Intelligent auxiliary system for predicting success rate of in-vitro fertilization-embryo transplantation pregnancy assistance based on LightGBM algorithm

Abstract

The invention discloses an intelligent auxiliary system for predicting the success rate of in-vitro fertilization embryo transplantation (IVF-ET) based on clinical characteristics. The prediction accuracy and the calculation efficiency of the success rate of the IVF-ET are improved by applying a Light GBM (Light Gray Boosting Mach) algorithm. The system performs efficient data processing and feature extraction by integrating a plurality of key clinical features, including basic health indexes such as blood glucose level, hormone level, semen quality and the like, so as to construct an accurate prediction model; the LightGBM algorithm is introduced to carry out efficient processing and accurate prediction of multi-dimensional features, and the algorithm is especially suitable for large-scale, high-dimensional and sparse data due to the rapid training and reasoning ability, and effectively reduces the calculation complexity and model overfitting. According to the method, the prediction accuracy of the IVF-ET success rate is improved, real-time success rate prediction support is provided for a doctor, a personalized treatment scheme can be formulated more scientifically, meanwhile, occupation of computing resources is reduced, and rapid feedback of prediction is achieved.

Description

Technical Field

[0001] The present invention relates to the application of artificial intelligence and machine learning technologies in the biomedical field, and in particular to an intelligent auxiliary system for predicting the success rate of in vitro fertilization-embryo transfer (IVF-ET) assisted pregnancy based on the LightGBM (Light Gradient Boosting Machine) algorithm. Background Art

[0002] With the continuous development of artificial intelligence technology, machine learning, as a key field, has gradually penetrated into various industries, especially in the medical field, showing great application potential. Traditional machine learning algorithms have been applied to disease prediction, personalized treatment recommendations, and patient health management. However, in complex medical processes such as in vitro fertilization-embryo transfer (IVF-ET), existing prediction models are still insufficient in accuracy and real-time performance. Especially when dealing with large and multi-dimensional data sets, the computational overhead and inference time of the model are often greatly increased, affecting the efficiency and timeliness of medical decision-making.

[0003] At present, predictions in clinical practice mainly rely on physician experience and traditional statistical methods, but they show high limitations in predicting the success rate of IVF-ET. First, doctors' experience and judgment are easily affected by subjectivity, which leads to deviations in the prediction of success rate. Second, traditional statistical models usually lack flexibility when dealing with multivariate and large-scale feature data, and it is difficult to capture the complex nonlinear relationship between IVF-ET results and multiple physiological factors. These deficiencies limit the accuracy of clinical predictions to a certain extent, and therefore a more intelligent and efficient prediction system is urgently needed.

[0004] In the past, there have been attempts to use machine learning methods for IVF-ET success rate prediction. XGBoost (Extreme Gradient Boosting), as a popular gradient boosting algorithm, has been widely studied for its powerful predictive performance. The XGBoost algorithm performs well in many machine learning tasks by building multiple decision trees for prediction. However, it has limitations when applied to IVF-ET success rate prediction: XGBoost requires large memory and computing resources when processing large-scale data sets. As the amount of data increases, the time for model training and inference increases significantly, which is a challenge for clinical environments that require rapid response. In addition, in IVF-ET data sets, sparse and unbalanced data distribution problems are often encountered. Although XGBoost can handle this data, its performance is not optimal, especially when the data is extremely sparse or the categories are extremely unbalanced. As clinical data continues to grow, it is difficult for the model to be flexibly expanded to adapt to new data.

[0005] In view of these limitations of XGBoost, technicians in this field have been seeking algorithms that are more efficient and more adaptable to specific data characteristics. Summary of the invention

[0006] The present invention provides an intelligent auxiliary system for predicting the success rate of in vitro fertilization and embryo transfer (IVF-ET) assisted pregnancy based on the LightGBM (Light Gradient Boosting Machine) algorithm. Through the LightGBM algorithm, high-precision prediction of the success rate of in vitro fertilization and embryo transfer (IVF-ET) assisted pregnancy is achieved, and customized health adjustment strategies are provided for patients based on the prediction results, aiming to significantly improve the treatment success rate and patient experience.

[0007] The following is a detailed description of the technical solution of the present invention:

[0008] This technical solution includes the following steps to form a complete data processing and prediction process:

[0009] Data collection and input: The first step of this protocol is to collect clinical data from patients. These data include, but are not limited to, key indicators such as blood sugar levels, hormone levels, and semen quality, which are considered to be closely related to the success rate of IVF-ET. The data is entered into the system through the data input module to provide raw materials for subsequent processing and analysis.

[0010] Data preprocessing: The collected raw data often contains missing values, outliers, or inconsistent formats, which requires the data preprocessing module to handle. This module is responsible for cleaning the data, filling or deleting missing values, filtering outliers, and standardizing the data to a uniform scale for model processing.

[0011] Feature extraction and selection: Based on data preprocessing, the feature extraction module uses the LightGBM algorithm to conduct in-depth analysis of the data. The LightGBM algorithm can calculate the split gain of each feature, automatically evaluate the contribution of the feature to the model's predictive ability, and sort the features accordingly. This step is key to improving the accuracy of model predictions because it can identify the most influential features.

[0012] Model training: The prediction model training module is responsible for using the above extracted features to train a prediction model based on the LightGBM algorithm. The model learns patterns in the data by building multiple decision trees and is able to handle large-scale, high-dimensional data while maintaining low resource consumption.

[0013] Prediction and result output: The trained model will be used to predict the success rate of IVF-ET. The result generation module will output the predicted success rate of each patient and generate personalized health adjustment suggestions based on the key influencing factors analyzed by the model, helping patients and doctors understand the factors that may affect the success rate and take corresponding measures.

[0014] Feedback and model optimization: In practical applications, the feedback data from patients after treatment is a valuable resource. The feedback optimization module compares this feedback data with the original prediction results, analyzes the prediction error of the model, and adjusts the model parameters accordingly to continuously optimize the model's prediction ability.

[0015] One of the core modules of the present invention, namely the feature extraction module, includes:

[0016] Split gain calculation mechanism, which uses the core function of the LightGBM algorithm, that is, calculating the split gain of each feature. For each feature and each possible split point, LightGBM calculates the sum of the gradient and second-order derivative of the leaf node after the split, as well as the corresponding loss reduction, to determine the optimal split point.

[0017] Feature sorting mechanism, according to the calculated splitting gain, the feature sorting mechanism sorts all features. The basis for sorting is the contribution of the feature to the model's predictive ability, that is, the size of the splitting gain. Features with high splitting gains are considered to contribute more to the model's predictive ability, so they are ranked first in the sorting.

[0018] Weight assignment mechanism,After sorting, the weight assignment mechanism assigns weight coefficients to each feature.,These weight coefficients reflect the importance of each feature in the,prediction model and are proportional to the split gain of the feature.

[0019] One of the core modules of the present invention, namely the prediction model training module, includes:

[0020] A decision tree construction mechanism that builds multiple decision trees based on the LightGBM algorithm. Each tree is further optimized based on the previous tree to reduce the residual (the difference between the predicted value and the true value). LightGBM uses a greedy algorithm to select the best split point to build each tree.

[0021] Parameter adjustment mechanism, in order to optimize model performance and ensure the accuracy and real-time nature of prediction, the parameter adjustment mechanism dynamically adjusts the training parameters of the LightGBM model (including learning rate, number of trees, tree depth, minimum number of leaf samples, etc.). By adjusting these parameters, the complexity of the model can be controlled, overfitting can be prevented, and the training speed and prediction performance of the model can be improved.

[0022] Gradient boosting mechanism,The prediction model training module also includes a gradient boosting mechanism, which reduces the prediction error by gradually adding new decision trees. Each tree tries to correct the residual of the previous tree, that is, iterative optimization in the gradient direction of the loss function.

[0023] The result generation module comprises:

[0024] a user interface for displaying predictions and health adjustment recommendations to physicians and patients;

[0025] A treatment planning tool that allows doctors to develop more reasonable treatment plans based on predictions and recommendations.

[0026] The feedback optimization module comprises:

[0027] A data comparison mechanism to compare the patient's post-treatment feedback data with the original predicted results;

[0028] A model update mechanism that adds feedback data to the model training set to improve the model's predictive accuracy.

[0029] A method for predicting the success rate of IVF-ET using the above intelligent auxiliary system is characterized in that the method comprises the following steps:

[0030] Step 1: Receive multiple key clinical data of the patient through the data input module;

[0031] Step 2: Preprocess the input clinical data through the data preprocessing module;

[0032] Step 3: Extract features and assign weights to the preprocessed data through the feature extraction module;

[0033] Step 4: Load the prediction model and train it through the prediction model training module;

[0034] Step 5: Output the prediction results of IVF-ET success rate and personalized health adjustment suggestions through the result generation module;

[0035] Step 6: Collect patient feedback data after treatment through the feedback optimization module to train and optimize the prediction model.

[0036] Furthermore, the feature extraction module in step 3 performs feature selection and weight allocation based on the automatic feature splitting mechanism of the LightGBM algorithm.

[0037] Furthermore, the prediction model training module in step 4 utilizes the constructed decision tree sequence to mine the complex nonlinear relationship between features and generate a prediction value of the IVF-ET success rate.

[0038] A computer-readable storage medium storing a computer program, wherein the computer program implements the above method when executed by a processor.

[0039] Compared with the prior art, the present invention has at least the following beneficial technical effects:

[0040] 1. High prediction accuracy: By utilizing the LightGBM algorithm, the system can effectively extract key features from large amounts of clinical data and build accurate prediction models.

[0041] 2. Real-time response: The system can quickly process data and provide prediction results, providing real-time support for clinical decision-making.

[0042] 3. Personalized suggestions: The system not only provides success rate predictions, but also provides personalized health adjustment suggestions based on the patient's specific circumstances to increase the targeted nature of treatment.

[0043] 4. Continuous optimization: Through the feedback mechanism, the system can continuously learn and adapt to new data, continuously improving the accuracy and reliability of predictions.

[0044] This invention combines the LightGBM algorithm with deep learning technology, giving full play to the advantages of both in feature selection, pattern recognition and prediction accuracy, which can maximize the prediction accuracy and computational efficiency, quickly analyze and process massive clinical data, and provide scientific decision support for IVF-ET. Based on the key influencing factors analyzed by the model, personalized health adjustment suggestions are provided to patients to improve the treatment success rate and patient satisfaction. A closed-loop feedback mechanism is established to continuously optimize the model using patient feedback data to ensure that the prediction performance keeps pace with the times. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0046] Figure 1 It is a flow chart of a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following is a detailed description of the embodiments of the present invention: This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given. It should be pointed out that for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

[0048] An embodiment of the present invention provides a method for predicting the performance of scientific computing applications based on deep learning. The method introduces deep learning technology and generative large models. First, the powerful representation ability of the large model is used to replace the complex process of traditional numerical modeling, which can better capture the nonlinear characteristics of performance under different configurations. Secondly, the visual feature learning ability of the generative large model is used to predict performance, so that complex parameter information can be input into the deep learning model in an intuitive and easy-to-process manner. The efficient prediction of the performance of scientific computing applications is achieved through vector encoding methods, which improves the prediction accuracy, reduces the computational overhead, improves the prediction effect, and enhances the generalization ability of the model.

[0049] The following is a specific embodiment to illustrate the specific implementation mode of the present invention:

[0050] Step 1: Data collection and input. The system calls the input module to collect and record the patient's key clinical data. Automatic collection includes but is not limited to multi-dimensional physiological indicators such as blood sugar level, hormone level, and semen quality, which are closely related to the success rate of IVF-ET. The data entry process supports automated verification and format adjustment to ensure data quality.

[0051] Step 2: Data preprocessing: The system preprocesses the patient's n key clinical data, including missing value filling, outlier filtering, and standardization.

[0052] Step 3: Feature extraction and weight assignment: The system calls the feature extraction module to analyze and screen n items of clinical data to determine the importance of each feature in the prediction of IVF-ET success rate.

[0053] Step 4: Prediction model loading and training: The system loads the prediction model based on the LightGBM algorithm and inputs the preprocessed patient data into the model.

[0054] Step 5: Generate prediction results and health adjustment suggestions. After the prediction model completes the calculation, the system outputs the prediction results of IVF-ET success rate and generates personalized health adjustment suggestions.

[0055] Step 6: Optimize result display and feedback. The system displays the prediction results and health adjustment suggestions to doctors and patients through the online platform. Doctors can formulate more reasonable treatment plans based on the results and suggestions. If the patient provides feedback data after treatment, the system will compare the data with the original prediction results and add the feedback data to the model training set, thereby gradually improving the prediction accuracy of the model under different patient conditions.

Claims

1. An intelligent auxiliary system for predicting the success rate of in vitro fertilization-embryo transfer based on the LightGBM algorithm, characterized in that: The system includes: The data input module is used to receive multiple key clinical data of patients, including basic health indicators such as blood sugar level, hormone level, and semen quality; Data preprocessing module, used to fill missing values, filter outliers and standardize the input clinical data; The feature extraction module uses the LightGBM algorithm to automatically calculate the split gain of each feature, sort the features by importance, and assign weight coefficients to each feature, so that in subsequent models, important features will have a greater impact on the prediction results; The prediction model training module builds a prediction model based on the LightGBM algorithm and inputs the preprocessed patient data into the model for training. Through the decision tree sequence, it mines the complex nonlinear relationship between features and generates the prediction value of IVF-ET success rate; The result generation module predicts the clinical data of new patients based on the trained model, outputs the predicted results of IVF-ET success rate, and generates personalized health adjustment suggestions based on the key influencing factors analyzed by the model; The feedback optimization module is used to compare the feedback data of patients after treatment with the original prediction results, analyze the prediction errors, and adjust the model parameters accordingly to achieve continuous optimization and iteration of the model.

2. The predictive intelligent assistance system according to claim 1, characterized in that: The feature extraction module comprises: The split gain calculation mechanism uses the core function of the LightGBM algorithm, which is to calculate the split gain of each feature. For each feature and each possible split point, LightGBM calculates the sum of the gradient and second-order derivative of the leaf node after the split, as well as the corresponding loss reduction, to determine the optimal split point; Feature sorting mechanism, based on the calculated split gain, the feature sorting mechanism sorts all features. The basis for sorting is the contribution of the feature to the model's predictive ability, that is, the size of the split gain; Weight Allocation Mechanism,After sorting, the weight allocation mechanism assigns a weight coefficient to each,feature.

3. The predictive intelligent assistance system according to claim 1, characterized in that: The prediction model training module includes: A parameter adjustment mechanism for dynamically adjusting the training parameters of the LightGBM model to optimize model performance and ensure prediction accuracy and real-time performance.

4. The predictive intelligent assistance system according to claim 1, characterized in that: The result generation module comprises: a user interface for displaying predictions and health adjustment recommendations to physicians and patients; A treatment planning tool that allows doctors to develop more reasonable treatment plans based on predictions and recommendations.

5. The predictive intelligent assistance system according to claim 1, characterized in that: The feedback optimization module comprises: A data comparison mechanism to compare the patient's post-treatment feedback data with the original predicted results; A model update mechanism that adds feedback data to the model training set to improve the model's predictive accuracy.

6. A method for predicting the success rate of IVF-ET using the intelligent assistance system according to any one of claims 1 to 5, characterized in that: The method comprises the following steps: Step 1: Receive multiple key clinical data of the patient through the data input module; Step 2: Preprocess the input clinical data through the data preprocessing module; Step 3: Extract features and assign weights to the preprocessed data through the feature extraction module; Step 4: Load the prediction model and train it through the prediction model training module; Step 5: Output the prediction results of IVF-ET success rate and personalized health adjustment suggestions through the result generation module; Step 6: Collect patient feedback data after treatment through the feedback optimization module to train and optimize the prediction model.

7. The method for predicting IVF-ET success rate according to claim 6, characterized in that: The feature extraction module in step 3 performs feature selection and weight allocation based on the automatic feature splitting mechanism of the LightGBM algorithm.

8. The method for predicting IVF-ET success rate according to claim 6, characterized in that: The prediction model training module in step 4 utilizes the constructed decision tree sequence to mine the complex nonlinear relationship between features and generate a prediction value of the IVF-ET success rate.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 6 to 8 is implemented.

Images