Application method of ferritin combined miRNA in osteoporosis diagnosis
Through machine learning methods and AI optimization technology of ferritin combined with miRNA, an iron-induced PMO diagnostic model was established, which solved the problem of accuracy and low efficiency of early diagnosis of osteoporosis, and achieved more efficient diagnosis and prevention.
Patent Information
- Application Number
- CN202410885281.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-03
- Publication Date
- 2024-11-15
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The treatment of osteoporosis is poor, and the existing diagnostic methods are inaccurate and inefficient, especially in postmenopausal women, it is difficult to detect and prevent osteoporosis early.
Using machine learning method of ferritin combined with miRNA, through training and verification of models, AI robots optimize model parameters, combined with peripheral blood markers for early diagnosis of osteoporosis, establish an iron-induced PMO diagnostic model, and improve the accuracy of diagnosis and efficiency.
It significantly improves the accuracy and efficiency of early diagnosis of osteoporosis, and provides more reliable diagnostic tools to help patients detect and intervene early, thereby reducing the burden of disease and improving quality of life.
Smart Images

Figure CN118969094A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of osteoporosis diagnosis, and in particular relates to an application method of ferritin combined with miRNA in osteoporosis diagnosis. Background Art
[0002] Degeneration of bone microstructure can easily lead to osteoporosis (OP). Postmenopausal osteoporosis (PMO) is more common in postmenopausal women. Currently, the treatment effect of PMO is poor. In the "prevention and treatment" of PMO, clinical workers should move the threshold forward and put prevention before treatment.
[0003] Current status of research on "iron and PMO" at home and abroad
[0004] In recent years, some studies have suggested that PMO is related to iron[5]. Iron metabolism indicators have a certain feasibility as diagnostic indicators of PMO.
[0005] Postmenopausal women often have elevated iron levels. When estrogen drops by 90%, serum ferritin levels increase by 2 to 3 times. In 2020, some scholars found a positive correlation between serum ferritin and the rate of bone loss in a large sample population[9]. It can be seen that serum ferritin, as an indicator of iron metabolism, has a certain diagnostic value in PMO.
[0006] However, in actual work, SF has collinearity with other factors and is difficult to be used alone as a diagnostic indicator for PMO. Other indicators besides SF need to be explored urgently.
[0007] Application of miRNA in PMO diagnostics
[0008] miRNA plays a key role in the regulation of bone metabolism. However, some miRNAs that are potential markers of OP are not common under normal conditions. Therefore, the use of miRNA alone to diagnose PMO has certain limitations.
[0009] Our team previously collected serum from 16 female patients with hip fragility fractures and performed miRNA sequencing, which showed that some miRNAs were associated with both PMO and iron accumulation. As mentioned above, neither SF nor miRNA can be used alone to screen for PMO. If a combined indicator of miRNA and SF is designed, it will have a better application prospect for screening and diagnosing PMO.
[0010] Application of artificial intelligence in PMO diagnosis
[0011] At present, AI research in the medical field is still in its early stages. In clinical work, the recently developed Chat GPT has brought certain challenges to the traditional medical system. Although it cannot currently replace offline medical treatment, the advantage of AI is that the multidisciplinary MDT level is stronger than that of specialists, so it is expected to break the traditional medical model in the future. In basic medical research, machine learning is widely used. Machine learning (ML) is a subfield of AI. Its goal is to allow computers to automatically discover patterns in data, derive disease diagnosis models based on existing big data, and incorporate certain indicators of the subject to be examined into the model to predict their future disease incidence
[25] . Therefore, ML is widely used in disease diagnosis research.
[0012] Feasibility of ML application in combined diagnosis of PMO with SF and miRNAs
[0013] There have been reports on the application of ML in the diagnosis of PMO. In clinical transformation, SF indicators can be used for PMO screening. Designing a joint indicator of miRNA and SF has a better application prospect for the diagnosis of PMO. PMO diagnosis is a discriminant analysis, so supervised learning in ML is more applicable. Applying ML technology to the joint diagnosis of SF and miRNAs for PMO has a high feasibility.
[0014] Through the above analysis, the problems and defects of the prior art are as follows:
[0015] Currently, the treatment effect of PMO is not good. In the "prevention and treatment" of PMO, clinical workers should move the threshold forward and put prevention before treatment. Summary of the invention
[0016] In view of the problems existing in the prior art, the present invention provides an application method of ferritin combined with miRNA in the diagnosis of osteoporosis.
[0017] The present invention is achieved by: a method for applying ferritin combined with miRNA in the diagnosis of osteoporosis comprising:
[0018] Step 1: Train the existing data to obtain a preliminary model and verify the model;
[0019] Step 2: Expand the sample and adjust the ML algorithm to obtain a better model;
[0020] In step 3, the last collected samples are used as test sets to evaluate the model quality, and the AI robot is used to continuously optimize the model parameters and update the PMO diagnostic model in real time.
[0021] Another object of the present invention is to provide an application system of ferritin combined with miRNA in the diagnosis of osteoporosis, comprising:
[0022] The verification module is used to train the existing data to obtain a preliminary model and verify the model;
[0023] The adjustment module is used to expand the sample and adjust the ML algorithm to obtain a better model;
[0024] The optimization module is used to collect samples as test sets to evaluate model quality, and uses AI robots to continuously optimize model parameters to update the PMO diagnostic model in real time.
[0025] Another object of the present invention is to provide a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for applying ferritin combined with miRNA in the diagnosis of osteoporosis.
[0026] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the method for applying ferritin combined with miRNA in the diagnosis of osteoporosis.
[0027] Another object of the present invention is to provide an information data processing terminal, which is used to implement the application system of ferritin combined with miRNA in the diagnosis of osteoporosis.
[0028] Another object of the present invention is to provide a method for diagnosing osteoporosis based on ferritin and miRNA, comprising the following steps:
[0029] (1) Collect clinical data of osteoporosis patients, including ferritin levels, miRNA expression profiles, and other relevant physiological indicators, and use machine learning algorithms to train a preliminary diagnostic model;
[0030] (2) Validate the preliminary diagnostic model and adjust the model parameters based on the validation results to optimize model performance;
[0031] (3) Continue to collect new data and expand the sample size, and use the new data to further train and adjust the optimized model to obtain a better model;
[0032] (4) The last collected samples were used as the test set to evaluate the model quality, and AI technology was used to continuously optimize the model parameters to achieve real-time updating of the osteoporosis diagnosis model.
[0033] Furthermore, the method also includes: using an AI robot to adjust and optimize the model parameters in real time to ensure the accuracy and timeliness of the model, and performing necessary updates and maintenance on the model based on newly collected data to adapt to the latest needs of osteoporosis diagnosis.
[0034] Another object of the present invention is to provide a method for constructing an iron-induced PMO diagnostic model, characterized in that it comprises the following steps:
[0035] Peripheral blood samples were collected from postmenopausal women with hip fracture and subjected to miRNA sequencing;
[0036] Based on the sequencing results, miRNAs related to iron accumulation and PMO were screened;
[0037] The screened miRNA, SF values, and known PMO risk factors (such as age and BMI) were used as feature variables to train a preliminary PMO diagnostic model;
[0038] Expand the sample size and adjust the machine learning algorithm to obtain a better model;
[0039] Use external test sets to evaluate model quality and use AI robots to optimize model parameters in real time.
[0040] Another object of the present invention is to provide a method for evaluating the performance of an iron-induced PMO diagnostic model, characterized in that it comprises the following steps:
[0041] Collect external test set data;
[0042] Use the cross matrix to calculate the model's precision, recall, and F1 value;
[0043] The quality of the model is evaluated based on the calculation results, and the AI robot is used to optimize the model parameters in real time.
[0044] Another object of the present invention is to provide a method for constructing a PMO diagnostic model using machine learning technology, which is characterized by comprising the following steps:
[0045] Collect peripheral blood samples from postmenopausal women and obtain relevant clinical data;
[0046] Perform miRNA sequencing on the samples and screen out miRNAs associated with PMO;
[0047] The screened miRNA, SF values, and known high-risk factors for PMO were used as characteristic variables;
[0048] Depending on the sample size, select an appropriate machine learning algorithm (such as linear regression, logistic regression, decision tree, SVM, random forest, KNN) for model training;
[0049] Use external test sets to evaluate model performance and perform real-time optimization of model parameters.
[0050] Another object of the present invention is to provide a method for optimizing PMO diagnostic model parameters based on an AI robot, characterized by comprising the following steps:
[0051] Import the initially constructed PMO diagnostic model into the AI robot;
[0052] Use external test set data to test the model;
[0053] Based on the test results, the reinforcement learning technology of the AI robot is used to optimize the model parameters in real time;
[0054] Output the optimized PMO diagnostic model.
[0055] Another object of the present invention is to provide a method for analyzing the correlation between biomarkers in peripheral blood and BMD, characterized by comprising the following steps:
[0056] Collect peripheral blood samples from postmenopausal women and obtain relevant clinical data;
[0057] Detection of biomarkers (such as SF, miRNA) in samples;
[0058] Statistical methods were used to analyze the correlation between biomarkers and BMD;
[0059] Based on the analysis results, biomarkers associated with BMD were screened out and used to construct a PMO diagnostic model.
[0060] Another object of the present invention is to provide a method for validating a PMO diagnostic model using a machine learning algorithm, characterized by comprising the following steps:
[0061] Peripheral blood samples and relevant clinical data were collected from postmenopausal women;
[0062] Perform miRNA sequencing on the samples and screen out miRNAs associated with PMO;
[0063] The screened miRNA, SF values, and known high-risk factors for PMO were used as characteristic variables;
[0064] Select an appropriate machine learning algorithm for model training and use the validation set to validate the model;
[0065] Based on the validation results, the diagnostic performance of the model is evaluated, and necessary adjustments and optimizations are made.
[0066] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:
[0067] First, the present invention uses AI technology and AI robots (the model is derived from ChatGPT) to empower the PMO diagnostic model, so that it can learn autonomously, continuously use new data to complete self-update, and feedback better model formulas. Therefore, research technology is one of the innovations of the present invention.
[0068] The present invention performs miRNAseq analysis on a large sample of peripheral blood specimens of postmenopausal women, and establishes an iron-induced PMO model in combination with the SF index. It has a high degree of originality in the field of PMO joint diagnosis, and the research results have an important guiding role in the screening of iron-induced PMO. Therefore, the research idea is one of the innovations of the present invention.
[0069] Compared with other projects that select healthy people as samples, the samples selected in this invention come from physical examination people and fracture people. It includes both healthy people and patients with traumatic stress. The obtained PMO diagnostic model is more generalized, and the population sources of the training and validation sets are different from those of the test set. External testing makes the model more universal in clinical testing. Therefore, sample selection is one of the innovations of this invention.
[0070] The present invention applies ML technology to the establishment of an early diagnosis model for PMO, using multiple methods such as linear / logistic regression, CART, SVM (linear kernel function / Gaussian kernel function), random forest, KNN, etc. Therefore, methodology is one of the innovations of the present invention.
[0071] Second, in the current field of medical diagnosis, there are significant challenges in the early diagnosis of postmenopausal osteoporosis (PMO). Existing diagnostic methods mostly rely on bone mineral density (BMD) measurements, but this method has limitations in predicting fracture risk, especially for individuals with reduced bone mass but who have not yet reached the osteoporosis standard. In addition, the traditional diagnostic process is time-consuming and costly, and the understanding of the complex relationship between biomarkers and PMO is still insufficient, resulting in the need to improve the diagnostic accuracy.
[0072] In view of the above problems, the present invention proposes a new method for constructing and optimizing the diagnostic model of iron-induced PMO. The method aims to improve the accuracy and efficiency of PMO diagnosis by integrating machine learning technology, AI robot optimization and in-depth analysis of specific biomarkers in peripheral blood. In particular, the present invention focuses on iron accumulation, an important cause of PMO, and provides a new perspective and tool for the early diagnosis of PMO by exploring biomarkers related to iron accumulation.
[0073] First, by introducing machine learning algorithms, the present invention realizes the rapid processing and analysis of a large amount of sample data, significantly improving the efficiency and accuracy of diagnostic model construction. Secondly, by using AI robots to optimize model parameters in real time, the diagnostic model can continuously adapt to new data changes, maintaining the stability and cutting-edge nature of diagnostic performance. In addition, the present invention also deeply analyzes the correlation between biomarkers and BMD in peripheral blood, providing new insights into the pathogenesis of PMO and providing strong support for further optimization of diagnostic models.
[0074] The proposal and application of the present invention not only solves many problems in the existing PMO diagnostic methods, but also significantly improves the accuracy and efficiency of diagnosis. The promotion and application of this method is expected to provide clinicians with more reliable diagnostic tools, helping patients to detect and intervene in PMO earlier, thereby reducing the burden of the disease and improving the quality of life. At the same time, the successful practice of the present invention also provides new ideas and directions for the early diagnosis of other complex diseases, and has broad application prospects and far-reaching social significance. BRIEF DESCRIPTION OF THE DRAWINGS
[0075] Figure 1 This is a flow chart of the application method of ferritin combined with miRNA in the diagnosis of osteoporosis provided by an embodiment of the present invention.
[0076] Figure 2 It is a structural block diagram of the application system of ferritin combined with miRNA in the diagnosis of osteoporosis provided by an embodiment of the present invention.
[0077] Figure 3 This is a flow chart of the application of ML in SF combined with miRNA in the diagnosis of iron-induced PMO provided by an embodiment of the present invention.
[0078] Figure 4 It is an experimental design roadmap provided by an embodiment of the present invention.
[0079] Figure 5 The present invention provides serum miRNAs sequencing of female patients with fragility hip fractures; AB, heat map and volcano map of differential miRNAs expression between low bone mass group (LB) and high bone mass group (HB); C, kernel density map; D, heat map of differential miRNAs expression between low iron group (LI) and high iron group (HI). DETAILED DESCRIPTION
[0080] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0081] The following are two specific examples of a method for using ferritin combined with miRNA in the diagnosis of osteoporosis (PMO) provided in an embodiment of the present invention:
[0082] Embodiment 1
[0083] 1. Data Collection and Preprocessing
[0084] A large number of clinical data of patients with confirmed osteoporosis are collected, including their ferritin levels, miRNA expression profiles, and other related physiological indicators. At the same time, the corresponding data of healthy people are also collected as controls. These data will be divided into training sets and validation sets.
[0085] 2. Model training and verification (S101)
[0086] Using the training set data, a preliminary diagnostic model was obtained through training with a machine learning (ML) algorithm. The model was able to predict the risk of osteoporosis based on the patient's ferritin level and miRNA expression profile. The model was then validated using the validation set data to evaluate its predictive accuracy and reliability.
[0087] 3. Model optimization (S102)
[0088] In the later stage, we will continue to collect more data from osteoporosis patients and healthy people to expand the sample size. At the same time, we will adjust and optimize the ML algorithm based on the newly collected data to improve the predictive performance of the model. This involves adjusting the parameters of the model, selecting different features, or trying different ML algorithms.
[0089] 4. Model testing and real-time updating (S103)
[0090] The last collected samples are used as test sets to evaluate the quality of the optimized model. AI robots are used to continuously optimize model parameters based on newly collected data to ensure the accuracy and timeliness of the PMO diagnostic model. At the same time, the model will be updated in real time according to actual conditions to meet the latest needs of osteoporosis diagnosis.
[0091] Embodiment 2
[0092] 1. Preliminary model construction
[0093] Based on the existing data of ferritin, miRNA and related physiological indicators of osteoporosis patients, a preliminary diagnostic model was constructed using appropriate ML algorithms (such as support vector machine, random forest, etc.). The model aims to predict the risk of osteoporosis of patients through their biomarker data.
[0094] 2. Model verification and adjustment
[0095] After the preliminary model is built, the model is validated using independent validation set data. Based on the validation results, the model parameters and algorithms are adjusted to improve its prediction accuracy and generalization ability.
[0096] 3. Expand samples and optimize models
[0097] Over time, we will continue to collect and integrate more patient data, including samples of different age groups, genders, and disease severity. We will use this new data to further train and optimize the model to improve its applicability to different populations.
[0098] 4. Real-time model update and maintenance
[0099] To ensure the accuracy and timeliness of the diagnostic model, AI robots are used to continuously monitor newly collected data and adjust the model parameters in real time based on this data. In addition, the performance of the model is regularly evaluated and necessary updates and maintenance are performed as needed to ensure that it always reflects the latest knowledge and technology in the diagnosis of osteoporosis.
[0100] These two examples demonstrate how to apply ferritin combined with miRNA in the diagnosis of osteoporosis and improve the accuracy and reliability of diagnosis through continuous data collection, model training and optimization.
[0101] like Figure 1 As shown, a method for applying ferritin combined with miRNA in the diagnosis of osteoporosis provided by an embodiment of the present invention comprises the following steps:
[0102] S101, train the existing data to obtain a preliminary model and verify the model;
[0103] S102, later expand the sample and adjust the ML algorithm to obtain a better model;
[0104] S103, the samples finally collected are used as test sets to evaluate the model quality, and the AI robot is used to continuously optimize the model parameters and update the PMO diagnosis model in real time.
[0105] The present invention provides a method for applying ferritin combined with miRNA in the diagnosis of osteoporosis, which mainly optimizes model parameters through machine learning (ML) algorithms and AI robots to improve diagnostic accuracy. The specific working principle is as follows:
[0106] 1. Preliminary model training and verification
[0107] Step S101: train the existing data to obtain a preliminary model, and verify the model.
[0108] Data preparation: Collect and organize existing ferritin and miRNA data, including sample data from osteoporosis patients and healthy people.
[0109] Model training: Use machine learning algorithms (such as logistic regression, support vector machine, random forest, etc.) to input ferritin and miRNA data into the algorithm to train a preliminary diagnostic model.
[0110] Model validation: Evaluate model performance through cross-validation methods to ensure that the preliminary model has a certain diagnostic accuracy and stability. Validation indicators can include accuracy, sensitivity, specificity, etc.
[0111] 2. Sample expansion and model optimization
[0112] Step S102: Expand the sample later and adjust the ML algorithm to obtain a better model.
[0113] Sample expansion: Over time, more ferritin and miRNA sample data are collected to increase the training sample size of the model.
[0114] Algorithm adjustment: Based on the expanded sample data, readjust and optimize the machine learning algorithm to improve the performance of the model. You can try different feature selection methods and algorithm parameter tuning to find the best model configuration.
[0115] Model retraining: Use new sample data and optimized algorithms to retrain the model to obtain a better diagnostic model.
[0116] 3. Test set evaluation and real-time optimization
[0117] Step S103: The samples collected at last are used as test sets to evaluate the model quality, and the AI robot is used to continuously optimize the model parameters and update the PMO diagnostic model in real time.
[0118] Model evaluation: The last collected sample data is used as a test set to evaluate the optimized model. The performance of the model in actual applications is evaluated through the test set data to further verify the accuracy and stability of the model.
[0119] AI robot optimization: Introduce AI robots to continuously monitor and optimize the model. AI robots can collect new data in real time and use online learning technology to continuously update model parameters to keep the model in the optimal state.
[0120] Real-time update: Based on the optimization results fed back by the AI robot, the PMO (osteoporosis) diagnostic model is updated in real time to ensure that the model can adapt to data changes and clinical needs and improve the accuracy and reliability of diagnosis.
[0121] Through the above steps, the present invention realizes the application method of ferritin combined with miRNA in the diagnosis of osteoporosis. By using machine learning algorithms and AI robots, the model parameters are continuously optimized and the diagnostic model is updated, the accuracy and stability of the model are improved, and the effectiveness in clinical applications is ensured. This method not only improves the early diagnosis rate of osteoporosis, but also provides new technical support for personalized medicine.
[0122] like Figure 2 As shown, an application system of ferritin combined with miRNA in the diagnosis of osteoporosis provided by an embodiment of the present invention includes:
[0123] The verification module is used to train the existing data to obtain a preliminary model and verify the model;
[0124] The adjustment module is used to expand the sample and adjust the ML algorithm to obtain a better model;
[0125] The optimization module is used to collect samples as test sets to evaluate model quality, and uses AI robots to continuously optimize model parameters to update the PMO diagnostic model in real time.
[0126] Another object of the present invention is to provide a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method for applying ferritin combined with miRNA in the diagnosis of osteoporosis.
[0127] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the method for applying ferritin combined with miRNA in the diagnosis of osteoporosis.
[0128] Another object of the present invention is to provide an information data processing terminal, which is used to implement the application system of ferritin combined with miRNA in the diagnosis of osteoporosis.
[0129] The present invention selects different supervised learning methods according to the size of the collected samples and the results of miRNAseq. As the number of collected samples increases, the ML algorithm is continuously adjusted, and the last 20% of the samples are used as the test set to detect the final model; the final model is subjected to reinforcement learning, the model formula is input into the AI robot, and the test set data is continuously input, and reward and punishment instructions are set to obtain a new model with AI feedback ( Figure 3 ).
[0130] Figure 3The process design diagram of ML applied to SF combined with miRNA for the diagnosis of iron-induced PMO. Different supervised learning algorithms are adjusted according to the number of samples collected, and clinical tests are performed. The final iron-induced PMO diagnostic model obtained after the test is subjected to reinforcement learning, and the parameters are continuously adjusted by AI robots to update the diagnostic model in real time.
[0131] The present invention adopts AI technology and aims to develop a diagnostic model for iron-induced PMO. It will: 1. Train the existing data to obtain a preliminary model and verify the model; 2. Expand the sample later and adjust the ML algorithm to obtain a better model; 3. Finally, use the collected samples as a test set to evaluate the quality of the model, and use AI robots to continuously optimize the model parameters and update the PMO diagnostic model in real time. The present invention is expected to use peripheral blood markers to obtain an early diagnostic model for iron-induced PMO, which is convenient for conducting PMO screening in secondary hospitals or community hospitals, and timely preventing the occurrence and development of osteoporosis and even brittle fractures in elderly women.
[0132] (1) Training existing small sample data to obtain a preliminary PMO diagnostic model
[0133] Our team has collected peripheral blood from 16 postmenopausal women with hip fractures and plans to conduct the following tests on the existing small samples:
[0134] 1) Train the original model based on the existing fracture patient data;
[0135] After sequencing samples from 16 postmenopausal women with hip fractures (training sample number m), 25 miRNAs with intersections of iron accumulation and PMO were obtained. These miRNAs were verified by PCR in peripheral blood. The verified miRNAs and SF values were used as feature attributes, combined with known high-risk factors for PMO (age, body mass index BMI), with a total of n≈28 feature variables. Since m is small and n is large, the amount of training set data is not enough to support the training of a complex nonlinear model. Therefore, the ML algorithm uses linear regression or support vector machine (SVM) without kernel function to obtain a preliminary PMO diagnostic model.
[0136] 2) Perform miRNASeq on peripheral blood of the physical examination population and set up a validation set;
[0137] 36 cases of physical examination had no fractures, and the SF value excluded the interference of traumatic stress, which can better reflect the correlation between iron metabolism and bone density. Therefore, the miRNASeq results of the physical examination population have more diagnostic value. In order to further dilute the impact of fractures, 16 cases of fractures were divided into 12 training groups + 4 validation groups, and 36 cases of physical examinations were divided into 27 training groups + 9 validation groups. m is still small, and the ML algorithm is not changed for the time being. The 13 validation groups used the repeated Kfold method with shuffled data to monitor whether the model is overfitting or underfitting, and to adjust the model hyperparameters accordingly.
[0138] Here, BMD is a continuous variable and linear regression is used. If the model fit is good, the applicant intends to classify the dependent variable into two categories according to the T value (T≤2.5 is PMO, T>2.5 is non-PMO) to construct a logistic regression, and calculate the corresponding prediction probability for each object, perform ROC analysis with the prediction probability (SF and miRNA joint diagnostic index), calculate the area under the curve AUC, and evaluate the modeling quality.
[0139] (2) Expand the sample and adjust the ML method to obtain a better model
[0140] During the research period, the team plans to continue to collect samples from physical examinations and hospitalized patients, and divide them into training sets and validation sets (ratio 3:1). The collection indicators still include peripheral blood SF, age, and BMI. Serum miRNASeq is performed and verified. Based on the expanded sample size, the following studies are planned:
[0141] 1) Decision tree or support vector machine modeling is used for medium sample size;
[0142] When the sample size m≈100, it can be considered as a medium sample size. We collected 40 postmenopausal female hospitalized patients (30 training sets + 10 validation sets) and 60 postmenopausal women for physical examination (45 training sets + 15 validation sets). We screened the intersection value of iron accumulation and PMO based on miRNASeq, and used SF, age, and BMI as features, and used decision trees and SVM models, respectively.
[0143] 2) Use random forest or K nearest neighbor modeling for large samples
[0144] Age, BMI and SF are not collinear, and PCA dimensionality reduction cannot be performed on the three, so there is no room for n to be reduced. When m is expanded to >300, m>10×n, and m can be considered a large sample. All sample detection indicators are the same as before. 100 hospitalized women (75 training sets + 25 validation sets) and 200 physical examination women (150 training sets + 50 validation sets) were modeled using random forest and K nearest neighbor (KNN) respectively.
[0145] (3) Check the quality of the obtained model and use AI robots to optimize model parameters
[0146] According to the plan, 60 samples of participants from other hospitals were collected until the total number of samples reached 300, and 20% (60 samples) collected at this time were used as external test sets. All sample detection indicators were the same as before. All the above modeling included linear regression, logistic regression, CART, SVM, random forest, and KNN. The test method used a cross matrix to evaluate the model.
[0147] According to the test results, the best PMO diagnostic model is obtained. The model formula is imported into the AI robot, and the model is optimized using the ML reinforcement learning method. The test set data is input one by one, and the AI is required to feedback the diagnostic results based on the input data. At the same time, reward and punishment instructions are input to enable the AI to continuously adjust the model parameters, thereby completing the real-time update of the iron-induced PMO diagnostic model.
[0148] (1) To analyze the correlation between biomarkers SF, miRNA and BMD in peripheral blood of postmenopausal women;
[0149] (2) Use different machine learning algorithms to establish and validate models for diagnosing PMO using SF, miRNA, and other characteristic variables;
[0150] (3) Obtain the best model of SF combined with miRNA for early diagnosis of PMO after external clinical testing;
[0151] (4) Train AI robots to learn autonomously, allowing them to optimize model parameters in real time.
[0152] Key scientific issues to be addressed
[0153] (1) Whether SF and miRNA are univariately / multivariately correlated with BMD;
[0154] (2) whether different machine learning algorithms are applicable to the modeling process of SF and miRNA diagnosis of PMO;
[0155] (3) Whether the final model can pass external clinical testing;
[0156] (4) Can AI robots use reinforcement learning technology to optimize model parameters?
[0157] The present invention explores the feasibility of SF combined with miRNA in diagnosing PMO in hospitalized fracture patients and healthy people undergoing physical examinations, in order to obtain a model that can be generalized and has strong universality. The specific research plan is as follows:
[0158] (1) Population selection and sample collection
[0159] 1) Crowd selection
[0160] 2) Sample collection.
[0161] (2) Application of ML technology in small sample data
[0162] ML was implemented using Python 3.11.2 software, and the data were pre-listed in an excel sheet and imported.
[0163] 1) Linear regression
[0164] 2) Logistic Regression
[0165] (3) Application of ML techniques in medium sample size data
[0166] 1) Decision Tree
[0167] 2) SVM with Gaussian kernel function
[0168] (4) Application of ML technology in large sample data
[0169] 1) Random Forest
[0170] 2) K nearest neighbors
[0171] (5) PMO diagnostic model performance test
[0172] The external test set data is used to test the model. The testing method uses a cross matrix to calculate the precision, recall and F1 value of each model to evaluate the quality of each model.
[0173] The final model needs to be optimized, and AI robots are used for reinforcement learning to optimize the model. AI uses ChatGPT3.5 as a model. Domestic companies use OpenAI's API service to implement interactive functions with ChatGPT (https: / / baike.z6j.cn / ). The charges are more reasonable than OpenAI, and the language is also easy to be accepted by secondary and community hospitals. The use plan of AI robots is as follows:
[0174] The data of the characteristic variables of one examinee in the test set is input into the AI robot, and an instruction (Prompt) is given, requiring the AI to feedback whether the examinee is diagnosed with PMO based on the imported model formula. At this time, the reward and punishment instructions are input. If the AI feedback result does not match the actual PMO diagnosis result of the examinee, the AI robot is punished (Prompt = "wrong") and required to include the examinee's data, adjust the parameters based on the imported model, and feedback the latest model; if the PMO diagnosis result is consistent, the AI robot is rewarded (Prompt = "very good"), and the parameter adjustment method is the same as before, so as to complete the continuous updating of the PMO diagnosis model.
[0175] Experimental design route Figure 4 ;
[0176] Our team previously collected serum from 16 female hospitalized patients with hip fragility fractures, performed miRNA sequencing in parallel, and divided them into high bone mass and low bone mass groups according to the median bone density. Cluster analysis verified Blondal's conclusion. Our team also divided them into groups based on serum ferritin and found that patients with higher serum ferritin had differentially expressed miRNAs. The differential RNAs of the two groups were analyzed using Wayne intersection to screen out 25 differentially expressed miRNAs. Tip: Some miRNAs are associated with both PMO and iron accumulation. As mentioned earlier, neither SF nor miRNA can be used alone for the screening of PMO. If a combined indicator of miRNA and SF is designed, it will have a better application prospect for screening and diagnosis of PMO. Figure 5 Serum miRNAs sequencing in female patients with fragility hip fractures. AB. Heat map and volcano map of differential miRNAs expression between low bone mass group (LB) and high bone mass group (HB); C. Kernel density map; D. Heat map of differential miRNAs expression between low iron group (LI) and high iron group (HI).
[0177] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. A person of ordinary skill in the art will appreciate that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVDROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.
[0178] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.
Claims
1. A method for constructing and optimizing an iron-induced PMO diagnostic model, characterized in that: The following steps are involved: Collect and pre-process samples; Use machine learning technology to build a preliminary PMO diagnostic model; Expand the sample size and adjust the machine learning algorithm to obtain a better model; Use external test sets to evaluate model quality; Use AI robots to optimize model parameters in real time.
2. The method according to claim 1, characterized in that The steps of collecting samples and preprocessing them include: Peripheral blood samples were collected from postmenopausal women with hip fracture and subjected to miRNA sequencing; Based on the sequencing results, miRNAs related to iron accumulation and PMO were screened; The screened miRNA, SF values and known high-risk factors for PMO were used as characteristic variables.
3. The method according to claim 1, characterized in that The steps to build a preliminary PMO diagnostic model using machine learning technology include: According to the sample size, select the appropriate machine learning algorithm for model training; The model is trained using the training set data and validated using the validation set.
4. The method according to claim 1, characterized in that The steps to evaluate model quality using an external test set include: Collect external test set data; Use the cross matrix to calculate the model's precision, recall, and F1 value; Evaluate the quality of the model based on the calculation results.
5. The method according to claim 1, wherein the step of using an AI robot to perform real-time optimization of model parameters specifically comprises: Import the initially constructed PMO diagnostic model into the AI robot; Use external test set data to test the model; Based on the test results, the reinforcement learning technology of the AI robot is used to optimize the model parameters in real time; Output the optimized PMO diagnostic model.
6. The method according to any one of claims 1 to 5, characterized in that It also includes steps to analyze the correlation between biomarkers and BMD in peripheral blood: Testing samples for biomarkers; Statistical methods were used to analyze the correlation between biomarkers and BMD; Based on the analysis results, biomarkers related to BMD were screened out and used to construct or optimize the PMO diagnostic model.
7. An application system of ferritin combined with miRNA in the diagnosis of osteoporosis for implementing the application method as claimed in claim 1, characterized in that: Application systems include: The verification module is used to train the existing data to obtain a preliminary model and verify the model; The adjustment module is used to expand the sample and adjust the ML algorithm to obtain a better model; The optimization module is used to collect samples as test sets to evaluate model quality, and uses AI robots to continuously optimize model parameters to update the PMO diagnostic model in real time.
8. A method for diagnosing osteoporosis based on ferritin and miRNA, characterized in that: The following steps are involved: (1) Collect clinical data of osteoporosis patients, including ferritin levels, miRNA expression profiles, and other relevant physiological indicators, and use machine learning algorithms to train a preliminary diagnostic model; (2) Validate the preliminary diagnostic model and adjust the model parameters based on the validation results to optimize model performance; (3) Continue to collect new data and expand the sample size, and use the new data to further train and adjust the optimized model to obtain a better model; (4) The last collected samples were used as the test set to evaluate the model quality, and AI technology was used to continuously optimize the model parameters to achieve real-time updating of the osteoporosis diagnosis model.
9. The osteoporosis diagnosis method according to claim 8, characterized in that: The method also includes: using an AI robot to adjust and optimize model parameters in real time to ensure the accuracy and timeliness of the model, and performing necessary updates and maintenance on the model based on newly collected data to adapt to the latest needs of osteoporosis diagnosis.
Citation Information
Patent Citations
Application of serum miRNAs as postmenopausal osteoporosis combined diagnosis marker
CN114959014A
Method for constructing interpretable breast cancer blood miRNA diagnosis system
CN116168826A
Hypertension early warning method based on pre-training and reinforcement learning
CN116403712A
Personalized patient treatment suggestion system based on large language model
CN118262933A
Cited By
Intelligent osteoporosis auxiliary diagnosis system and method based on multi-mode serum marker
CN120613121A
