ASD development trajectory prediction method based on PERS and interval censored data
By combining a multi-environmental risk scoring model and a Bayesian regression model with censored data, the problems of multiple environmental risk factors and data censoring in the prediction of ASD developmental trajectory were solved, achieving more accurate prediction of developmental processes and supporting early screening and personalized intervention.
Patent Information
- Application Number
- CN202511588300.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-03
- Publication Date
- 2026-02-27
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing technologies fail to effectively consider various environmental risk factors and data censoring issues in predicting ASD developmental trajectories, resulting in insufficient prediction accuracy and an inability to provide reliable evidence for early screening and precise intervention.
By combining a multi-environmental risk scoring model with a Bayesian regression model based on interval censoring data, high-confidence environmental risk factors are extracted from the Simons Autism Database. After data preprocessing and correction, the Bayesian regression model is used to predict milestone events and their probabilities in the future.
It significantly improves the accuracy of predicting the timing and probability of future milestone events, providing a scientific basis for early screening and personalized intervention, and ensuring the credibility and practicality of the prediction results.
Smart Images
Figure CN121583518A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of ASD development trajectory prediction methods, in particular to an ASD development trajectory prediction method based on PERS and interval censored data. BACKGROUND
[0002] Research on autism spectrum disorder (ASD) has received extensive attention in recent years. Many research teams are committed to finding methods that can effectively predict the development trajectory of ASD patients in order to achieve early screening and precise intervention. In terms of environmental risk factors, some studies have pointed out that various factors in the environment may be related to the occurrence and development of ASD. However, there is currently no multi-environmental risk score model containing 20 environmental risks for ASD development trajectory prediction. In terms of development milestone event prediction, traditional methods often do not fully consider the interval censored characteristics of the data. For example, some studies simply analyze based on complete data, ignoring the fact that many data are observed within a certain interval in practice.
[0003] Bayesian regression models have been widely used in other fields such as economics and biology. They can use prior information and sample information to estimate model parameters, and have good flexibility and adaptability. However, their application in predicting the occurrence time of ASD patient development milestones is still in the exploratory stage.
[0004] Current research on ASD development trajectory mainly focuses on the analysis of the influence of a single environmental factor or a few environmental factors. When constructing prediction models, traditional statistical methods such as linear regression and logistic regression are usually used. These methods have limited ability to describe complex environmental risk factors and development processes.
[0005] When dealing with development milestone event data, most studies do not consider the interval censored problem of the data. For example, when recording the occurrence time of a patient's development milestone event (such as language development, social development, etc.), due to practical observation limitations, we can only know that the event occurred within a certain time period, rather than the exact time point. Traditional methods directly ignore this interval censored information, which may cause model bias and inaccuracy. SUMMARY
[0006] The technical problem to be solved by the present application is how to improve the prediction accuracy of the occurrence time and probability of milestone events at future time points in the development process. To overcome the defects of the above prior art (or related technology), the present application provides an ASD development trajectory prediction method based on PERS and interval censored data. The present application provides an ASD development trajectory prediction method based on PERS and interval censored data, comprising: Step S1, for a plurality of ASD patients, obtaining the environmental exposure record and the existing milestone event record in the development process corresponding to each ASD patient, and extracting a plurality of environmental risk factors and covariants related to each environmental risk factor in the environmental exposure record; Step S2, pre-processing each environmental risk factor to obtain a pre-processed environmental risk factor; Step S3, inputting each pre-processed environmental risk factor into a pre-constructed multi-environmental risk scoring model to analyze and obtain an environmental risk score; Step S4, correcting each covariant using the icenReg package in R language to obtain a corrected covariant; Step S5, inputting the environmental risk score, the corrected covariant and the existing milestone event record into a pre-constructed ASD development trajectory prediction model based on interval censored data to predict a milestone event and a probability of occurrence at a future time.
[0007] Compared with the prior art, the ASD development trajectory prediction method based on PERS and interval censored data has the following advantages: In the present application, the multi-environmental risk scoring model is combined with the ASD development trajectory prediction model based on interval censored data. The multi-environmental risk scoring model integrates a plurality of environmental risk factors, avoiding the limitations of single factor analysis. The processing of interval censored data makes full use of the inaccurate observation data in the real world, avoiding information waste and bias. The combination of the two from the two dimensions of "input data breadth" and "data processing depth" significantly improves the accuracy and reliability of predicting the occurrence time and probability of a milestone event at a future time. The final output is the milestone event and the probability of occurrence at a future time. Such quantitative and time-dimension prediction results can provide valuable scientific basis for doctors and parents, making early screening and personalized intervention program possible.
[0008] In a possible implementation, in step S1, the environmental exposure record and the existing milestone event record of a plurality of ASD patients aged 4 to 18 years old are extracted from the Simons Foundation Autism Research Initiative (SFARI) database.
[0009] Compared with the prior art, the above technical solution can use the internationally recognized, large and high-quality ASD professional database of the Simons Foundation Autism Research Initiative (SFARI) as a data source, ensuring the authenticity, standardization and authority of the data on which the present application is based, thereby ensuring the credibility and persuasiveness of the prediction results predicted thereby.
[0010] In a possible implementation, in the step S1, family history, maternal disease during pregnancy, drug use records and perinatal records of each ASD patient are obtained as the environmental exposure records.
[0011] Compared with the prior art, the types of the obtained environmental exposure records are specifically listed, which are all key factors highly related to ASD development confirmed by existing research, which indicates that the multi-environmental risk scoring model is constructed based on comprehensive and reliable environmental risk factors, rather than randomly selected, further ensuring the scientificity and effectiveness of the multi-environmental risk scoring model.
[0012] In a possible implementation, in the step S2, for each environmental risk factor, the environmental risk factor is sequentially subjected to missing value filling processing, abnormal value correction processing and variable uniform coding processing to obtain the corresponding preprocessed environmental risk factor.
[0013] Compared with the prior art, the above technical solution can effectively solve the problems of incomplete data, errors or inconsistent formats commonly existing in real clinical data, and provide high-quality and standardized input data for the multi-environmental risk scoring model.
[0014] In a possible implementation, in the step S3, a first risk score is obtained by respectively scoring and accumulating each preprocessed environmental risk factor according to the credibility level of the preprocessed environmental risk factor, a second risk score is obtained by weighted accumulation of each preprocessed environmental risk factor according to the effect size of the preprocessed environmental risk factor, and a third risk score is obtained by weighted accumulation of the score of each preprocessed environmental risk factor according to the effect size of the preprocessed environmental risk factor, and the optimal risk score is selected from the first risk score, the second risk score and the third risk score as the environmental risk score.
[0015] Compared with the prior art, the above technical solution can provide three different scoring construction strategies including simple accumulation, effect size weighting and comprehensive weighting, and determine the “optimal risk score” through a screening mechanism, which can adaptively select the most suitable scoring method for the current data, ensure the best prediction effect at all times, and evaluate the risk from different angles through different scoring methods, and comprehensive screening ensures the scientificity and rigor of the scoring system.
[0016] In one possible implementation, in step S3, each of the pretreated environmental risk factors is scored using a standard of 6 points, 4 points, 2 points, and 1 point. Pretreated environmental risk factors that are positively correlated with ASD are assigned positive scores, and pretreated environmental risk factors that are negatively correlated with ASD are assigned negative scores. Then, the scores of each of the pretreated environmental risk factors are accumulated to obtain the first risk score.
[0017] In one possible implementation, in step S3, each of the pretreated environmental risk factors is scored using a standard of 4 points, 3 points, 2 points, and 1 point. Pretreated environmental risk factors that are positively correlated with ASD are assigned positive scores, and pretreated environmental risk factors that are negatively correlated with ASD are assigned negative scores. Then, the scores of each of the pretreated environmental risk factors are accumulated to obtain the first risk score.
[0018] Compared with existing technologies, the above-mentioned technical solution can obtain the first risk score through two scoring standards. By comparing these two solutions, a more suitable scoring method can be found, thereby improving the predictive efficiency of the ASD developmental trajectory prediction model.
[0019] In one possible implementation, in step S3, the second risk score is obtained by multiplying and summing the pretreated environmental risk factors with the corresponding effect sizes of each pretreated environmental risk factor.
[0020] In one possible implementation, in step S3, the effect size corresponding to each of the pretreated environmental risk factors is logarithmically transformed to obtain the transformed effect size, and the pretreated environmental risk factors and the transformed effect size corresponding to each of the pretreated environmental risk factors are multiplied and accumulated to obtain the second risk score.
[0021] Compared with existing technologies, the above-mentioned technical solution can obtain the second risk score through two different weighted accumulation methods: direct accumulation and logarithmic transformation. The direct accumulation method can directly reflect the sum of the effect sizes of environmental risk factors, while the logarithmic transformation can make the data more in line with a normal distribution, reduce the skewness of the data, and thus improve the stability and accuracy of the multi-environmental risk scoring model.
[0022] In one possible implementation, in step S3, the third risk score is obtained by multiplying and summing the effect size and score corresponding to each of the pretreated environmental risk factors.
[0023] Compared with existing technologies, the above-mentioned technical solution can comprehensively consider the impact of credibility level and effect size, and give a more reasonable third risk score under normal circumstances. Attached Figure Description
[0024] Figure 1 A step flow chart of the present application. DETAILED DESCRIPTION
[0025] First of all, those skilled in the art should understand that these embodiments are only used to explain the technical principles of the embodiments of the present application, and are not intended to limit the protection scope of the embodiments of the present application. Those skilled in the art can adjust them as needed in order to adapt to specific application occasions.
[0026] The present application will be further described in detail below in conjunction with the drawings and specific embodiments.
[0027] Reference Figure 1 , the embodiments of the present application disclose an ASD development trajectory prediction method based on PERS and interval censored data, mainly comprising five steps of step S1, step S2, step S3, step S4 and step S5, wherein in step S1, for a plurality of ASD patients, the corresponding environmental exposure records and existing milestone event records in the development process of each ASD patient are obtained, and a plurality of environmental risk factors and covariates related to each environmental risk factor are extracted from the environmental exposure records; in step S2, each environmental risk factor is preprocessed to obtain preprocessed environmental risk factors; in step S3, each preprocessed environmental risk factor is input into a pre-constructed multi-environmental risk scoring model to analyze and obtain an environmental risk score; in step S4, the icenReg package in R language is used to correct each covariate to obtain a corrected covariate; in step S5, the environmental risk score, the corrected covariate and the existing milestone event record are input into a pre-constructed ASD development trajectory prediction model based on interval censored data to predict the milestone event and the occurrence probability at the future time.
[0028] In the embodiments of the present application, considering that most of the existing researches on autism spectrum disorder (ASD) development trajectory are mainly concentrated on the analysis of the influence of single or a small number of environmental factors, lack comprehensive consideration of multiple environmental risk factors, and cannot fully reflect the influence of environmental factors on ASD development trajectory, the present application aims to solve this problem and construct a multi-environmental risk scoring model that can comprehensively consider multiple environmental risk factors; at present, when processing development milestone event data, most studies do not consider the interval censored data problem, and the traditional method directly ignores such interval censored data, resulting in inaccurate prediction of the development milestone event occurrence time by the model, and unable to provide reliable scientific basis for early screening and precise intervention, the present application is dedicated to solving this problem and improving the prediction accuracy of the milestone event occurrence time at the future time in the development process.
[0029] In the embodiments of the present application, two models are mainly constructed, one is a PERS model, and the other is an ASD development trajectory prediction model. The PERS model relies on the SSC database, extracts the environmental exposure records of ASD patients, screens high-confidence environmental risk factors, and performs data standardization processing. A variety of scoring methods are used to construct different PERS models. The optimal scoring method is selected through correlation statistical analysis, and the covariates are corrected. The ASD development trajectory prediction model is based on interval censored data, uses a Bayesian regression model to predict the time of milestone events in the future of ASD patients, evaluates the development differences of different risk groups through correlation test, establishes a web platform for precise development milestone event prediction, and finally integrates the PERS model to develop a prediction tool and carry out clinical pilot study.
[0030] In the embodiments of the present application, in step S1, the environmental exposure records of 2508 ASD patients aged 4-18 years old are carefully extracted from the SSC database. These records cover multiple important variables, including family medical history, maternal pregnancy disease, drug use, and perinatal records. These data are the basis for subsequent research and are crucial for a comprehensive understanding of the environmental exposure of ASD patients.
[0031] In the embodiments of the present application, in step S1, 18 high-confidence environmental risk factors are selected from the numerous environmental factors in the environmental exposure records. These environmental risk factors include increased parental age, maternal autoimmune disease family history, maternal gestational hypertension, use of antidepressants during pregnancy, and breastfeeding. These environmental risk factors are strictly screened and are believed to be closely related to the occurrence and development of ASD.
[0032] In the embodiments of the present application, in step S2, the selected environmental risk factors are comprehensively standardized. First, missing values are filled in by reasonable methods such as mean filling and median filling to ensure data integrity. Then, abnormal value correction is performed to identify and process abnormal values in the data to avoid interference with subsequent analysis. Finally, variable encoding is performed using a binary encoding method to encode environmental risk factors as 0 and 1, where 0 indicates that the factor does not exist, and 1 indicates that the factor exists. This encoding method is convenient for subsequent data analysis and model construction.
[0033] In the embodiment of the present application, in step S3, the first risk score is obtained by respectively scoring each pre-processed environmental risk factor according to the credibility level of each pre-processed environmental risk factor, the second risk score is obtained by weighted accumulation of each pre-processed environmental risk factor according to the effect amount of each pre-processed environmental risk factor, and the third risk score is obtained by weighted accumulation of the score of each pre-processed environmental risk factor according to the effect amount of each pre-processed environmental risk factor, and the optimal risk score is selected from the first risk score, the second risk score and the third risk score as the environmental risk score.
[0034] In the embodiment of the present application, when calculating the first risk score, scoring is performed according to the credibility level of the pre-processed environmental risk factor, and two different scoring schemes are adopted. In scheme one, scoring is performed according to the standard of 6 / 4 / 2 / 1 points, positive points are given to the pre-processed environmental risk factor positively correlated with ASD, and negative points are given to the pre-processed environmental risk factor negatively correlated with ASD, then the scores of all pre-processed environmental risk factors are accumulated to generate the first risk score. This scoring method is simple and direct, and can quickly quantify and evaluate the environmental risk factor. The specific calculation method is as follows:
[0035] wherein, represents the first risk score in scheme one, represents the first pre-processed environmental risk factor, represents the second pre-processed environmental risk factor, represents the third pre-processed environmental risk factor, represents the fourth pre-processed environmental risk factor. In scheme two, the scoring standard of 4 / 3 / 2 / 1 points is adopted, and positive and negative points are assigned according to the correlation of the pre-processed environmental risk factor with ASD, and the total score is accumulated. The specific calculation method is as follows:
[0036] wherein, represents the first risk score in scheme two. By comparing the two schemes, a more suitable scoring method can be found to improve the prediction performance.
[0037] In the embodiment of the present application, when calculating the second risk score, two different weighted accumulation methods are adopted based on the effect amount (OR / RR) of the pre-processed environmental risk factor. In scheme one, the effect amount of each pre-processed environmental risk factor is directly accumulated to obtain the weighted total score as the second risk score. This method is simple and intuitive, and can directly reflect the sum of the effect amount. The specific calculation method is as follows:
[0038] wherein, represents the second risk score in scheme one, represents the effect size of the first pre-processed risk factor, represents the first pre-processed environmental risk factor, represents the effect size of the second pre-processed risk factor, represents the second pre-processed environmental risk factor; Scheme two is a logarithmic conversion method (ln(OR / RR)), the effect size of the pre-processed environmental risk factor is logarithmically converted to obtain a converted effect size, and then weighted accumulation is performed, the logarithmic conversion can make the data more conform to the normal distribution, reduce the skewness of the data, and thus improve the stability and accuracy of the model, and the specific calculation mode is as follows:
[0039] wherein, represents the second risk score in scheme two, represents the converted effect size of the first pre-processed risk factor, represents the converted effect size of the second pre-processed risk factor.
[0040] In the embodiment of the present application, when calculating the third risk score, the evidence level score and the effect size are combined for weighting, a double-weighted calculation equation is established, the credibility level and the effect size of the pre-processed environmental risk factor are comprehensively considered, the influence of the pre-processed environmental risk factor on ASD can be more comprehensively and accurately evaluated, and thus the prediction accuracy is improved, and the specific calculation mode is as follows:
[0041]
[0042]
[0043]
[0044] From , , and , the optimal one is selected as the third risk score.
[0045] In the embodiment of the present application, data of key developmental events such as social and language, action, eating skills and toilet behavior are collected from the SSC database, which are important indicators for evaluating the development of ASD patients and are of great significance for verifying the correlation between the PERS model and the ASD phenotype, and then the Bayesian regression model in the icenReg package of R language is used to predict the occurrence time and probability of the milestone event at the future time based on the interval censored data. Coefficient of determination comparison 、 、 The prediction performance of the three calculation methods, The coefficient of determination can measure the fitting degree of the calculation method to the data, and the closer the value is to 1, the better the prediction performance is, and the optimal calculation method is screened out by comparing the coefficients of determination of the three calculation methods.
[0046] In the embodiment of the present application, in step S4, the risk model based on interval censored data (icenReg package of R language) is used to correct the covariates such as gender, age, parent demographic characteristics and the like, which may have an impact on the development trajectory of ASD, and by correcting the covariates, the synergistic mechanism of other pre-processed environmental risk factors can be revealed, and the relationship between the pre-processed environmental risk factors and ASD can be more accurately reflected.
[0047] In the embodiment of the present application, in step S5, based on the interval censored data, the Bayesian regression model in the icenReg package of R language is used to predict the occurrence time and probability of the milestone event at the future time of the ASD patient, the interval censored data means that only the time range of the event occurrence is known, and the Bayesian regression model can well process such interval censored data, and by introducing prior information, the occurrence time and probability of the milestone event at the future time are more accurately speculated; according to the environmental risk score, the covariates and the existing developmental milestone event record, the future developmental milestone event is speculated, and due to the existence of the interval censored data, the Bayesian regression model can be used in the present application to speculate the probability of the occurrence of a certain milestone event at a certain time, specifically, the speculation that the future developmental milestone event has 5%, 25%, 45%, 50%, 55%, 75% and 95% possibility of occurring at a certain time can be given, and these probability information can help doctors and parents better understand the development trend of the ASD patient and make intervention preparation in advance.
[0048] In the embodiments of the present application, a webpage platform for predicting ASD precise development milestones can be established, which integrates the above-mentioned multi-environment risk scoring model and the ASD development trajectory prediction model based on interval censored data. Users can input the relevant information of ASD patients (such as environmental exposure records, covariate information, and existing development milestone event records) to obtain the prediction results of future development milestone events. The establishment of the webpage platform can facilitate doctors, parents and researchers to use the model, and improve the practicability and popularization of the model.
[0049] In the embodiments of the present application, the ASD development trajectory prediction tool can be developed by integrating the PERS model. The ASD development trajectory prediction tool can comprehensively consider the preprocessed environmental risk factors and the existing development milestone event records, and can more comprehensively and accurately predict the development trajectory of ASD patients. In addition, the ASD development trajectory prediction tool can be used for clinical pilot study in cooperation with medical institutions. Through the application of the ASD development trajectory prediction tool in the actual clinical environment, the practical application value of the model can be verified. The clinical pilot study can collect real ASD patient data, evaluate the accuracy and reliability of the model in actual application, and provide a basis for further optimization and popularization of the model. At the same time, through the clinical pilot study, the implementation of precise screening and individualized intervention programs can be promoted, and more effective treatment and intervention measures can be provided for ASD patients.
[0050] In the description of the present application, the description of the terms "one embodiment", "some embodiments", "in this embodiment", "specific examples", or "some examples" means that the specific features, mechanisms, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, mechanisms, materials or characteristics described can be combined in any suitable manner in any one or more embodiments or examples. In addition, the skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples without contradiction.
[0051] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any skilled person in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.
Claims
1. A method for predicting ASD developmental trajectories based on PERS and interval censoring data, characterized in that, Includes the following steps: Step S1: For multiple ASD patients, obtain the environmental exposure records and existing milestone event records during the development process for each ASD patient, and extract multiple environmental risk factors and covariates related to each environmental risk factor from the environmental exposure records; Step S2: Preprocess each of the environmental risk factors to obtain preprocessed environmental risk factors; Step S3: Input each of the preprocessed environmental risk factors into a pre-constructed multi-environmental risk scoring model to analyze and obtain an environmental risk score; Step S4: Use the ichenReg package in R language to correct each covariate to obtain the corrected covariates; Step S5: Input the environmental risk score, the corrected covariates, and the existing milestone event records into a pre-built ASD developmental trajectory prediction model based on interval censoring data to predict future milestone events and their probabilities of occurrence.
2. The ASD developmental trajectory prediction method according to claim 1, characterized in that, In step S1, the environmental exposure records and existing milestone event records of multiple ASD patients aged 4 to 18 years are extracted from the Simmons Autism Database.
3. The ASD developmental trajectory prediction method according to claim 1, characterized in that, In step S1, the family medical history, maternal pregnancy-related diseases, medication use records, and perinatal records of each ASD patient are obtained as the environmental exposure records.
4. The ASD developmental trajectory prediction method according to claim 1, characterized in that, In step S2, for each environmental risk factor, missing value imputation, outlier correction, and variable unified encoding are performed sequentially to obtain the corresponding preprocessed environmental risk factor.
5. The ASD developmental trajectory prediction method according to claim 1, characterized in that, In step S3, a first risk score is obtained by assigning and accumulating scores to each pre-treated environmental risk factor based on the confidence level of each pre-treated environmental risk factor. A second risk score is obtained by weighted accumulating scores to each pre-treated environmental risk factor based on the effect size of each pre-treated environmental risk factor. A third risk score is obtained by weighted accumulating scores to each pre-treated environmental risk factor based on the effect size of each pre-treated environmental risk factor. The optimal risk score is selected from the first risk score, the second risk score, and the third risk score as the environmental risk score.
6. The ASD developmental trajectory prediction method according to claim 5, characterized in that, In step S3, each of the pretreated environmental risk factors is scored using a standard of 6 points, 4 points, 2 points, and 1 point. Pretreated environmental risk factors that are positively correlated with ASD are assigned positive scores, and pretreated environmental risk factors that are negatively correlated with ASD are assigned negative scores. Then, the scores of each of the pretreated environmental risk factors are accumulated to obtain the first risk score.
7. The ASD developmental trajectory prediction method according to claim 5, characterized in that, In step S3, each of the pretreated environmental risk factors is scored using a standard of 4 points, 3 points, 2 points, and 1 point. Pretreated environmental risk factors that are positively correlated with ASD are assigned positive scores, and pretreated environmental risk factors that are negatively correlated with ASD are assigned negative scores. Then, the scores of each of the pretreated environmental risk factors are accumulated to obtain the first risk score.
8. The ASD developmental trajectory prediction method according to claim 5, characterized in that, In step S3, the second risk score is obtained by multiplying and summing the pretreated environmental risk factors with the corresponding effect sizes of each pretreated environmental risk factor.
9. The ASD developmental trajectory prediction method according to claim 5, characterized in that, In step S3, the effect size corresponding to each of the pretreated environmental risk factors is logarithmically transformed to obtain the transformed effect size, and the pretreated environmental risk factors and the transformed effect size corresponding to each of the pretreated environmental risk factors are multiplied and accumulated to obtain the second risk score.
10. The ASD developmental trajectory prediction method according to claim 5, characterized in that, In step S3, the third risk score is obtained by multiplying and summing the pretreated environmental risk factors with the corresponding effect size and score.