

[0063] Specifically, data related to ovarian function, such as follicle count, average follicle diameter, follicle-stimulating hormone (FSH), luteinizing hormone (LH), and estradiol concentration, are read point-by-point from multi-source baseline data packages. The follicle count and average follicle diameter data are arranged chronologically to form a time series of follicle morphology changes. The FSH, LH, and estradiol concentration data are arranged in the same chronological order to form a time series of hormone level changes. Based on similar indicators at adjacent time points, interpolation methods are used to estimate missing indicators to complete the continuous time series. The follicle morphology change time series and the hormone level change time series are mapped one-to-one according to a time baseline. The follicle count, average follicle diameter, FSH, LH, and estradiol concentration data corresponding to each time point are combined into a multidimensional feature vector for the same time point, generating time series data related to follicle growth.

[0064] S2.2. Based on time series data related to follicle growth, analyze hormone changes and combine them with the time changes in follicle number and average follicle diameter to obtain the temporal coupling relationship between hormone levels and follicle growth.

[0065] Specifically, the concentrations of follicle-stimulating hormone, luteinizing hormone, and estradiol at each time point are paired with the number of follicles and the average diameter of follicles at the same time point in chronological order. The changes in the concentrations of each hormone and the changes in the number of follicles and the average diameter at adjacent time points are calculated to form a hormone change sequence, a follicle number change sequence, and a follicle diameter change sequence.

[0066] Pearson correlation analysis and Spearman rank correlation analysis were used to calculate the correlation coefficients between hormone change sequences and follicle number change sequences, and between hormone change sequences and follicle diameter change sequences, respectively. At the same time, linear regression analysis was used to estimate the influence coefficients of hormone changes on follicle number and diameter changes. By combining the correlation coefficients and influence coefficients, the temporal correspondence between hormone level changes and follicle number and diameter changes was extracted, and the temporal coupling relationship between hormone levels and follicle growth was obtained.

[0067] S2.3. Combining the temporal coupling relationship and the patient's ovarian sensitivity parameters, the change rate of follicle number and the change rate of average follicle diameter are simultaneously converted to form an expression for the change rate of follicle volume, and the calculation relationship of follicle volume change over time is obtained.

[0068] Specifically, based on the influence coefficients of changes in follicle-stimulating hormone (FSH), luteinizing hormone (LH), and estradiol concentration on changes in follicle number and average follicle diameter in the temporal coupling relationship between hormone levels and follicle growth, the changes in FSH, LH, and estradiol concentration at each time point in the follicle growth-related time series data are substituted into the calculation of the rate of change in follicle number and average follicle diameter at each time point. The patient's ovarian sensitivity parameter is used as a multiplicative modulatory factor to affect the rate of change in follicle number and average follicle diameter, so that the rate of change in follicle number and average follicle diameter can reflect the intensity of the individual patient's ovary response to changes in hormone levels.

[0069] After obtaining the individualized rate of change in the number of follicles and the rate of change in the average diameter of follicles, based on the premise that follicles are approximately spherical, and according to the geometric conversion relationship between the volume and diameter of a sphere, the rate of change in the average diameter of follicles is converted into the rate of change in the volume of a single follicle. The rate of change in the volume of a single follicle is then derived by combining the rate of change in the number of follicles with the rate of change in the number of follicles. By differentiating the product of the number of follicles and the volume of a single follicle with respect to time, the expression for the rate of change in the total volume of follicles is obtained. The expression for the rate of change in the total volume of follicles is then presented as a calculation relationship of the change in follicle volume over time, with time as the independent variable and the total volume of follicles as the dependent variable.

[0070] It should be noted that the expression for the rate of change of follicular volume is:

[0071] ;

[0072] in, Indicates the total volume of follicles over time rate of change, Indicates time The number of follicles, Indicates the average volume of a single follicle over time. rate of change, Indicates the time of a single follicle The average volume, Indicates the number of follicles over time rate of change, This represents a time index variable.

[0073] ;

[0074] in, This indicates the patient's ovarian sensitivity parameters. This represents the coefficient indicating the influence of changes in follicle-stimulating hormone (FSH) concentration on changes in the number of follicles. This represents the coefficient indicating the effect of changes in luteinizing hormone (LH) concentration on changes in follicle count. This indicates the concentration of follicle-stimulating hormone (FSH) over time. rate of change Indicates luteinizing hormone concentration over time The rate of change of the change.

[0075] ;

[0076] in, Indicates time The average diameter of the follicles, This represents the geometric coefficients obtained by differentiating the volume with respect to the diameter. Indicates the average diameter of follicles over time The rate of change.

[0077] ;

[0078] in, This represents the influence coefficient of changes in follicle-stimulating hormone (FSH) concentration on changes in follicle diameter. This represents the influence coefficient of changes in luteinizing hormone (LH) concentration on changes in follicle diameter. This represents the coefficient indicating the effect of changes in estradiol concentration on changes in follicle diameter. Indicates estradiol concentration over time The rate of change.

[0079] S2.4. Based on the calculation relationship of follicle volume change over time, the time step recursive algorithm is used to calculate the follicle volume change sequence, and interpolation and smoothing are performed to form a continuous time curve, thus obtaining the follicle response curve.

[0080] Specifically, a discrete time step is selected on the time axis related to ovulation induction treatment. The total follicle volume corresponding to the start monitoring time is used as the initial total follicle volume value. The calculation relationship of follicle volume change over time is rewritten into a time step recursive form. In each time step, the follicle volume change is calculated using the total follicle volume value of the previous time step and the follicle number change rate and follicle volume change rate at the corresponding time point. The follicle volume change is added to the total follicle volume value of the previous time step to obtain the total follicle volume value of the current time step. This process is continuously recursively applied along the time axis to form a follicle volume change sequence covering the entire ovulation induction process.

[0081] After obtaining the follicle volume change sequence, a one-dimensional interpolation method is used to supplement the estimated total follicle volume at intermediate time points between adjacent time steps. A smoothing filter method is then used to smooth the interpolated total follicle volume change sequence over time, making the curve of total follicle volume change over time more numerically continuous and stable. The smoothed curve of total follicle volume change over time is used as a continuous expression of follicle volume change over time to obtain the follicle response curve.

[0082] S2.5. Extract the three-dimensional spatial coordinates of the ovary, the direction of adjacent blood vessels, the spatial distribution of follicles, the feasible puncture angle range and depth from the multi-source baseline data package, and unify them into the anatomical coordinate system to obtain the puncture simulation input set.

[0083] Specifically, based on the ovarian three-dimensional spatial coordinates, follicular boundary information, and ovarian blood flow velocity data contained in the multi-source baseline data package, the three-dimensional contour of the ovary is reconstructed and its spatial location is determined through a continuous sequence of ultrasound images. Based on the vascular signals and Doppler blood flow images recorded in the multi-source baseline data package, the orientation, diameter, and relative positional relationship with the ovarian capsule of the main blood vessels around the ovary (vascular structures with continuously visible blood flow signals, relatively large diameters, and potential risks of crossing with the ovarian puncture path) are identified and extracted. The center coordinates of each follicle are determined by combining the follicular boundary information, and the spatial distribution density of follicles within the ovarian parenchyma is calculated. The feasible puncture angle range is set according to the standard operating procedures of the transvaginal ultrasound probe, with an example value of 0 to 30 degrees offset from the midline of the ovarian longitudinal axis to both sides. The puncture depth is obtained based on the distance from the ovarian three-dimensional spatial coordinates to the vaginal wall. The ovarian three-dimensional spatial coordinates, the orientation of adjacent blood vessels, the spatial distribution of follicles, the puncture angle range, and the puncture depth data are uniformly registered to the standard pelvic anatomical coordinate system with the sacral promontory as the origin, forming a puncture simulation input set.

[0084] S2.6. Based on the puncture simulation input set, reconstruct the spatial representation of ovarian blood vessels and generate the ovarian blood vessel space volume. Perform random collision detection within the ovarian blood vessel space volume according to the puncture angle and depth to obtain the probability of blood vessel damage.

[0085] Specifically, based on the spatial data in the puncture simulation input set, the outer boundary of the ovary is constructed in the anatomical coordinate system with the three-dimensional spatial coordinates of the ovary as the center, and the center line of the blood vessel is represented by a continuous spatial curve in the anatomical coordinate system along the direction of the adjacent blood vessels. The center line of the blood vessel is expanded into a three-dimensional channel of blood vessel with a cross-sectional radius, forming a spatial representation of the ovarian blood vessel. The spatial representation of the ovarian blood vessel is then encapsulated into an ovarian blood vessel space volume in the anatomical coordinate system.

[0086] Within the ovarian vascular space, a large number of puncture angle and depth combinations are randomly generated using the Monte Carlo method based on the feasible puncture angle range and depth in the puncture simulation input set. The geometric collision detection method is used to determine whether each puncture path segment intersects with any three-dimensional channel of any blood vessel in the ovarian vascular space. The ratio of the number of intersecting puncture path segments to the total number of randomly generated puncture path segments is calculated to obtain the probability of vascular injury.

[0087] S2.7. Perform physiological correction on the probability of vascular injury to obtain the probability of bleeding. Align the probability of bleeding with the follicular response curve on the egg retrieval time axis and extract the slope and curvature of the follicular response curve. Regularly combine the aligned probability of bleeding with the characteristics of the follicular response curve.

[0088] Specifically, based on the probability of vascular injury, combined with the bleeding-related medical history recorded in the patient's medical history information, and gene loci related to vascular permeability and coagulation function in the gene characteristics, a physiological correction coefficient is calculated for the probability of vascular injury. The product of the probability of vascular injury and the physiological correction coefficient is taken as the bleeding probability. Taking the oocyte retrieval time axis as a reference, the bleeding probability is located to the time interval corresponding to the oocyte retrieval operation, and a one-to-one correspondence is established with the follicular response curve on the oocyte retrieval time axis. The slope and curvature of the follicular response curve at several representative time points on the oocyte retrieval time axis are calculated using numerical differentiation methods. The bleeding probability, the slope and curvature of the follicular response curve are combined in a fixed order to form a fusion feature set. The bleeding probability, the slope and curvature of the follicular response curve in the fusion feature set are weighted and summed. The weighted fusion result is used as a joint representation of the patient's ovarian hyperstimulation risk and bleeding risk during oocyte retrieval, and a patient-characterized risk vector is output.

[0089] S3. The patient-characterized risk vector is fused with multi-source baseline data in a multimodal manner to obtain a comprehensive risk score and confidence level, and a multi-factor comprehensive risk table is formed.

[0090] S3.1. Extract hormone time series, medical record text features and gene features from multi-source baseline data packages and align them with the patient-characterized risk vector on the same time reference to obtain a joint fusion input set.

[0091] Specifically, based on the time-series data of follicle-stimulating hormone (FSH), luteinizing hormone (LH), and estradiol indicators included in the multi-source baseline data package, the temporal variation information of the calculated average and volatility results is retained. Simultaneously, based on the text records included in the medical history information in the multi-source baseline data package, unstructured text such as drug use history and ovulation induction treatment history are transformed into TF-IDF feature vectors through natural language processing to obtain medical record text features. At the same time, gene locus encoding values ​​related to follicle development and vascular permeability are extracted from the gene features in the multi-source baseline data package. The hormone time series, medical record text features, and gene features, together with the key parameters of the follicle response curve and the bleeding probability value contained in the patient's characteristic risk vector, are uniformly aligned to a standard time axis with the first administration time of ovulation induction treatment as the zero point. The hormone time series is resampled to uniform time points with a 24-hour interval using linear interpolation. The medical record text features and gene features are expanded by timestamp matching to obtain a joint fusion input set.

[0092] S3.2. Perform numerical normalization, image encoding compression, text semantic vectorization, and gene locus vectorization on the joint fusion input set, and calculate the temporal difference, slope, and volatility to obtain the temporal feature matrix.

[0093] Specifically, hormone time-series features, image features, medical record text features, gene features, and patient-specific risk vectors are extracted from the joint fusion input set. Quantitative indicators are linearly normalized, and high-dimensional image features are encoded and compressed to obtain image-coded features. Medical record text features are converted into fixed-length semantic vectors, and gene features are expanded into continuous numerical vectors. With a unified time reference as the horizontal axis, the time difference, slope, and volatility of each time-series feature are calculated. All processed features at the same time point are concatenated in a fixed order to form a single-row feature vector and arranged in chronological order to form a time-series feature matrix.

[0094] S3.3. Perform numerical synthesis of the feature correlation between the time series feature matrix and the patient characteristic risk vector to obtain the comprehensive risk score and confidence level, and integrate the comprehensive risk score and confidence level into a multi-factor comprehensive risk table.

[0095] Specifically, hormonal changes, imaging changes, medical record semantic features, and gene locus features related to ovarian hyperstimulation are selected from the time-series feature matrix. The components reflecting the intensity of follicular response and the probability of bleeding in the patient's characteristic risk vector are used as weighting references. Weighting coefficients corresponding to the patient's characteristic risk vector are assigned to the numerical values ​​of each dimension of the time-series feature matrix. A single scalar comprehensive risk score is obtained by weighted summation of the time-series feature matrix column-wise. Based on the dispersion of the comprehensive risk-related components of the time-series feature matrix at different time slices and the stability of the patient's characteristic risk vector across each risk component, variance calculation and confidence intervals are used to estimate the stability of the comprehensive risk score over the entire time range. This stability is then converted into a confidence level reflecting the reliability of the comprehensive risk score. The stability of the comprehensive risk score is determined by comparing the fluctuation range of the comprehensive risk score across different time slices with the overall average level to obtain a stability index. This stability index serves as a dimensionless quantity characterizing whether the comprehensive risk score changes smoothly over time. After obtaining the stability index, it is converted to a confidence level by mapping it to a range of 0 to 1. For example, a smaller stability index corresponds to a confidence level close to 1, reflecting high stability of the comprehensive risk score over time; a larger stability index corresponds to a confidence level close to 0, reflecting significant fluctuations in the comprehensive risk score over time. The comprehensive risk score, confidence level, and the feature weights, time slice markers, and feature source information used in the weighting process are recorded and organized to generate a multi-factor comprehensive risk table.

[0096] S4. Based on the confidence level of the multifactor comprehensive risk table, determine the stability of the comprehensive risk score, adjust the simulation parameters of the ovarian response simulation and the puncture path bleeding simulation, generate a corrected risk vector, recalculate the corrected risk vector, and output the ovarian hyperstimulation risk score and bleeding risk score.

[0097] S4.1. Based on the comprehensive risk score and confidence level in the multi-factor comprehensive risk table, calculate the score stability index. Based on the score stability index and the feature contribution degree in the multi-factor comprehensive risk table, adjust the parameters in the ovarian response simulation and puncture path bleeding simulation to obtain the adjusted simulation parameters.

[0098] Specifically, the comprehensive risk score and confidence level are retrieved from the multi-factor comprehensive risk table. A weighted calculation method is used, with the numerical values ​​of the comprehensive risk score and confidence level as inputs, to calculate the score stability index. Feature contribution information is also retrieved from the multi-factor comprehensive risk table. The feature contribution is the proportion of feature influence obtained by normalizing the feature weights recorded during the multimodal fusion process, used to characterize the relative influence strength of each feature on the comprehensive risk score. The feature contribution levels related to follicle-stimulating hormone (FSH), luteinizing hormone (LH), estradiol (estradiol), and follicular response are mapped to the patient's ovarian sensitivity parameters and follicular volume changes in the ovarian response simulation. The parameters are mapped to the feature contribution values ​​related to the three-dimensional spatial coordinates of the ovary, the direction of adjacent blood vessels, and the distribution of puncture angles, and then to the simulation parameters such as the puncture angle range, puncture depth, and blood vessel radius in the puncture path bleeding simulation. The stability level of the comprehensive risk score is determined based on the value of the scoring stability index. If the stability level is insufficient, the patient's ovarian sensitivity parameters, follicle volume change parameters, puncture angle range, puncture depth, and blood vessel radius are proportionally adjusted. The adjusted patient's ovarian sensitivity parameters, follicle volume change parameters, puncture angle range, puncture depth, and blood vessel radius are then recombined to obtain the adjusted simulation parameters.

[0099] S4.2. Re-execute the ovarian response simulation and puncture path bleeding simulation using the adjusted simulation parameters, obtain the corrected follicular response curve and corrected bleeding probability, fuse them, and output the corrected risk vector.

[0100] Specifically, the adjusted ovarian sensitivity parameters and follicular volume change parameters are substituted into the calculation relationship of follicular volume change over time. The time step recursive algorithm is used to recalculate the follicular volume change sequence at each time step. The new follicular volume change sequence is then interpolated and smoothed to obtain the corrected follicular response curve that reflects the follicular growth process after parameter adjustment.

[0101] The adjusted simulation parameters, such as the puncture angle range, puncture depth, and vessel radius, are substituted into the calculation step of the puncture trajectory and vessel spatial relationship in the puncture path bleeding simulation. Under the updated puncture angle range and puncture depth conditions, random collision detection is repeatedly performed to count the intersection of the puncture trajectory and the vessel space volume and calculate the new vessel damage probability. Combined with the bleeding risk-related indicators in the patient's medical history and genetic characteristics, the corrected bleeding probability is calculated. After obtaining the corrected follicular response curve and the corrected bleeding probability, the slope and curvature features of the corrected follicular response curve are extracted along the oocyte retrieval time axis. The corrected follicular response curve features and the corrected bleeding probability are recombined into a new fusion feature set. Numerical summarization is performed on the fusion feature set to output a corrected risk vector that represents the combined characteristics of the patient's ovarian hyperstimulation risk and bleeding risk under the current simulation parameters.

[0102] S4.3. Align the corrected risk vector with the hormone time series, ovarian reserve features and image features in the multi-source baseline data package on the same time reference to obtain the fusion input set; perform weighted summation of the features in the fusion input set to obtain the updated initial value of the comprehensive risk score and uncertainty measure.

[0103] Specifically, the risk-related time markers included in the modified risk vector are uniformly converted to the detection time points of hormone time series, assessment time points of ovarian reserve characteristics, and acquisition time points of imaging features in the multi-source baseline data package, and all are projected onto the time reference related to ovulation induction treatment. Based on the principle of time series positional consistency, the risk values ​​corresponding to the modified risk vector, hormone time series values, ovarian reserve characteristic values, and follicle number, average follicle diameter, and ovarian blood flow velocity values ​​in the imaging features are combined according to time points to construct a fusion input set. In the fusion input set, each type of feature is weighted and summarized, and the comprehensive expression level of the fusion feature is calculated using a weighted average method. This comprehensive expression level is used as the initial value for updating the comprehensive risk score. The feature fluctuations in the fusion input set are statistically analyzed, and variance calculation is used to measure the magnitude of time-varying changes, resulting in an uncertainty measure.

[0104] S4.4. Based on the uncertainty measure, calculate the corresponding confidence level and perform stability correction on the updated initial value of the comprehensive risk score. Based on the correspondence between the stability-corrected comprehensive risk score and the hormone time series, ovarian reserve characteristics and imaging characteristics in the multi-source baseline data package that characterize the risk of ovarian hyperstimulation, obtain the ovarian hyperstimulation risk score.

[0105] Specifically, the uncertainty measure is read from the updated initial value of the comprehensive risk score and the uncertainty measure. The uncertainty measure is used as an indicator to measure the degree of fluctuation of the updated initial value of the comprehensive risk score over time. The confidence level is obtained by mapping the uncertainty measure to a value range of 0 to 1. When the confidence level is low, the stability correction is performed on the updated initial value of the comprehensive risk score. The stability-corrected comprehensive risk score is obtained by smoothing the updated initial value of the comprehensive risk score in the time dimension or by weighting the updated initial value of the comprehensive risk score with the historical average comprehensive risk score.

[0106] The stability-corrected comprehensive risk score is compared with the hormone time series, ovarian reserve characteristics, and imaging characteristics in the multi-source baseline data package that characterize the risk of ovarian hyperstimulation. Using a segmented mapping method or regression analysis method, the risk level of the stability-corrected comprehensive risk score is converted according to the rate of increase of estradiol in the hormone time series, the antral follicle count level in the ovarian reserve characteristics, and the follicle volume growth in the imaging characteristics, and the ovarian hyperstimulation risk score is output.

[0107] S4.5. Based on the correspondence between the stability-corrected comprehensive risk score and the image features representing bleeding risk in the multi-source baseline data package, a bleeding risk score is obtained.

[0108] Specifically, after obtaining the stability-corrected comprehensive risk score, imaging features representing bleeding risk are selected from the multi-source baseline data package. These features include ovarian blood flow velocity, ovarian three-dimensional spatial coordinates, adjacent vessel orientation, adjacent vessel diameter, and the shortest distance between the ovarian surface and the planned puncture path. The imaging features representing bleeding risk are numerically normalized to eliminate dimensional differences. The normalized imaging features representing bleeding risk are combined with the stability-corrected comprehensive risk score. The stability-corrected comprehensive risk score is used as a reference for overall risk intensity. Increased ovarian blood flow velocity, larger adjacent vessel diameter, and densely packed vessels near the puncture path are considered factors that increase bleeding risk. Conversely, lower ovarian blood flow velocity and larger distance between the ovarian surface and major vessels are considered factors that decrease bleeding risk. A segmented scoring method is used to weight and correct the imaging features representing bleeding risk based on the stability-corrected comprehensive risk score. A numerical result reflecting the patient's overall bleeding risk level under oocyte retrieval-related procedures is obtained through continuous approximation and output as the bleeding risk score.

[0109] S5. Based on the comparison of ovarian hyperstimulation risk score and bleeding risk score with risk grading threshold, the risk range is obtained. A two-dimensional risk matrix is ​​established through multi-objective hierarchical calculation. The two-dimensional risk matrix is ​​used to calculate the suggestions for adjusting ovulation induction dose, triggering method selection, puncture path adjustment, and post-egg retrieval monitoring plan, and integrate them into a strategy list.

[0110] S5.1. Compare the ovarian hyperstimulation risk score and bleeding risk score with the risk grading threshold to obtain the ovarian hyperstimulation risk interval and bleeding risk interval respectively, and combine them into a risk interval.

[0111] Specifically, the ovarian hyperstimulation risk score is compared with the threshold used for ovarian hyperstimulation risk grading. By determining the segment position of the ovarian hyperstimulation risk score within the risk grading threshold, the corresponding ovarian hyperstimulation risk interval is determined. The bleeding risk score is compared with the risk grading threshold used for bleeding risk grading. Based on the segment position of the bleeding risk score, the bleeding risk interval is determined. After obtaining the ovarian hyperstimulation risk interval and the bleeding risk interval, the ovarian hyperstimulation risk interval and the bleeding risk interval are combined to form a risk interval by constructing an interval combination relationship.

[0112] It should be noted that clinical outcomes related to ovarian hyperstimulation (e.g., occurrence of ovarian hyperstimulation syndrome) and bleeding-related clinical outcomes (e.g., abnormally increased vaginal bleeding after oocyte retrieval) from historical cases were collected and labeled separately. The sensitivity and specificity of the ovarian hyperstimulation risk score and bleeding risk score were calculated using receiver operating characteristic (ROC) curve analysis, and the cutoff points for distinguishing different risk levels were determined in the ROC curve. The risk score distribution was segmented using hierarchical clustering to obtain multiple interval boundaries that can reflect the differences in risk levels. These were then compared and integrated to form the risk grading threshold.

[0113] S5.2. Based on the risk intervals, obtain the corresponding stratification levels on the ovarian hyperstimulation risk dimension and the bleeding risk dimension respectively, and match them in a coordinate manner to obtain two-dimensional stratification coordinate points; use the two-dimensional stratification coordinate points as the positioning basis to establish a two-dimensional risk matrix of multiple risk level regions on the two-dimensional coordinate plane.

[0114] Specifically, the level coordinate position in the ovarian hyperstimulation risk dimension is determined based on the stratification level corresponding to the ovarian hyperstimulation risk interval.

[0115] The stratified level coordinates in the ovarian hyperstimulation risk dimension and the stratified level coordinates in the bleeding risk dimension are combined in a coordinate axis manner to form two-dimensional stratified coordinate points. Based on the two-dimensional stratified coordinate points, multiple risk level regions are divided in the two-dimensional coordinate plane. By setting the stratified level scale of the ovarian hyperstimulation risk dimension on the horizontal axis and the stratified level scale of the bleeding risk dimension on the vertical axis, different level combinations are mapped to different rectangular or gridded regions to form a two-dimensional risk matrix.

[0116] S5.3. Based on the risk level regions corresponding to the ovarian hyperstimulation risk level and the bleeding risk level in the two-dimensional risk matrix, the risk control parameter group is obtained.

[0117] Specifically, the risk level region where the two-dimensional stratified coordinate point is located is identified based on its position in the two-dimensional risk matrix. According to the clinical management rules corresponding to the risk level region, the ovulation induction drug adjustment parameters, monitoring frequency adjustment parameters, and oocyte retrieval strategy adjustment parameters are extracted sequentially from the control parameter items associated with the risk level region. By clarifying the correspondence between each type of control parameter and the risk level of ovarian hyperstimulation and bleeding, for example, in the region with a combination of high ovarian hyperstimulation risk and bleeding risk, the risk control parameter group may include parameters for reducing the dosage of ovulation induction drugs, parameters for shortening the monitoring time interval, and operation prompts for adjusting the puncture path during oocyte retrieval. The ovulation induction drug adjustment parameters, monitoring frequency adjustment parameters, and oocyte retrieval strategy adjustment parameters are combined in a fixed order to obtain the risk control parameter group.

[0118] It should be noted that clinical management rules can be illustrated using existing reproductive endocrine management pathways. For example, in a two-dimensional risk level region with both high ovarian hyperstimulation risk and high bleeding risk, the risk control parameter set can include parameters for reducing the dosage of ovulation-inducing drugs (e.g., a 20% reduction in the example value), parameters for shortening the monitoring interval (e.g., once every 24 hours in the example value), and parameters for choosing a more lateral approach during puncture (e.g., changing the puncture angle by 5°–10° in the example value). In a two-dimensional risk level region with both high ovarian hyperstimulation risk and low bleeding risk, the risk control parameter set can include parameters for reducing the dosage of ovulation-inducing drugs (e.g., a 10%–15% reduction in the example value), parameters for the frequency of routine monitoring, and parameters for choosing a triggering method that favors the use of GnRH agonists. In a two-dimensional risk level region with both low ovarian hyperstimulation risk and high bleeding risk, the risk control parameter set can include parameters for maintaining the dosage of ovulation-inducing drugs unchanged, parameters for shortening the monitoring interval, and parameters for avoiding high-flow-rate blood vessel areas during oocyte retrieval.

[0119] S5.4. Based on the risk control parameter group, calculate the adjustment range of the stimulation dose, the selection result of the triggering method, the adjustment amount of the angle and depth of the puncture path, and the time interval and number of monitoring after oocyte retrieval. Obtain the stimulation dose adjustment suggestions, triggering method selection suggestions, puncture path adjustment suggestions and oocyte retrieval monitoring plan, and arrange and structure them according to the risk control logic to output the strategy list.

[0120] Specifically, based on the ovulation induction medication adjustment parameters, triggering method control parameters, puncture path adjustment parameters, and post-egg retrieval monitoring adjustment parameters included in the risk control parameter group, the adjustment range of ovulation induction dosage, the selection result of triggering method, the adjustment amount of puncture path angle and depth, and the time interval and number of post-egg retrieval monitoring are calculated respectively.

[0121] Furthermore, the ovulation induction medication adjustment parameters are used to determine recommendations for adjusting the ovulation induction dose. For example, the ovulation induction medication adjustment parameters correspond to a reduction in the ovulation induction dose or a slight adjustment by a fixed range. The triggering method control parameters are used to determine recommendations for triggering method selection. For example, the triggering method control parameters should select gonadotropin-releasing hormone agonist triggering or human chorionic gonadotropin triggering. The puncture path adjustment parameters are used to determine recommendations for puncture path adjustment. By calculating the changes in the angle and depth of the puncture path, the puncture path maintains a low probability of vascular damage in the ovarian vascular space representation. The post-ovulation retrieval monitoring adjustment parameters are used to determine the post-ovulation retrieval monitoring plan. According to the risk mitigation logic implicit in the risk control parameter group, the recommendations for ovulation induction dose adjustment, triggering method selection, puncture path adjustment, and post-ovulation retrieval monitoring plan are arranged in a fixed order to form a strategy list.

[0122] S6. Based on the intervention recommendations in the structured strategy list, correlate them with the ovarian hyperstimulation risk score and bleeding risk score to generate a risk change description, and then generate a risk assessment report based on the risk change description.

[0123] S6.1. Based on the intervention recommendations in the structured strategy checklist, match them with the ovarian hyperstimulation risk score and bleeding risk score on the egg retrieval timeline to obtain phased risk information.

[0124] Specifically, the calibrated intervention recommendations were matched one-to-one with the ovarian hyperstimulation risk score and bleeding risk score on the same oocyte retrieval timeline. By determining the risk level of the ovarian hyperstimulation risk score and bleeding risk score at each time point, the time points corresponding to the ovulation induction dose adjustment recommendations were compared with the overlapping intervals of risk fluctuations. The execution time points of the triggering method selection recommendations were matched with the changes in the ovarian hyperstimulation risk score before and after triggering. The puncture operation time points corresponding to the puncture path adjustment recommendations were matched point-by-point with the bleeding risk score curve. The monitoring time points of the post-ovarian retrieval monitoring protocol were matched with the changes in the ovarian hyperstimulation risk score and bleeding risk score after oocyte retrieval. The intervention recommendations in the structured strategy list were arranged together with the ovarian hyperstimulation risk score and bleeding risk score on the oocyte retrieval timeline, and staged risk information was generated based on the risk level at different stages.

[0125] S6.2. Combine the phased risk information with the ovarian hyperstimulation risk score and bleeding risk score to obtain a description of risk changes, and combine it with the intervention logic in the structured strategy checklist to form a risk assessment report.

[0126] Specifically, the phased risk information is synchronously integrated with the ovarian hyperstimulation risk score and bleeding risk score on the oocyte retrieval timeline. The direction and magnitude of change in the ovarian hyperstimulation risk score and bleeding risk score are calculated during the stimulation, triggering, oocyte retrieval, and post-retrieval monitoring phases to obtain the risk change trend at each stage. Based on the risk change trend, risk-increasing, high-risk, and decreasing risk intervals are identified. The change characteristics of each interval are compared with the intervention execution time and intervention content recorded in the phased risk information. From the comparison results, a risk change description reflecting the impact of intervention measures on risk change is formed. After obtaining the risk change description, it is matched item by item with the intervention logic in the structured strategy checklist. Following the implementation sequence of medication strategy, triggering strategy, oocyte retrieval operation strategy, and post-retrieval monitoring strategy during the stimulation phase, the risk change trend and its corresponding intervention effect are sequentially explained to form a risk assessment report with the oocyte retrieval timeline as the main line.

[0127] In summary, this invention, by constructing a patient-characterized risk vector that integrates dynamic follicular response and puncture bleeding risk, and introducing a confidence-based adaptive correction mechanism for simulation parameters, unifies physiological response and surgical operation risk into the same quantitative framework. This provides precise, interpretable, and executable intervention strategies for optimizing ovulation induction protocols, planning puncture pathways, and postoperative monitoring, effectively enhancing the safety and personalization of assisted reproductive treatment.

[0128] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multifactor risk assessment method for ovarian hyperstimulation syndrome, characterized in that: include, The patient's past medical history, baseline hormone levels, ovarian reserve parameters, gene polymorphism information, and transvaginal ultrasound imaging data were collected and standardized to obtain a multi-source baseline data package. Based on multi-source baseline data, the trend of follicle volume change over time is calculated through ovarian response simulation to obtain the follicle response curve. The probability of vascular injury is calculated through puncture path bleeding simulation to obtain the bleeding probability. The follicle response curve and the bleeding probability are then fused to form a patient-characterized risk vector. The specific steps for forming a patient-specific risk vector are as follows: The three-dimensional spatial coordinates of the ovary, the direction of adjacent blood vessels, the spatial distribution of follicles, the feasible puncture angle range and depth are extracted from the multi-source baseline data package and unified into the anatomical coordinate system to obtain the puncture simulation input set. Based on the puncture simulation input set, the spatial representation of ovarian blood vessels is reconstructed and the ovarian blood vessel space volume is generated. Random collision detection is performed in the ovarian blood vessel space volume according to the puncture angle and depth to obtain the probability of blood vessel damage. Physiological correction is applied to the probability of vascular injury to obtain the probability of bleeding. The probability of bleeding is aligned with the follicular response curve on the oocyte retrieval time axis, and the slope and curvature of the follicular response curve are extracted. By regularizing the combination of the aligned probability of bleeding and the features of the follicular response curve, a patient-characterized risk vector is obtained. The patient-characterized risk vector is fused with multi-source baseline data in a multimodal manner to obtain a comprehensive risk score and confidence level, and a multi-factor comprehensive risk table is formed. The stability of the comprehensive risk score is determined based on the confidence level of the multifactor comprehensive risk table. The simulation parameters of the ovarian response simulation and the puncture path bleeding simulation are adjusted to generate a modified risk vector. The modified risk vector is then re-fused and calculated to output the ovarian hyperstimulation risk score and the bleeding risk score. Based on the comparison of ovarian hyperstimulation risk score and bleeding risk score with risk grading threshold, risk intervals were obtained. A two-dimensional risk matrix was established through multi-objective hierarchical calculation. The two-dimensional risk matrix was used to calculate suggestions for adjusting ovulation induction dose, selection of triggering method, adjustment of puncture path, and post-egg retrieval monitoring plan, and integrated into a strategy list. The specific steps for establishing a two-dimensional risk matrix through multi-objective hierarchical calculation are as follows: The ovarian hyperstimulation risk score and bleeding risk score were compared with the risk grading threshold to obtain the ovarian hyperstimulation risk interval and bleeding risk interval respectively, and then combined into a risk interval. Based on the risk intervals, corresponding stratification levels were obtained on the ovarian hyperstimulation risk dimension and the bleeding risk dimension, and then matched according to the coordinate method to obtain two-dimensional stratification coordinate points; A two-dimensional risk matrix is ​​established in a two-dimensional coordinate plane with multiple risk level areas based on two-dimensional hierarchical coordinate points for positioning. The risk change description is generated by correlating the intervention recommendations in the structured strategy list with the ovarian hyperstimulation risk score and bleeding risk score, and a risk assessment report is formed based on the risk change description.

2. The multifactor risk assessment method for ovarian hyperstimulation syndrome as described in claim 1, characterized in that: The specific steps for obtaining multi-source baseline data packets are as follows. Collect patients' past medical history data and extract medication history to obtain medical history information; collect basal hormone level data and extract follicle-stimulating hormone, luteinizing hormone and estradiol indicators, and calculate the average value and volatility to generate hormone characteristics; collect ovarian reserve parameters and extract antral follicle count and ovarian volume to generate ovarian reserve characteristics. Gene polymorphism information was collected and gene loci related to follicle development and vascular permeability were extracted and sequenced to generate gene features; Transvaginal ultrasound image data were acquired and follicular boundaries, follicular diameter, follicular number, and ovarian blood flow velocity were extracted to generate image features; The medical history information, hormonal characteristics, ovarian reserve characteristics, genetic characteristics and imaging characteristics are normalized and aligned with the time reference to generate a multi-source baseline data package.

3. The multifactor risk assessment method for ovarian hyperstimulation syndrome as described in claim 1, characterized in that: The specific steps for obtaining the follicle response curve are as follows. Time point extraction, time base mapping, and missing feature interpolation completion were performed on multi-source baseline data packets to obtain time series data related to follicular growth. Based on time series data related to follicular growth, we analyzed hormone changes and combined them with the time changes in follicle number and average follicle diameter to obtain the temporal coupling relationship between hormone levels and follicular growth. By combining the temporal coupling relationship and the patient's ovarian sensitivity parameters, the change rate of follicle number and the change rate of average follicle diameter were simultaneously converted to form an expression for the change rate of follicle volume, and the calculation relationship of follicle volume change over time was obtained. Based on the calculation relationship of follicle volume change over time, a time step recursive algorithm is used to calculate the follicle volume change sequence, and interpolation and smoothing are performed to form a continuous time curve, thus obtaining the follicle response curve.

4. The multifactor risk assessment method for ovarian hyperstimulation syndrome as described in claim 1, characterized in that: The specific steps for generating a multi-factor comprehensive risk table are as follows: Hormone time series, medical record text features, and gene features are extracted from multi-source baseline data packages and aligned with patient-characterized risk vectors on the same time reference to obtain a joint fusion input set; The joint fusion input set is numerically normalized, image encoded and compressed, text semantic vectorized and gene locus vectorized, and the temporal difference, slope and volatility are calculated to obtain the temporal feature matrix; Numerical synthesis of feature correlation between the time-series feature matrix and the patient-characterized risk vector is performed to obtain a comprehensive risk score and confidence level. The comprehensive risk score and confidence level are then integrated into a multi-factor comprehensive risk table.

5. The multifactor risk assessment method for ovarian hyperstimulation syndrome as described in claim 1, characterized in that: The specific steps for generating the corrected risk vector are as follows: Based on the comprehensive risk score and confidence level in the multi-factor comprehensive risk table, the score stability index is calculated. Based on the score stability index and the feature contribution degree in the multi-factor comprehensive risk table, the parameters in the ovarian response simulation and puncture path bleeding simulation are adjusted to obtain the adjusted simulation parameters. The ovarian response simulation and puncture path bleeding simulation were re-executed using the adjusted simulation parameters. The corrected follicular response curve and corrected bleeding probability were obtained and fused to output the corrected risk vector.

6. The multifactor risk assessment method for ovarian hyperstimulation syndrome as described in claim 1, characterized in that: The specific steps for generating the ovarian hyperstimulation risk score and bleeding risk score are as follows. The corrected risk vector is aligned with the hormone time series, ovarian reserve features and imaging features in the multi-source baseline data package on the same time reference to obtain the fusion input set; The features in the fused input set are weighted and summarized to obtain the updated initial value of the comprehensive risk score and the uncertainty measure; Based on the uncertainty measure, the corresponding confidence level is calculated and the initial value of the updated comprehensive risk score is subjected to stability correction. Based on the correspondence between the stability-corrected comprehensive risk score and the hormone time series, ovarian reserve characteristics and imaging characteristics characterizing the risk of ovarian hyperstimulation in the multi-source baseline data package, the ovarian hyperstimulation risk score is obtained. The bleeding risk score is obtained based on the correspondence between the stability-corrected comprehensive risk score and the image features representing bleeding risk in the multi-source baseline data package.

7. The multifactor risk assessment method for ovarian hyperstimulation syndrome as described in claim 1, characterized in that: The specific steps for integrating these into a strategy list are as follows. Based on the risk level regions corresponding to the ovarian hyperstimulation risk level and the bleeding risk level in the two-dimensional risk matrix, the risk control parameter set is obtained; Based on the risk control parameter group, the adjustment range of the stimulation dose, the selection result of the triggering method, the adjustment amount of the angle and depth of the puncture path, and the time interval and number of monitoring after oocyte retrieval are calculated to obtain the stimulation dose adjustment suggestions, triggering method selection suggestions, puncture path adjustment suggestions and oocyte retrieval monitoring plan, and arrange and structure them according to the risk control logic to output a strategy list.

8. The multifactor risk assessment method for ovarian hyperstimulation syndrome as described in claim 1, characterized in that: The risk assessment report is generated based on the description of risk changes. The specific steps are as follows: Based on the intervention recommendations in the structured strategy checklist, the risk scores for ovarian hyperstimulation and bleeding were matched with the oocyte retrieval timeline to obtain phased risk information. By combining the phased risk information with the ovarian hyperstimulation risk score and bleeding risk score, a description of risk changes is obtained, and a risk assessment report is formed by combining the intervention logic in the structured strategy checklist.

Images