A method and apparatus for clinical prediction of fertility potential in patients with uterine fibroids

Abstract

The present application relates to a kind of clinical prediction method and device for the fertility potential of uterine fibroid patient, wherein the method comprises: encoding anatomical integrated dataset to generate interpretable standardized feature matrix, and cross-validation is carried out by Elastic Net elastic network model, the minimum feature subset with significant and clinical relevance to composite reproductive outcome endpoint is screened out, and interpretable baseline prediction model is trained according to the minimum feature subset;The original clinical data of the individual to be predicted is input into interpretable baseline prediction model, and the preliminary individualized fertility potential probability value is output;Individualized fertility potential probability value is corrected, and time-varying calibrated uterine fibroid related fertility potential index is generated and mapped to clinical decision threshold interval to generate suggestion report.The present application carries out longitudinal tracking and individualized correction to fertility potential by dynamic calibration coefficient vector and time-varying UFP-DI index, significantly improves the accuracy and individualized level of uterine fibroid patient fertility potential evaluation.

Description

Technical Field

[0001] This invention relates to the field of gynecological assisted reproductive technology, specifically to a clinical prediction method and device, and computer equipment for the fertility potential of patients with uterine fibroids. Background Technology

[0002] Uterine fibroids are the most common benign pelvic tumors in women of reproductive age. Although most fibroids are benign, their location, size, and number can significantly interfere with the morphology of the uterine cavity, endometrial receptivity, fallopian tube patency, and ovarian microenvironment, thereby reducing the natural pregnancy rate and increasing the risk of adverse pregnancy outcomes such as miscarriage and premature birth. Clinically, it is necessary to assess the impact of fibroids on an individual's fertility potential.

[0003] Currently, clinical assessment mainly relies on the FIGO classification system. While the FIGO 0–8 classification can describe the anatomical location of fibroids, it lacks the ability to quantify their impact on fertility. Furthermore, existing guidelines (such as ESHRE and ACOG) primarily focus on whether to remove fibroids, without providing dynamic predictive tools for different fertility goals. In addition, traditional scoring systems (such as MUSA and STEPW) neglect the interaction of multiple dimensions of factors, including endocrine axis function, ovarian reserve, intrauterine microenvironment, and comorbidities, leading to biased assessments and poor reproducibility.

[0004] To address the aforementioned issues, this invention discloses a clinical prediction method for the fertility potential of patients with uterine fibroids, aiming to improve the accuracy and individualization of fertility potential assessment in these patients. Summary of the Invention

[0005] In view of the above problems, the present invention provides a clinical prediction method and device, and a computer device for the fertility potential of patients with uterine fibroids.

[0006] According to one aspect of the present invention, a clinical prediction method for the fertility potential of patients with uterine fibroids is provided, comprising:

[0007] A phenotypic-deeply annotated fertility cohort for uterine fibroids was constructed, and an integrated anatomical dataset was extracted from women of reproductive age with uterine fibroids who wished to conceive naturally or planned to undergo assisted reproductive technology. The integrated anatomical dataset included baseline population parameters, reproductive endocrine axis function indicators, fertility reserve quantitative parameters, reproductive history, imaging features of uterine fibroids, anatomical features of uterine fibroids, intrauterine microenvironment integrity index, perturbation degree of fallopian tube-ovarian anatomical relationship, and comorbidity burden index.

[0008] The anatomical integrated dataset was standardized and encoded according to FIGO classification and clinical guidelines to generate an interpretable standardized feature matrix. The standardized feature matrix was cross-validated using the Elastic Net model. Under L1 / L2 mixed regularization constraints, the minimum feature subset with preset significance and clinical relevance to the composite fertility outcome endpoint was selected. An interpretable baseline prediction model was trained based on the minimum feature subset.

[0009] The original clinical data of the individual to be predicted is input into the interpretable baseline prediction model, and a preliminary individualized fertility potential probability value is output. Based on the fibroid growth rate, ART protocol type and ovarian stimulation responsiveness information of the individual to be predicted, the individualized fertility potential probability value is corrected to generate a time-varying calibrated uterine fibroid-related fertility potential index.

[0010] The calibrated uterine fibroid-related fertility potential index values ​​are mapped to clinical decision threshold ranges and a recommendation report is generated. The recommendation report includes a heatmap of key risk drivers, a fibroid intervention necessity score, and personalized fertility pathway recommendations.

[0011] In one alternative approach, the baseline demographic parameters include age and body mass index; the reproductive endocrine axis function indicators include menstrual cycle stability indicators and ovulation function assessment indicators; the fertility reserve quantification parameters include anti-Müllerian hormone (AMH) levels and basal antral follicle count (AFC); the reproductive history includes past pregnancy history, delivery history, miscarriage history, and duration of infertility; the imaging characteristics of uterine fibroids include the number of lesions, maximum diameter, three-dimensional volume percentage, MRI enhancement pattern, and apparent diffusion coefficient; the anatomical characteristics of uterine fibroids include type 0-8 classification, number of fibroids, maximum fibroid diameter, and total fibroid volume percentage; the uterine cavity microenvironment integrity index includes the uterine cavity morphology distortion index and endometrial blood perfusion parameters; the tubo-ovarian anatomical relationship disturbance degree includes the tubal mechanical compression index and the spatial positional relationship of ovarian fibroids; and the comorbidity burden index includes the depth of adenomyosis invasion, endometriosis FAF staging, and adnexal lesions.

[0012] In one alternative approach, standardizing the integrated anatomical dataset according to FIGO classification and clinical guidelines to generate an interpretable standardized feature matrix further includes:

[0013] The continuous variables in the anatomical integrated dataset were subjected to a normality test. For continuous variables that did not meet the normality assumption, an adaptive Box-Cox power transformation was performed to obtain a continuous feature matrix. The continuous variables included age, body mass index (BMI), anti-Müllerian hormone (AMH) level, basal antral follicle count (AFC), three-dimensional volume ratio, apparent diffusion coefficient (ADC) value, uterine cavity morphology distortion index, endometrial blood perfusion parameters, fallopian tube mechanical compression index, and adenomyosis infiltration depth.

[0014] The categorical variables in the anatomical integrated dataset are input into the FIGO typing encoder. Hierarchical semantic mapping of uterine fibroid anatomical features is performed according to the International Federation of Gynecology and Obstetrics' (IFO) uterine fibroid typing standards to generate a fibroid typing one-hot encoding vector. A classification feature matrix is ​​generated based on the fibroid typing one-hot encoding vector, endometriosis FAF staging, and adnexal lesions. The categorical variables include menstrual cycle stability indicators, ovulation function assessment indicators, past pregnancy history, delivery history, miscarriage history, type 0-8 classification, MRI enhancement pattern, and the spatial location of ovarian fibroids.

[0015] The continuous feature matrix and the classification feature matrix are encoded in a temporal sequence to generate an interpretable standardized feature matrix.

[0016] In one alternative approach, the individualized fertility potential probability value is corrected based on the fibroid growth rate, ART protocol type, and ovarian stimulation responsiveness information of the individual to be predicted, and the generation of a time-varying calibrated uterine fibroid-related fertility potential index further includes:

[0017] The longitudinal imaging follow-up data of the individual to be predicted is obtained, and MRI or ultrasound images of uterine fibroids at at least two different time points are extracted and input into the fibroid segmentation model to calculate the three-dimensional volume of the fibroids at each time point; wherein, the time span of the longitudinal imaging follow-up data is at least 3 menstrual cycles;

[0018] Ovarian responsiveness is dynamically assessed based on the clinical data of the ART cycle of the individuals to be predicted, and the ovarian responsiveness index is calculated. The ART cycle clinical data includes the type of ovulation induction protocol, initiation dose, number of Gn days, total Gn dose, daily estrogen level on HCG day, daily progesterone level on HCG day, number of oocytes retrieved, MII oocyte rate, fertilization rate, and high-quality embryo rate. The ovarian responsiveness index includes the ratio of actual oocytes retrieved to expected oocytes retrieved and the follicle release rate.

[0019] The influence weight of fertility potential is estimated based on the three-dimensional volume of the fibroid and the ovarian responsiveness index to obtain a dynamic calibration coefficient vector; the individualized fertility potential probability value is used as the baseline predicted value and the dynamic calibration coefficient vector is modified by multiplication effect to generate a preliminary calibration value; the number of previous failed ART cycles is negatively weighted according to the ART cycle number decay factor to generate a cycle cumulative effect correction term.

[0020] The initial calibration value and the periodic cumulative effect correction term are dynamically synthesized to generate a time-varying calibrated uterine fibroid-related fertility potential index.

[0021] In one alternative approach, mapping the calibrated uterine fibroid-related fertility potential index value to a clinical decision threshold range and generating a recommendation report further includes:

[0022] The uterine fibroid-related fertility potential index is input into a multi-level threshold mapper, and the UFP-DI is mapped to the corresponding fertility potential level based on the dynamic threshold range determined by large-sample multi-center clinical validation.

[0023] The UFP-DI and its corresponding fertility potential level are analyzed using risk driver factors to obtain analytical features. The marginal contribution value of each analytical feature to UFP-DI is calculated based on the feature contribution weight vector of the individual to be predicted, and a feature attribution weight matrix is ​​generated. The feature attribution weight matrix is ​​sorted in descending order, and the top-K key risk driver factors with a cumulative contribution rate of more than 85% are selected.

[0024] The top-K key risk drivers and their corresponding attribution weights are input into a heatmap generator to generate a heatmap of key risk drivers. The heatmap displays the contribution direction and intensity of each risk factor in matrix form, and provides a network topology diagram of the interaction effects between factors.

[0025] The UFP-DI, fertility potential level, and heatmap of key risk drivers are input into a path planning decision tree generator to generate a phased fertility path planning decision tree with the goal of maximizing the cumulative pregnancy probability. The decision tree is rooted at the current UFP-DI and branches along the time axis to output personalized fertility path suggestions.

[0026] In one alternative approach, the formula for calculating the dynamic calibration coefficient vector is:

[0027]

[0028] in, A sigmoid mapping kernel with adjustable kurtosis. , The median threshold for a multi-center queue; It is an exponentially decaying memory kernel. , , The half-life of the effect of fibroid volume on pregnancy rate; For anisotropic diffusion smoothing function; The distribution function is used as a reference for healthy women; Embed functions for the ART scheme; Deadline; The zero point is the time when the patient begins ART treatment; This refers to the type of ovulation induction protocol.

[0029] In one alternative approach, the formula for calculating the UFP-DI is:

[0030]

[0031] in, This is a function representing the instantaneous risk intensity associated with fibroids. ; This refers to the total volume of the uterus. The optimal power exponent for the Box-Cox transform; For the spatial distribution tensor of fibroids; The gradient is the Frobenius norm. For the cumulative penalty functional of composite failure, ; Deadline Ovarian responsiveness index; This is an indicator function for periods of poor endometrial receptivity. For at a certain point in time The total volume of uterine fibroids was obtained through 3D ultrasound or MRI and smoothed by anisotropic diffusion.

[0032] In an alternative approach, the method further includes:

[0033] A uterine-adnexal functional coupling factor was constructed based on the uterine cavity microenvironment integrity index and the degree of perturbation of the fallopian tube-ovarian anatomical relationship. This uterine-adnexal functional coupling factor was then input as an enhancement feature into the interpretable baseline prediction model. The calculation formula for the uterine-adnexal functional coupling factor is as follows:

[0034]

[0035] in, This is a local perfusion activation function; For implantation weighting; The sensitivity index is The degree of disturbance of the attachments; The endometrial cavity region obtained through 3D ultrasound or MRI reconstruction; Additional perturbation penalty term corresponding to FAF phases; The FAF staging system for endometriosis; This is the position coordinate vector in the three-dimensional space of the uterine cavity.

[0036] According to another aspect of the present invention, a clinical prediction device for the fertility potential of patients with uterine fibroids is provided, comprising:

[0037] The anatomical integrated data extraction module is used to construct a phenotypic-deeply labeled fertility cohort of uterine fibroids and extract an anatomical integrated dataset of women of reproductive age with uterine fibroids who wish to conceive naturally or plan to undergo assisted reproductive technology. The anatomical integrated dataset includes baseline population parameters, reproductive endocrine axis function indicators, fertility reserve quantitative parameters, reproductive history, imaging features of uterine fibroids, anatomical features of uterine fibroids, intrauterine microenvironment integrity index, perturbation degree of fallopian tube-ovarian anatomical relationship, and comorbidity burden index.

[0038] An interpretable baseline prediction model building module is used to standardize the anatomical integrated dataset according to FIGO classification and clinical guidelines to generate an interpretable standardized feature matrix; the standardized feature matrix is ​​cross-validated by the Elastic Net elastic network model, and the minimum feature subset with preset significance and clinical relevance to the composite fertility outcome endpoint is selected under L1 / L2 mixed regularization constraints; the interpretable baseline prediction model is trained based on the minimum feature subset.

[0039] The individualized fertility potential calibration module is used to input the original clinical data of the individual to be predicted into the interpretable baseline prediction model and output a preliminary individualized fertility potential probability value; based on the fibroid growth rate, ART protocol type and ovarian stimulation responsiveness information of the individual to be predicted, the individualized fertility potential probability value is corrected to generate a time-varying calibrated uterine fibroid-related fertility potential index.

[0040] The clinical decision mapping module is used to map the calibrated uterine fibroid-related fertility potential index value to the clinical decision threshold range and generate a recommendation report. The recommendation report includes a heatmap of key risk drivers, a fibroid intervention necessity score, and personalized fertility pathway recommendations.

[0041] According to another aspect of the present invention, a computer device is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus;

[0042] The memory is used to store at least one executable instruction that causes the processor to perform the operation corresponding to the above-described clinical prediction method for the fertility potential of patients with uterine fibroids.

[0043] According to the scheme provided by the present invention, a phenotypic-deeply annotated fertility cohort of uterine fibroids is constructed, and an integrated anatomical dataset of women of reproductive age with uterine fibroids who wish to conceive naturally or intend to undergo assisted reproductive technology is extracted. The integrated anatomical dataset includes baseline population parameters, reproductive endocrine axis function indicators, fertility reserve quantitative parameters, reproductive history, imaging features of uterine fibroids, anatomical features of uterine fibroids, intrauterine microenvironment integrity index, tubo-ovarian anatomical relationship disturbance degree, and comorbidity burden index. The integrated anatomical dataset is standardized and encoded according to FIGO classification and clinical guidelines to generate an interpretable standardized feature matrix. This matrix is ​​then processed using Elastic... The Net elastic network model performs cross-validation on the standardized feature matrix, and selects the minimum feature subset with pre-defined significance and clinical relevance to the composite fertility outcome endpoint under L1 / L2 mixed regularization constraints. An interpretable baseline prediction model is trained based on this minimum feature subset. The original clinical data of the individual to be predicted is input into the interpretable baseline prediction model, outputting a preliminary individualized fertility potential probability value. Based on the individual's fibroid growth rate, ART protocol type, and ovarian stimulation responsiveness information, the individualized fertility potential probability value is corrected to generate a time-varying calibrated uterine fibroid-related fertility potential index. The calibrated uterine fibroid-related fertility potential index value is mapped to a clinical decision threshold range, generating a recommendation report. This recommendation report includes a heatmap of key risk drivers, a fibroid intervention necessity score, and personalized fertility pathway recommendations. This invention, through the product dynamic calibration coefficient vector and the time-varying UFP-DI index, longitudinally tracks and personally corrects fertility potential, significantly improving the accuracy and individualization of fertility potential assessment for patients with uterine fibroids.

[0044] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0045] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0046] Figure 1 A flowchart illustrating a clinical prediction method for fertility potential in patients with uterine fibroids according to an embodiment of the present invention is shown.

[0047] Figure 2A schematic diagram of the process for generating a recommendation report according to an embodiment of the present invention is shown;

[0048] Figure 3a A schematic diagram of a uterine fibroid fertility prediction model according to an embodiment of the present invention is shown;

[0049] Figure 3b This diagram illustrates the individual predictions of the most integrated model according to an embodiment of the present invention.

[0050] Figure 4 A schematic diagram of the framework of a clinical prediction device for the fertility potential of patients with uterine fibroids according to an embodiment of the present invention is shown.

[0051] Figure 5 A schematic diagram of the structure of a computer device according to an embodiment of the present invention is shown. Detailed Implementation

[0052] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0053] Figure 1 A flowchart illustrating a clinical prediction method for the fertility potential of patients with uterine fibroids according to an embodiment of the present invention is shown. Figure 2 A schematic diagram illustrating the process of generating a suggestion report according to an embodiment of the present invention is shown. Specifically, as... Figure 1 , Figure 2 As shown, it includes the following steps:

[0054] Step S101: Construct a phenotypic-deeply annotated fertility cohort for uterine fibroids and extract an integrated anatomical dataset of women of reproductive age with uterine fibroids who wish to conceive naturally or intend to undergo assisted reproductive technology; wherein, the integrated anatomical dataset includes population baseline parameters, reproductive endocrine axis function indicators, fertility reserve quantitative parameters, reproductive history, uterine fibroid imaging features, uterine fibroid anatomical features, intrauterine microenvironment integrity index, perturbation degree of fallopian tube-ovarian anatomical relationship, and comorbidity burden index.

[0055] In this embodiment, women of reproductive age with uterine fibroids who wish to conceive naturally or plan to undergo assisted reproductive technology (ART) are excluded, while those without fertility needs or who have lost their fertility are excluded. This ensures the cohort closely matches the practical application scenario of fertility potential assessment, avoiding confounding bias and improving the model's responsiveness to real clinical problems. Nine categories of indicators—including demographics, endocrine function, ovarian reserve, medical history, imaging anatomy, intrauterine microenvironment, adnexal spatial relationships, and comorbidities—comprehensively capture structural and functional factors affecting fertility, achieving a shift from "local lesions" to "overall functional status of the reproductive system." The intrauterine microenvironment integrity index and the degree of perturbation in the anatomical relationship between the fallopian tube and ovary not only focus on the presence of fibroids but also quantify their interference with key aspects such as the embryo implantation window, fallopian tube egg pickup function, and ovarian blood supply, more realistically reflecting the biological impact pathway of fibroids on fertility. Data collection strictly adheres to international guidelines such as FIGO classification and FAF staging to ensure cross-center comparability; simultaneously, individual details (such as endometrial blood flow and fibroid spatial location) are preserved to support the generation of personalized risk heat maps and fertility pathway recommendations, promoting the implementation of precision reproductive medicine. As shown in Table 1, the patient presented with uterine fibroids 3 years ago and had been trying to conceive for 2 years without success. The key parameters of the anatomical integrated dataset reflect a systematic quantitative assessment of the patient's fertility potential. The assessment of fertility potential has been deepened from a simple qualitative description of "submucosal fibroids" to a systematic quantitative characterization of "mildly decreased AMH at age 38 + type 2 fibroids causing uterine cavity torsion + impaired endometrial blood flow + mild compression of fallopian tube openings".

[0056] Table 1

[0057]

[0058] In one alternative approach, the baseline demographic parameters include age and body mass index; the reproductive endocrine axis function indicators include menstrual cycle stability indicators and ovulation function assessment indicators; the fertility reserve quantification parameters include anti-Müllerian hormone (AMH) levels and basal antral follicle count (AFC); the reproductive history includes past pregnancy history, delivery history, miscarriage history, and duration of infertility; the imaging characteristics of uterine fibroids include the number of lesions, maximum diameter, three-dimensional volume percentage, MRI enhancement pattern, and apparent diffusion coefficient; the anatomical characteristics of uterine fibroids include type 0-8 classification, number of fibroids, maximum fibroid diameter, and total fibroid volume percentage; the uterine cavity microenvironment integrity index includes the uterine cavity morphology distortion index and endometrial blood perfusion parameters; the tubo-ovarian anatomical relationship disturbance degree includes the tubal mechanical compression index and the spatial positional relationship of ovarian fibroids; and the comorbidity burden index includes the depth of adenomyosis invasion, endometriosis FAF staging, and adnexal lesions.

[0059] In this embodiment, instead of relying solely on rough descriptions such as fibroid size, quantitative radiomics parameters such as three-dimensional volume ratio, apparent diffusion coefficient (ADC), and MRI enhancement mode are used to objectively reflect the cell density, blood supply, and invasiveness of the fibroid, providing a more refined pathophysiological basis for prediction. By using the uterine cavity morphology distortion index, fallopian tube mechanical compression index, and the spatial relationship of ovarian fibroids, the interference effect of fibroids is assessed within the entire reproductive pathway system, with particular attention to their impact on key links such as embryo implantation (uterine cavity) and oocyte transport (fallopian tube-ovary), avoiding the oligarchic interference of comorbidities and making the prediction more closely reflect complex clinical realities. As shown in Table 2, the differences between the two patients are clear. Patient A's problem is relatively simple: 32 years old, with good reproductive reserves, only one type 3 fibroid in contact with the endometrium, a basically intact uterine cavity microenvironment, and no complications. The core of her fertility potential assessment is "evaluating the isolated impact of this type 3 fibroid on embryo implantation." Patient B's situation is much more complicated. She is 38 years old, has decreased fertility, multiple fibroids (including type 2), high tumor burden, uterine cavity tortuosity and poor endometrial blood supply, fallopian tube compression, and adenomyosis and endometriosis. Her fertility potential assessment faces a multiple dilemma of "age + fertility + multifactorial anatomical and functional abnormalities".

[0060] Table 2

[0061]

[0062] Step S102: Standardize the anatomical integrated dataset according to FIGO classification and clinical guidelines to generate an interpretable standardized feature matrix; cross-validate the standardized feature matrix using the Elastic Net model, and select the minimum feature subset that has preset significance and clinical relevance to the composite fertility outcome endpoint under L1 / L2 mixed regularization constraints; train an interpretable baseline prediction model based on the minimum feature subset.

[0063] In this embodiment, the messy raw clinical data (such as "submucosal fibroids" and "irregular menstruation") is transformed into a standardized feature matrix that conforms to statistical assumptions through FIGO typing encoder and Box-Cox transformation of continuous variables. This not only preserves the clinical interpretability of the features (e.g., one-hot encoding clearly distinguishes between fibroids of types 0-8) but also eliminates the influence of dimensions and skewed distribution. The Elastic Net model achieves automatic feature selection through L1 norm (LASSO), eliminating redundant and irrelevant variables; at the same time, it handles multicollinearity among features through L2 norm (Ridge), ensuring the stability of the model. It can screen the smallest subset of features that are truly relevant to fertility outcomes from high-dimensional clinical data, effectively preventing overfitting. The output of Elastic Net is a linear weighted model. The magnitude and direction (positive / negative) of the coefficient of each selected feature directly reflect the direction and strength of the factor's contribution to fertility outcomes, allowing doctors to intuitively understand "why the predicted probability is this value"—is it because of advanced age? Or because of type 2 fibroids? This further enhances clinical trust.

[0064] In one alternative approach, standardizing the integrated anatomical dataset according to FIGO classification and clinical guidelines to generate an interpretable standardized feature matrix further includes:

[0065] The continuous variables in the anatomical integrated dataset were subjected to a normality test. For continuous variables that did not meet the normality assumption, an adaptive Box-Cox power transformation was performed to obtain a continuous feature matrix. The continuous variables included age, body mass index (BMI), anti-Müllerian hormone (AMH) level, basal antral follicle count (AFC), three-dimensional volume ratio, apparent diffusion coefficient (ADC) value, uterine cavity morphology distortion index, endometrial blood perfusion parameters, fallopian tube mechanical compression index, and adenomyosis infiltration depth.

[0066] The categorical variables in the anatomical integrated dataset are input into the FIGO typing encoder. Hierarchical semantic mapping of uterine fibroid anatomical features is performed according to the International Federation of Gynecology and Obstetrics' (IFO) uterine fibroid typing standards to generate a fibroid typing one-hot encoding vector. A classification feature matrix is ​​generated based on the fibroid typing one-hot encoding vector, endometriosis FAF staging, and adnexal lesions. The categorical variables include menstrual cycle stability indicators, ovulation function assessment indicators, past pregnancy history, delivery history, miscarriage history, type 0-8 classification, MRI enhancement pattern, and the spatial location of ovarian fibroids.

[0067] The continuous feature matrix and the classification feature matrix are encoded in a temporal sequence to generate an interpretable standardized feature matrix.

[0068] In this embodiment, one-hot encoding is used for unordered categorical variables such as menstrual cycle stability, previous pregnancy history, and MRI enhancement patterns to avoid the model incorrectly assuming an order relationship between these categories (e.g., mistakenly assuming "homogeneous enhancement" is greater than "non-homogeneous enhancement"). Simultaneously, prior clinical knowledge such as FAF staging and adnexal lesions is incorporated into the encoding process to ensure the clinical rationality of the categorical feature matrix. By encoding the temporal relationship between the continuous feature matrix and the categorical feature matrix, the changing trends of key parameters over time can be captured (e.g., the rate of change in fibroid volume between two visits, the rate of decrease in AMH, etc.), enabling the model to perceive the speed of disease progression and the dynamic changes in the patient's condition.

[0069] Step S103: Input the original clinical data of the individual to be predicted into the interpretable baseline prediction model and output a preliminary individualized fertility potential probability value; based on the fibroid growth rate, ART protocol type and ovarian stimulation responsiveness information of the individual to be predicted, correct the individualized fertility potential probability value and generate a time-varying calibrated uterine fibroid-related fertility potential index.

[0070] In this embodiment, the changes in fibroid growth rate, ART protocol type, and ovarian stimulation responsiveness over time allow the predicted results to reflect the patient's current true physiological state. Fibroid growth rate quantifies the cumulative effect of local mechanical disturbances to the uterus (e.g., increased uterine cavity deformation); ovarian stimulation responsiveness (e.g., number of retrieved oocytes, follicle release rate) reflects the actual functional output of the endocrine follicle axis; their synergistic calibration provides a more comprehensive portrayal of the dynamic trajectory of fertility. Different ovulation induction protocols (e.g., long protocol, antagonist protocol, microstimulation) have varying effects on patients with uterine fibroids. By using the ART protocol type as a calibration factor, the expected success rate under different treatment pathways can be automatically adjusted. As shown in Table 3, although the ovarian response was good in this IVF cycle (number of retrieved oocytes exceeded expectations), the fibroids were observed to be continuously growing over the past 6 months. This negative impact on the uterine environment offset the advantage of the good ovarian response, resulting in a dynamic potential index (UFP-DI) of 12% for achieving pregnancy in the current cycle, lower than the baseline prediction (25%) based on static information.

[0071] Table 3

[0072]

[0073] In one alternative approach, the individualized fertility potential probability value is corrected based on the fibroid growth rate, ART protocol type, and ovarian stimulation responsiveness information of the individual to be predicted, and the generation of a time-varying calibrated uterine fibroid-related fertility potential index further includes:

[0074] The longitudinal imaging follow-up data of the individual to be predicted is obtained, and MRI or ultrasound images of uterine fibroids at at least two different time points are extracted and input into the fibroid segmentation model to calculate the three-dimensional volume of the fibroids at each time point; wherein, the time span of the longitudinal imaging follow-up data is at least 3 menstrual cycles;

[0075] Ovarian responsiveness is dynamically assessed based on the clinical data of the ART cycle of the individuals to be predicted, and the ovarian responsiveness index is calculated. The ART cycle clinical data includes the type of ovulation induction protocol, initiation dose, number of Gn days, total Gn dose, daily estrogen level on HCG day, daily progesterone level on HCG day, number of oocytes retrieved, MII oocyte rate, fertilization rate, and high-quality embryo rate. The ovarian responsiveness index includes the ratio of actual oocytes retrieved to expected oocytes retrieved and the follicle release rate.

[0076] The influence weight of fertility potential is estimated based on the three-dimensional volume of the fibroid and the ovarian responsiveness index to obtain a dynamic calibration coefficient vector; the individualized fertility potential probability value is used as the baseline predicted value and the dynamic calibration coefficient vector is modified by multiplication effect to generate a preliminary calibration value; the number of previous failed ART cycles is negatively weighted according to the ART cycle number decay factor to generate a cycle cumulative effect correction term.

[0077] The initial calibration value and the periodic cumulative effect correction term are dynamically synthesized to generate a time-varying calibrated uterine fibroid-related fertility potential index.

[0078] In this embodiment, MRI / ultrasound images from at least two time points (interval ≥ 3 menstrual cycles) are required. Three-dimensional volume changes are calculated using a segmentation model, avoiding reliance solely on a rough estimate based on the maximum diameter. This approach more sensitively captures fibroid growth trends (such as volume doubling time). Combining the number of retrieved oocytes, the actual / expected oocyte retrieval ratio, and the follicle output rate (FORT) comprehensively reflects the ovary's true responsiveness to stimulation, providing a more accurate representation of the quality and quantity of available oocytes in the current cycle than a single AMH or AFC value. The three-dimensional volume change of the fibroid represents the dynamic accumulation of local mechanical disturbances in the uterus, while the ovarian responsiveness index represents the functional state of the endocrine-follicle axis. Both are mapped to the cumulative mechanical disturbance potential and the endocrine-follicle coupling impedance, respectively, achieving dual-channel calibration of anatomy and function, more closely reflecting the evolution of actual fertility. The type of ovulation induction protocol (e.g., long protocol, antagonist, PPOS) is used as a calibration factor. A pre-trained embedding function is used to adjust the weights of different protocols on fibroid patients, enabling the model to adapt to different clinical pathways. Negative weighting of past ART failure cycles (e.g., high weight for recent failures and low weight for long-term failures) prevents the neglect of patients' real-life setbacks and enhances predictive robustness.

[0079] In one alternative approach, the formula for calculating the dynamic calibration coefficient vector is:

[0080]

[0081] in, It is a sigmoid mapping kernel with adjustable kurtosis. , The median threshold for a multi-center queue; It is an exponentially decaying memory kernel. , , The half-life of the effect of fibroid volume on pregnancy rate; For anisotropic diffusion smoothing function; The distribution function is used as a reference for healthy women; Embed functions for the ART scheme; Deadline; The zero point is the time when the patient begins ART treatment; This refers to the type of ovulation induction protocol.

[0082] In this embodiment, the exponentially decaying kernel assigns a higher weight to recent changes in fibroids, reflecting clinical observations that the impact of fibroids on pregnancy partially diminishes over time (e.g., (For 6 months). To avoid excessive penalty caused by linear accumulation and improve the predictive validity of patients in long-term follow-up. The actual ovarian responsiveness (ORI) of patients is compared with reference values ​​for healthy women of the same age, ensuring that the calibration results have a population benchmark rather than absolute numerical judgment. Antagonist regimens, microstimulation, PPOS, etc., are mapped to continuous vectors, automatically learning the relative advantages and disadvantages of different regimens in the fibroid population (e.g., some regimens may stimulate fibroid growth), thereby dynamically adjusting the predicted values. Figure 3a , Figure 3b As shown, although the patient's baseline predicted probability was 60%, due to the rapid growth of the fibroids (the largest fibroid had a maximum diameter of 3 cm), lower-than-expected ovarian response, and the use of a high-stimulation protocol, her current effective fertility potential was adjusted to 60% × 0.293 ≈ 17.59%. It is recommended to suspend the high-stimulation protocol and switch to a milder protocol, so that the prediction is no longer a "black box probability".

[0083] In an alternative approach, the method further includes:

[0084] A uterine-adnexal functional coupling factor was constructed based on the uterine cavity microenvironment integrity index and the degree of perturbation of the fallopian tube-ovarian anatomical relationship. This uterine-adnexal functional coupling factor was then input as an enhancement feature into the interpretable baseline prediction model. The calculation formula for the uterine-adnexal functional coupling factor is as follows:

[0085]

[0086] in, This is a local perfusion activation function; For implantation weighting; The sensitivity index is The degree of disturbance of the attachments; The endometrial cavity region obtained through 3D ultrasound or MRI reconstruction; Additional perturbation penalty term corresponding to FAF phases; The FAF staging system for endometriosis; This is the position coordinate vector in the three-dimensional space of the uterine cavity.

[0087] In this embodiment, the functional matching degree between the endometrial implantation microenvironment and the oocyte pickup and transport pathway is quantified by coupling factors, more realistically reflecting the complete reproductive pathway efficiency required for natural conception. Voxel-level integration is performed on the reconstructed uterine cavity region, where... This indicates the importance of different uterine cavity locations for embryo implantation (e.g., higher weight for the fundus and lower weight for the lower segment). Reflecting local endometrial blood perfusion activity (obtained by contrast-enhanced ultrasound or MRI perfusion imaging), it upgrades the integrity of the intrauterine microenvironment from a qualitative description (such as "mild adhesions") to a calculable continuous functional indicator.

[0088] Step S104: Map the calibrated uterine fibroid-related fertility potential index value to the clinical decision threshold range and generate a recommendation report, wherein the recommendation report includes a heatmap of key risk drivers, a fibroid intervention necessity score, and personalized fertility pathway recommendations.

[0089] In this embodiment, the UFP-DI is a continuous index that has undergone complex dynamic calibration. Its numerical values ​​(e.g., 6.2%, 15.8%) may lack intuitive clinical significance for patients and even some clinicians. A multi-level threshold mapper maps abstract numerical values ​​to fertility potential levels (e.g., "low," "medium," "high"), making the assessment results easier to understand. Through a feature attribution weight matrix and a heatmap of key risk drivers, the system visually demonstrates the extent (contribution weight) to which specific factors (e.g., "type 2 fibroids," "reduced endometrial blood flow") drive the final UFP-DI value, directly indicating the target of clinical intervention. Addressing the clinical dilemma faced by patients with uterine fibroids (e.g., whether or not surgery should be performed first), a fibroid intervention necessity score quantifies the urgency and expected benefit of treating fibroids to improve fertility potential under the combined effect of all current factors.

[0090] In one alternative approach, mapping the calibrated uterine fibroid-related fertility potential index value to a clinical decision threshold range and generating a recommendation report further includes:

[0091] The uterine fibroid-related fertility potential index is input into a multi-level threshold mapper, and the UFP-DI is mapped to the corresponding fertility potential level based on the dynamic threshold range determined by large-sample multi-center clinical validation.

[0092] The UFP-DI and its corresponding fertility potential level are analyzed using risk driver factors to obtain analytical features. The marginal contribution value of each analytical feature to UFP-DI is calculated based on the feature contribution weight vector of the individual to be predicted, and a feature attribution weight matrix is ​​generated. The feature attribution weight matrix is ​​sorted in descending order, and the top-K key risk driver factors with a cumulative contribution rate of more than 85% are selected.

[0093] The top-K key risk drivers and their corresponding attribution weights are input into a heatmap generator to generate a heatmap of key risk drivers. The heatmap displays the contribution direction and intensity of each risk factor in matrix form, and provides a network topology diagram of the interaction effects between factors.

[0094] The UFP-DI, fertility potential level, and heatmap of key risk drivers are input into a path planning decision tree generator to generate a phased fertility path planning decision tree with the goal of maximizing the cumulative pregnancy probability. The decision tree is rooted at the current UFP-DI and branches along the time axis to output personalized fertility path suggestions.

[0095] In this embodiment, the top-K key risk drivers are selected based on a cumulative contribution rate exceeding 85%, achieving an optimal balance of information condensation. This retains the vast majority (over 85%) of predictive information while eliminating interference from secondary factors. A path planning decision tree generator outputs phased fertility path planning with the time axis as the branch. The optimization objective of the decision tree is to maximize the cumulative pregnancy probability, rather than the single success rate or the shortest time. The root node of the decision tree is the current patient's UFP-DI value and its corresponding fertility potential level, meaning that all subsequent branches and probability predictions are completely personalized (anchored to the patient's specific state). Patients with different UFP-DI values, even facing similar clinical problems, will have different branch structures and expected probabilities in their decision trees.

[0096] In one alternative approach, the formula for calculating the UFP-DI is:

[0097]

[0098] in, This is a function representing the instantaneous risk intensity associated with fibroids. ; This refers to the total volume of the uterus. The optimal power exponent for the Box-Cox transform; For the spatial distribution tensor of fibroids; The gradient is the Frobenius norm. For the cumulative penalty functional of composite failure, ; Deadline Ovarian responsiveness index; This is an indicator function for periods of poor endometrial receptivity. For at a certain point in time The total volume of uterine fibroids was obtained through 3D ultrasound or MRI and smoothed by anisotropic diffusion.

[0099] In this embodiment, Describe the anatomical and structural risks that fibroids pose over time (such as volume compression and spatial distortion). Quantifying the burden of functional failure (such as ART cycle failure, poor endometrial receptivity) is closer to clinical reality. Among these, the more ART failures, the greater the penalty; however, if the patient has good cumulative ovarian responsiveness ( If the AMH level is high, the penalty for failure is exponentially reduced, indicating that individuals with high reserves can withstand more trial and error, which is consistent with clinical observations. Even if individuals with high AMH experience 1–2 failures, they still have a good prognosis.

[0100] According to the scheme provided by the present invention, a phenotypic-deeply annotated fertility cohort of uterine fibroids is constructed, and an integrated anatomical dataset of women of reproductive age with uterine fibroids who wish to conceive naturally or intend to undergo assisted reproductive technology is extracted. The integrated anatomical dataset includes baseline population parameters, reproductive endocrine axis function indicators, fertility reserve quantitative parameters, reproductive history, imaging features of uterine fibroids, anatomical features of uterine fibroids, intrauterine microenvironment integrity index, tubo-ovarian anatomical relationship disturbance degree, and comorbidity burden index. The integrated anatomical dataset is standardized and encoded according to FIGO classification and clinical guidelines to generate an interpretable standardized feature matrix. This matrix is ​​then processed using Elastic... The Net elastic network model performs cross-validation on the standardized feature matrix, and selects the minimum feature subset with pre-defined significance and clinical relevance to the composite fertility outcome endpoint under L1 / L2 mixed regularization constraints. An interpretable baseline prediction model is trained based on this minimum feature subset. The original clinical data of the individual to be predicted is input into the interpretable baseline prediction model, outputting a preliminary individualized fertility potential probability value. Based on the individual's fibroid growth rate, ART protocol type, and ovarian stimulation responsiveness information, the individualized fertility potential probability value is corrected to generate a time-varying calibrated uterine fibroid-related fertility potential index. The calibrated uterine fibroid-related fertility potential index value is mapped to a clinical decision threshold range, generating a recommendation report. This recommendation report includes a heatmap of key risk drivers, a fibroid intervention necessity score, and personalized fertility pathway recommendations. This invention, through the product dynamic calibration coefficient vector and the time-varying UFP-DI index, longitudinally tracks and personally corrects fertility potential, significantly improving the accuracy and individualization of fertility potential assessment for patients with uterine fibroids.

[0101] Figure 4 A schematic diagram of the framework of a clinical prediction device for the fertility potential of patients with uterine fibroids, according to an embodiment of the present invention, is shown. The clinical prediction device for the fertility potential of patients with uterine fibroids includes:

[0102] The anatomical integrated data extraction module 410 is used to construct a phenotypic-deeply labeled uterine fibroid fertility cohort and extract an anatomical integrated dataset of women of reproductive age with uterine fibroids who have a desire for natural conception or intend to undergo assisted reproductive technology; wherein, the anatomical integrated dataset includes population baseline parameters, reproductive endocrine axis function indicators, fertility reserve quantitative parameters, reproductive history, uterine fibroid imaging characteristics, uterine fibroid anatomical characteristics, uterine cavity microenvironment integrity index, fallopian tube-ovarian anatomical relationship disturbance degree, and comorbidity burden index;

[0103] The interpretable baseline prediction model building module 420 is used to standardize the anatomical integrated dataset according to FIGO classification and clinical guidelines to generate an interpretable standardized feature matrix; cross-validate the standardized feature matrix through the Elastic Net elastic network model, and select the minimum feature subset with preset significance and clinical relevance to the composite fertility outcome endpoint under L1 / L2 mixed regularization constraints, and train the interpretable baseline prediction model based on the minimum feature subset.

[0104] The individualized fertility potential calibration module 430 is used to input the original clinical data of the individual to be predicted into the interpretable baseline prediction model and output a preliminary individualized fertility potential probability value; based on the fibroid growth rate, ART protocol type and ovarian stimulation responsiveness information of the individual to be predicted, the individualized fertility potential probability value is corrected to generate a time-varying calibrated uterine fibroid-related fertility potential index.

[0105] The clinical decision mapping module 440 is used to map the calibrated uterine fibroid-related fertility potential index value to the clinical decision threshold range and generate a recommendation report, wherein the recommendation report includes a heatmap of key risk drivers, a fibroid intervention necessity score, and personalized fertility pathway recommendations.

[0106] Figure 5 The diagram shows a structural schematic of an embodiment of the computer device of the present invention. The specific embodiments of the present invention do not limit the specific implementation of the computer device.

[0107] like Figure 5 As shown, the computer device may include: a processor 502, a communications interface 504, a memory 506, and a communications bus 508.

[0108] The processor 502, communication interface 504, and memory 506 communicate with each other via communication bus 508. Communication interface 504 is used to communicate with other network elements such as clients or other servers. The processor 502 executes program 510, specifically performing the relevant steps in the above-described clinical prediction method embodiment for the fertility potential of patients with uterine fibroids.

[0109] Specifically, program 510 may include program code that includes computer operation instructions.

[0110] Processor 502 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The computer device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0111] Memory 506 is used to store program 510. Memory 506 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0112] According to the scheme provided by the present invention, a phenotypic-deeply annotated fertility cohort of uterine fibroids is constructed, and an integrated anatomical dataset of women of reproductive age with uterine fibroids who wish to conceive naturally or intend to undergo assisted reproductive technology is extracted. The integrated anatomical dataset includes baseline population parameters, reproductive endocrine axis function indicators, fertility reserve quantitative parameters, reproductive history, imaging features of uterine fibroids, anatomical features of uterine fibroids, intrauterine microenvironment integrity index, tubo-ovarian anatomical relationship disturbance degree, and comorbidity burden index. The integrated anatomical dataset is standardized and encoded according to FIGO classification and clinical guidelines to generate an interpretable standardized feature matrix. This matrix is ​​then processed using Elastic... The Net elastic network model performs cross-validation on the standardized feature matrix, and selects the minimum feature subset with pre-defined significance and clinical relevance to the composite fertility outcome endpoint under L1 / L2 mixed regularization constraints. An interpretable baseline prediction model is trained based on this minimum feature subset. The original clinical data of the individual to be predicted is input into the interpretable baseline prediction model, outputting a preliminary individualized fertility potential probability value. Based on the individual's fibroid growth rate, ART protocol type, and ovarian stimulation responsiveness information, the individualized fertility potential probability value is corrected to generate a time-varying calibrated uterine fibroid-related fertility potential index. The calibrated uterine fibroid-related fertility potential index value is mapped to a clinical decision threshold range, generating a recommendation report. This recommendation report includes a heatmap of key risk drivers, a fibroid intervention necessity score, and personalized fertility pathway recommendations. This invention, through the product dynamic calibration coefficient vector and the time-varying UFP-DI index, longitudinally tracks and personally corrects fertility potential, significantly improving the accuracy and individualization of fertility potential assessment for patients with uterine fibroids.

[0113] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination of all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed can be employed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose. Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of the invention and form different embodiments. For example, in the following claims, any of the claimed embodiments can be used in any combination. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims listing several devices, several of these devices may be embodied by the same hardware item. Unless otherwise specified, the steps in the above embodiments should not be construed as limiting the order of execution.

Claims

1. A clinical method for predicting the fertility potential of patients with uterine fibroids, characterized in that, include: A phenotypic-deeply annotated fertility cohort for uterine fibroids was constructed, and an integrated anatomical dataset was extracted from women of reproductive age with uterine fibroids who wished to conceive naturally or planned to undergo assisted reproductive technology. The integrated anatomical dataset included baseline population parameters, reproductive endocrine axis function indicators, fertility reserve quantitative parameters, reproductive history, imaging features of uterine fibroids, anatomical features of uterine fibroids, intrauterine microenvironment integrity index, perturbation degree of fallopian tube-ovarian anatomical relationship, and comorbidity burden index. The anatomical integrated dataset was standardized and encoded according to FIGO classification and clinical guidelines to generate an interpretable standardized feature matrix. The standardized feature matrix was cross-validated using the Elastic Net model. Under L1 / L2 mixed regularization constraints, the minimum feature subset with preset significance and clinical relevance to the composite fertility outcome endpoint was selected. An interpretable baseline prediction model was trained based on the minimum feature subset. The original clinical data of the individual to be predicted is input into the interpretable baseline prediction model, and a preliminary individualized fertility potential probability value is output. Based on the fibroid growth rate, ART protocol type and ovarian stimulation responsiveness information of the individual to be predicted, the individualized fertility potential probability value is corrected to generate a time-varying calibrated uterine fibroid-related fertility potential index. The calibrated uterine fibroid-related fertility potential index values ​​are mapped to clinical decision threshold ranges and a recommendation report is generated. The recommendation report includes a heatmap of key risk drivers, a fibroid intervention necessity score, and personalized fertility pathway recommendations.

2. The clinical prediction method for fertility potential in patients with uterine fibroids according to claim 1, characterized in that, The baseline demographic parameters include age and body mass index; the reproductive endocrine axis function indicators include menstrual cycle stability indicators and ovulation function assessment indicators; the fertility reserve quantification parameters include anti-Müllerian hormone (AMH) levels and basal antral follicle count (AFC); the reproductive history includes past pregnancy history, delivery history, miscarriage history, and duration of infertility; the imaging characteristics of uterine fibroids include the number of lesions, maximum diameter, three-dimensional volume percentage, MRI enhancement pattern, and apparent diffusion coefficient; the anatomical characteristics of uterine fibroids include type 0-8 classification, number of fibroids, maximum fibroid diameter, and total fibroid volume percentage; the uterine cavity microenvironment integrity index includes the uterine cavity morphology distortion index and endometrial blood perfusion parameters; the tubo-ovarian anatomical relationship disturbance degree includes the tubal mechanical compression index and the spatial positional relationship of ovarian fibroids; the comorbidity burden index includes the depth of adenomyosis invasion, endometriosis FAF staging, and adnexal lesions.

3. The clinical prediction method for fertility potential in patients with uterine fibroids according to claim 2, characterized in that, The standardized encoding of the anatomical integrated dataset based on FIGO classification and clinical guidelines to generate an interpretable standardized feature matrix further includes: The continuous variables in the anatomical integrated dataset were subjected to a normality test. For continuous variables that did not meet the normality assumption, an adaptive Box-Cox power transformation was performed to obtain a continuous feature matrix. The continuous variables included age, body mass index (BMI), anti-Müllerian hormone (AMH) level, basal antral follicle count (AFC), three-dimensional volume ratio, apparent diffusion coefficient (ADC) value, uterine cavity morphology distortion index, endometrial blood perfusion parameters, fallopian tube mechanical compression index, and adenomyosis infiltration depth. The categorical variables in the anatomical integrated dataset are input into the FIGO typing encoder. Hierarchical semantic mapping of uterine fibroid anatomical features is performed according to the International Federation of Gynecology and Obstetrics' (IFO) uterine fibroid typing standards to generate a fibroid typing one-hot encoding vector. A classification feature matrix is ​​generated based on the fibroid typing one-hot encoding vector, endometriosis FAF staging, and adnexal lesions. The categorical variables include menstrual cycle stability indicators, ovulation function assessment indicators, past pregnancy history, delivery history, miscarriage history, type 0-8 classification, MRI enhancement pattern, and the spatial location of ovarian fibroids. The continuous feature matrix and the classification feature matrix are encoded in a temporal sequence to generate an interpretable standardized feature matrix.

4. The clinical prediction method for fertility potential in patients with uterine fibroids according to claim 1, characterized in that, Based on the fibroid growth rate, ART protocol type, and ovarian stimulation responsiveness information of the individual to be predicted, the individualized fertility potential probability value is corrected to generate a time-varying calibrated uterine fibroid-related fertility potential index, which further includes: The longitudinal imaging follow-up data of the individual to be predicted is obtained, and MRI or ultrasound images of uterine fibroids at at least two different time points are extracted and input into the fibroid segmentation model to calculate the three-dimensional volume of the fibroids at each time point; wherein, the time span of the longitudinal imaging follow-up data is at least 3 menstrual cycles; Ovarian responsiveness is dynamically assessed based on the clinical data of the ART cycle of the individuals to be predicted, and the ovarian responsiveness index is calculated. The ART cycle clinical data includes the type of ovulation induction protocol, initiation dose, number of Gn days, total Gn dose, daily estrogen level on HCG day, daily progesterone level on HCG day, number of oocytes retrieved, MII oocyte rate, fertilization rate, and high-quality embryo rate. The ovarian responsiveness index includes the ratio of actual oocytes retrieved to expected oocytes retrieved and the follicle release rate. The influence weight of fertility potential is estimated based on the three-dimensional volume of the fibroid and the ovarian responsiveness index to obtain a dynamic calibration coefficient vector; the individualized fertility potential probability value is used as the baseline predicted value and the dynamic calibration coefficient vector is modified by multiplication effect to generate a preliminary calibration value; the number of previous failed ART cycles is negatively weighted according to the ART cycle number decay factor to generate a cycle cumulative effect correction term. The initial calibration value and the periodic cumulative effect correction term are dynamically synthesized to generate a time-varying calibrated uterine fibroid-related fertility potential index.

5. The clinical prediction method for fertility potential in patients with uterine fibroids according to claim 4, characterized in that, The calibrated uterine fibroid-related fertility potential index values ​​are mapped to clinical decision threshold ranges, and a recommendation report is generated, which further includes: The uterine fibroid-related fertility potential index is input into a multi-level threshold mapper, and the UFP-DI is mapped to the corresponding fertility potential level based on the dynamic threshold range determined by large-sample multi-center clinical validation. The UFP-DI and its corresponding fertility potential level are analyzed using risk driver factors to obtain analytical features. The marginal contribution value of each analytical feature to UFP-DI is calculated based on the feature contribution weight vector of the individual to be predicted, and a feature attribution weight matrix is ​​generated. The feature attribution weight matrix is ​​sorted in descending order, and the top-K key risk driver factors with a cumulative contribution rate of more than 85% are selected. The top-K key risk drivers and their corresponding attribution weights are input into a heatmap generator to generate a heatmap of key risk drivers. The heatmap displays the contribution direction and intensity of each risk factor in matrix form, and provides a network topology diagram of the interaction effects between factors. The UFP-DI, fertility potential level, and heatmap of key risk drivers are input into a path planning decision tree generator to generate a phased fertility path planning decision tree with the goal of maximizing the cumulative pregnancy probability. The decision tree is rooted at the current UFP-DI and branches along the time axis to output personalized fertility path suggestions.

6. The clinical prediction method for fertility potential in patients with uterine fibroids according to claim 4, characterized in that, The formula for calculating the dynamic calibration coefficient vector is as follows: ； in, A sigmoid mapping kernel with adjustable kurtosis. , The median threshold for a multi-center queue; It is an exponentially decaying memory kernel. , , The half-life of the effect of fibroid volume on pregnancy rate; For anisotropic diffusion smoothing function; The distribution function is used as a reference for healthy women; Embed functions for the ART scheme; Deadline; The zero point is the time when the patient begins ART treatment; This refers to the type of ovulation induction protocol.

7. The clinical prediction method for fertility potential in patients with uterine fibroids according to claim 5, characterized in that, The calculation formula for UFP-DI is as follows: ； in, This is a function representing the instantaneous risk intensity associated with fibroids. ; This refers to the total volume of the uterus. The optimal power exponent for the Box-Cox transform; For the spatial distribution tensor of fibroids; The gradient is the Frobenius norm. For the cumulative penalty functional of composite failure, ; Deadline Ovarian responsiveness index; This is an indicator function for periods of poor endometrial receptivity. For at a certain point in time The total volume of uterine fibroids was obtained through 3D ultrasound or MRI and smoothed by anisotropic diffusion.

8. The clinical prediction method for fertility potential in patients with uterine fibroids according to claim 2, characterized in that, The method further includes: A uterine-adnexal functional coupling factor was constructed based on the uterine cavity microenvironment integrity index and the degree of perturbation of the fallopian tube-ovarian anatomical relationship. This uterine-adnexal functional coupling factor was then input as an enhancement feature into the interpretable baseline prediction model. The calculation formula for the uterine-adnexal functional coupling factor is as follows: ； in, This is a local perfusion activation function; For implantation weighting; The sensitivity index is The degree of disturbance of the attachments; The endometrial cavity region obtained through 3D ultrasound or MRI reconstruction; Additional perturbation penalty term corresponding to FAF phases; The FAF staging system for endometriosis; This is the position coordinate vector in the three-dimensional space of the uterine cavity.

9. A clinical predictive device for the fertility potential of patients with uterine fibroids, characterized in that, include: The anatomical integrated data extraction module is used to construct a phenotypic-deeply labeled fertility cohort of uterine fibroids and extract an anatomical integrated dataset of women of reproductive age with uterine fibroids who wish to conceive naturally or plan to undergo assisted reproductive technology. The anatomical integrated dataset includes baseline population parameters, reproductive endocrine axis function indicators, fertility reserve quantitative parameters, reproductive history, imaging features of uterine fibroids, anatomical features of uterine fibroids, intrauterine microenvironment integrity index, perturbation degree of fallopian tube-ovarian anatomical relationship, and comorbidity burden index. An interpretable baseline prediction model building module is used to standardize the anatomical integrated dataset according to FIGO classification and clinical guidelines to generate an interpretable standardized feature matrix; the standardized feature matrix is ​​cross-validated by the Elastic Net elastic network model, and the minimum feature subset with preset significance and clinical relevance to the composite fertility outcome endpoint is selected under L1 / L2 mixed regularization constraints; the interpretable baseline prediction model is trained based on the minimum feature subset. The individualized fertility potential calibration module is used to input the original clinical data of the individual to be predicted into the interpretable baseline prediction model and output a preliminary individualized fertility potential probability value; based on the fibroid growth rate, ART protocol type and ovarian stimulation responsiveness information of the individual to be predicted, the individualized fertility potential probability value is corrected to generate a time-varying calibrated uterine fibroid-related fertility potential index. The clinical decision mapping module is used to map the calibrated uterine fibroid-related fertility potential index value to the clinical decision threshold range and generate a recommendation report. The recommendation report includes a heatmap of key risk drivers, a fibroid intervention necessity score, and personalized fertility pathway recommendations.

10. A computer device, comprising: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction that causes the processor to perform the operation corresponding to the clinical prediction method for fertility potential of patients with uterine fibroids as described in any one of claims 1-8.

Images