Method and system for correcting biological age prediction model deviation

Abstract

The invention discloses a method for correcting the deviation of a biological age prediction model, and the method comprises the steps: S1, collecting physiological health data, carrying out the standardization processing of the physiological health data, removing abnormal values and missing values, and forming a sample set; s2, based on industrial standards, dividing data in the sample set into a healthy group and a non-healthy group, and using the data in the healthy physiological aging sample set to preliminarily establish a biological age prediction model; s3, purifying data in the healthy physiological aging sample set to realize cleaning of the sample set, and forming an approximate absolute healthy physiological aging sample set; s4, constructing a new prediction model on the cleaned sample set, recording the new prediction model as a new prediction model, and establishing a corrected biological age prediction model through the approximate absolute health physiological aging sample set; s5, predicting deviation is corrected through relative individual risk differences; and S6, presetting a verification index, and verifying the corrected biological age prediction model through the verification index.

Description

Technical Field

[0001] The present invention relates to the technical field of biological age prediction, and in particular to a method and system for correcting deviations of a biological age prediction model. Background Art

[0002] Physiological age refers to the age reflected by a person's physical health, which is different from their actual age. Physiological age can be assessed through a variety of physiological indicators, usually including indicators of physical fitness level, immune system, metabolic function, cardiovascular health, etc. The prediction of physiological age can help assess the health status of an individual and provide appropriate health advice.

[0003] With the increasing demand for risk management and early prevention of physiological diseases, accurate assessment of the degree of physiological aging has become a key area of ​​health management and medical research. The aging of physiological age is not only one of the signs of physiological aging, but also an independent risk factor for the occurrence and development of physiological diseases. Current physiological age calculation methods are mainly based on biomarkers. Usually, these methods use actual age as a regression label and establish prediction models by analyzing a series of biomarkers, such as blood test indicators, DNA methylation, and brain MRI. However, using only actual age as a prediction label will cause age bias in the physiological age prediction model. This bias is manifested in that the model predicts the physiological age of young people too high and the elderly too low, which affects the accuracy of health assessment and makes physiological age limited in assessing individual health status. Summary of the invention

[0004] The purpose of the present invention is to solve the problem that the existing physiological age prediction model is prone to age deviation, and to propose a method and system for correcting the deviation of the biological age prediction model.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A method for correcting a biological age prediction model bias, comprising:

[0007] S1. Collect physiological health data, standardize the physiological health data, remove abnormal values ​​and missing values, and form a sample set;

[0008] S2. Based on industry standards, the data in the sample set are divided into a healthy group and an unhealthy group. The data set formed by the samples in the healthy group is recorded as a healthy physiological aging sample set. The data in the healthy physiological aging sample set is used to preliminarily establish a biological age prediction model. The preliminarily established biological age prediction model is used to generate a predicted biological age for each individual.

[0009] S3, purifying the data in the healthy physiological aging sample set to clean the sample set, wherein the purification method is to further screen samples from the healthy physiological aging sample set based on relative individual risk differences to form an approximate absolute healthy physiological aging sample set;

[0010] The relative individual risk difference is the difference between the predicted biological age of each individual and the corresponding mean biological age of the healthy group, which is used to characterize the health risk difference of the individuals;

[0011] S4. constructing a new prediction model on the cleaned sample set, denoted as a new prediction model, wherein the new prediction model is a supervised regression model in machine learning or deep learning, and establishing a corrected biological age prediction model through the approximate absolute healthy physiological aging sample set, wherein the corrected biological age prediction model is used to generate a corrected biological age;

[0012] The corrected biological age prediction model is a newly constructed prediction model;

[0013] S5. In the process of predicting physiological age, the corrected biological age prediction model corrects the prediction deviation by the relative individual risk difference, wherein the prediction deviation is the difference between the predicted physiological age and the actual age;

[0014] S6. Preset verification indicators, and verify the corrected biological age prediction model through the verification indicators.

[0015] As a further solution of the present invention: in S1, the physiological health data includes blood test indicators, DNA methylation data, and brain MRI data.

[0016] As a further solution of the present invention: in S3, the calculation formula of the relative individual risk difference is:

[0017] RIRD i =|PA i -MPA i |

[0018] Among them, RIRD i is the relative individual risk difference value of the i-th sample, PA i is the predicted physiological age of the ith sample, MPA i is the mean physiological age of the group in the healthy physiological aging sample set, MPA i Obtained by fitting PA and chronological age in the healthy group.

[0019] As a further solution of the present invention: in S3, an evaluation value is preset, and the relative individual risk difference is less than or equal to the evaluation value as a sample screening criterion.

[0020] As a further solution of the present invention: in S3, the samples whose relative individual risk difference is less than or equal to the evaluation value are used as the approximate absolute healthy physiological aging group to purify the sample data of healthy physiological aging of the healthy group, thereby forming the approximate absolute healthy physiological aging sample set.

[0021] As a further solution of the present invention: in S5, during the process of correcting the prediction deviation, it is necessary to perform model training on the corrected biological age prediction model.

[0022] As a further solution of the present invention: in S5, during the model training process, it is necessary to clean the samples screened out in the previous training level.

[0023] As a further solution of the present invention: in S6, the verification index includes a root mean square error. During verification, it is necessary to compare the root mean square error values ​​before and after correction.

[0024] A system for correcting deviations in a physiological age prediction model is applicable to a method for correcting deviations in a biological age prediction model, characterized in that it comprises: a data acquisition and preprocessing module, the data acquisition and preprocessing module is used to collect physiological health data, standardize the physiological health data, remove abnormal values ​​and missing values, and form a sample set; a sample screening module, the sample screening module is used to divide the data in the sample set into a healthy group and an unhealthy group, the data set formed by the samples in the healthy group is recorded as a healthy physiological aging sample set, the data in the healthy physiological aging sample set is purified to achieve cleaning of the sample set, and the sample set is selected from the healthy physiological aging sample set based on relative individual risk differences. In this set, samples are further screened to form an approximate absolute healthy physiological aging sample set; a model construction module, the model construction module is used to preliminarily establish a biological age prediction model, construct a new prediction model, and establish a corrected biological age prediction model. The preliminarily established biological age prediction model is used to generate a predicted biological age for each individual, the new prediction model is a supervised regression model in machine learning or deep learning, and the corrected biological age prediction model is a new prediction model reconstructed through the approximate absolute healthy physiological aging sample set; a deviation correction module, the deviation correction module is used to correct prediction deviations; a result verification module, the result verification module is used to verify the corrected biological age prediction model.

[0025] Beneficial effects of the present invention:

[0026] (1) In the present invention, the corrected biological age prediction model established by introducing relative individual risk differences can reduce prediction errors and improve the accuracy of the model. In the corrected biological age prediction model, the age bias phenomenon is reduced, and the prediction results are more closely matched with the individual's actual age and health status.

[0027] (2) In the present invention, the corrected biological age prediction model established by introducing relative individual risk differences can better identify unhealthy physiological aging individuals than the traditional prediction model, thereby effectively improving the disease screening ability of the model and providing a more accurate risk assessment tool for clinical practice.

[0028] (3) In the present invention, the model optimized relative to individual risk differences achieves stronger generalization ability through standardized samples, making it applicable to a wider range of health management scenarios and adaptable to individuals of different ages and health conditions, with higher potential for promotion and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The present invention will be further described below in conjunction with the accompanying drawings.

[0030] Figure 1 It is a flow chart of the method for correcting the deviation of the biological age prediction model of the present invention;

[0031] Figure 2 It is a schematic diagram of the structure of the system for correcting the deviation of the physiological age prediction model of the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] Age bias is a common problem in biological age prediction because the prediction model does not fully consider other risk factors of the individual, such as race, region, and lifestyle, which are crucial to physiological health assessment. Lack of attention to these individual differences can lead to bias in the prediction model, making it impossible for physiological age to accurately reflect the true health status. For example, a young person has a lower health risk, but because the prediction model is based only on actual age, the predicted physiological age may be higher than the actual situation. Correspondingly, the physiological age of older people may be underestimated. The age bias phenomenon may mislead doctors and patients about their health status, thereby affecting the effectiveness of health management and early intervention.

[0034] At present, the existing implementation schemes for biological age prediction mainly include two categories. The first category is the classic statistical model. KDM is a widely used statistical model that uses multiple health indicators to estimate biological age. KDM considers the linear relationship between age and different health indicators in model construction, thereby effectively improving the accuracy of age prediction. However, the prediction accuracy of the KDM model is dependent on data, which is relatively limited when dealing with nonlinear relationships in biomarkers. It is difficult to adapt to the complex relationship of multidimensional health indicators, so it is difficult to provide the best prediction effect. At the same time, the KDM model also has obvious age bias. The second category is machine learning models. Machine learning models such as multivariate linear regression are widely used in biological age prediction and are favored for their strong interpretability and relatively low computational cost. The multivariate linear regression model predicts physiological age by establishing a regression relationship between health indicators and actual age. However, since the multivariate linear regression model often uses actual age as a label, it ignores individual risk differences, resulting in a more obvious age bias in the predicted physiological age in different age groups. In addition, there are problems such as large differences in health status and non-standardization in the samples constructed by the multivariate linear regression model, which affects the generalization ability of the model. Although other models such as support vector machines and deep learning models can fit complex data better, they still produce age bias and limit the prediction accuracy in some cases.

[0035] See also Figure 1 As shown, the present invention is a method for correcting the deviation of a biological age prediction model, comprising:

[0036] S1. Collect physiological health data, standardize the physiological health data, remove outliers and missing values, and form a sample set. Standardization can remove outliers and missing values ​​to ensure the integrity and uniformity of physiological health data. The preprocessed data provides a reliable input for subsequent model construction and calculation of relative individual risk differences.

[0037] Among them, physiological health data include blood test indicators, DNA methylation data, brain MRI data, and physiological health data are multi-source, including but not limited to facial images, tongue coating pictures, and fundus images.

[0038] Specifically, blood test data usually include a set of numerical variables, such as blood sugar, blood lipids, and white blood cell counts. DNA methylation data usually comes from high-throughput sequencing technology (such as Illumina BeadChip). This type of data is based on CpG sites, and the methylation level of each CpG site is generally expressed as a β value, ranging from 0 to 1. Brain MRI data can be image data (such as DICOM format) or processed feature data (such as brain region volume, gray-white matter ratio, fiber bundle data). For image data, image processing methods are needed to extract relevant features; for feature data, the processing method is similar to other numerical data.

[0039] In this step, standardization is to convert various types of data into the same scale so that each indicator is comparable.

[0040] Specifically, each blood test indicator is standardized. The commonly used standardization method is Z-score standardization, and its formula is as follows:

[0041] Z=(X-μ) / σ

[0042] Among them, X is the original data, μ is the mean, and σ is the standard deviation.

[0043] DNA methylation data usually need to be standardized. Common methods include:

[0044] Quantile normalization: used to process methylation data in multiple samples to make the distribution consistent in different samples.

[0045] Logit transformation: For β values ​​(0 to 1), a logit transformation can be performed to map to the range from negative infinity to positive infinity, making the data more suitable for downstream analysis. The formula is as follows:

[0046] logit(X)=log[X / (1-X)]

[0047] For brain MRI feature data, the following normalization methods are usually used:

[0048] Z-score standardization: Like blood test data, the volume, gray-white matter ratio and other indicators of each brain region are standardized.

[0049] Min-Max Normalization: Scale each eigenvalue to a specified interval, usually [0, 1].

[0050] Furthermore, regarding outliers and missing values, since outliers will seriously affect the results of data analysis, they need to be removed. Statistical methods (including but not limited to standard deviation method, box plot method) or machine learning methods (including but not limited to isolation forest, local outlier factor) can be adopted. In addition, the missing value processing method should be selected according to the characteristics of the data and the type of missingness.

[0051] Furthermore, the missing value processing of blood test indicators is generally processed by mean / median filling method, KNN interpolation method, and multiple interpolation method; the missing value processing of DNA methylation data is generally processed by zero filling method and multiple interpolation method; the missing value processing of brain MRI data is generally processed by linear interpolation method and model-based interpolation method (such as using regression model to predict missing feature values).

[0052] The above-mentioned methods for processing outliers and missing values ​​are all existing technologies and will not be elaborated here.

[0053] S2. Based on industry standards, the data in the sample set are divided into a healthy group and an unhealthy group. The data set formed by the samples in the healthy group is recorded as the healthy physiological aging sample set, and the data set formed by the samples in the unhealthy group is recorded as the unhealthy physiological aging sample set. The data in the healthy physiological aging sample set is used to preliminarily establish a biological age prediction model, and the preliminarily established biological age prediction model is used to generate the predicted biological age of each individual.

[0054] Industry standards are different for different fields, and they also vary for different data objects. For example, if you are predicting brain biological age, you use MRI data, and the standard can be the Alzheimer's risk detection scale; if it is vascular biological age, you can use arteriosclerosis detection data, and the standard can be arterial flow velocity; if it is blood biological age, you can use blood test indicators, and the standard is the range of blood test indicators.

[0055] S3. Purify the data in the healthy physiological aging sample set to clean the sample set. The purification method is to further screen samples from the healthy physiological aging sample set based on relative individual risk differences to form an approximate absolute healthy physiological aging sample set.

[0056] The relative individual risk difference is the difference between the predicted biological age of each individual and the mean biological age of the corresponding healthy group, which is used to characterize the individual health risk differences.

[0057] The sample set is cleaned and the samples are screened based on the relative individual risk differences. In this step, the relative individual risk differences are introduced into the healthy group to screen out the approximate absolute healthy physiological aging group (approximate absolute healthy physiological aging sample set).

[0058] In the process of cleaning the sample set, it is necessary to calculate the relative individual risk difference. The relative individual risk difference is the difference between the predicted physiological age of each individual and the mean physiological age of the healthy physiological aging group, which is used to characterize the individual health risk differences.

[0059] The formula for calculating the relative individual risk difference is:

[0060] RIRD i =|PA i -MPA i |;

[0061] Among them, RIRD i is the relative individual risk difference value of the i-th sample; PA i is the predicted physiological age of the ith sample. The predicted physiological age of the ith sample needs to be calculated using the initially established biological age prediction model; MPA i is the mean physiological age of the group in the healthy physiological aging sample set, MPA i Obtained by fitting PA and chronological age in the healthy group.

[0062] An evaluation value is preset. In the present embodiment, the evaluation value is 3, and the relative individual risk difference is less than or equal to 3 as the sample screening criterion. The samples with the relative individual risk difference less than or equal to 3 are used as the approximate absolutely healthy physiological aging group to purify the sample data of healthy physiological aging of the healthy group, so as to ensure the consistency of the health status of these samples and reduce the interference of abnormal samples on the model, thereby forming an approximate absolutely healthy physiological aging sample set.

[0063] It should be noted that the screening criterion of relative individual risk difference less than or equal to 3 can also provide new technical ideas for data preprocessing and standardization of the model.

[0064] S4. Construct a new prediction model on the cleaned sample set, denoted as the new prediction model. The new prediction model is a supervised regression model in machine learning or deep learning. By approximating the absolute healthy physiological aging sample set, a corrected biological age prediction model is established. The corrected biological age prediction model is used to generate the corrected physiological age.

[0065] Among them, the corrected biological age prediction model is a newly reconstructed prediction model.

[0066] S5. In the process of predicting physiological age, the corrected biological age prediction model corrects the prediction bias by relative individual risk differences. The prediction bias is the difference between the predicted physiological age and the actual age.

[0067] Specifically, in the process of correcting the prediction bias, the corrected biological age prediction model needs to be trained. During the model training process, the samples screened out in the previous training level need to be cleaned so that the corrected biological age prediction model can reduce the impact of age bias in different age groups. After introducing relative individual risk differences, the corrected biological age prediction model can more accurately reflect the real health risk differences of individuals, thereby reducing age bias.

[0068] The method of correcting prediction bias by using relative individual risk differences, especially by systematically eliminating age bias during model construction, can ensure the consistency and accuracy of the corrected biological age prediction model in different age groups.

[0069] S6. Preset verification indicators and verify the corrected biological age prediction model through the verification indicators.

[0070] Verification indicators include root mean square error. During verification, it is necessary to compare the root mean square error values ​​before and after correction to ensure that the corrected biological age prediction model has reduced errors in prediction. In addition, in terms of the effect of eliminating age bias, the consistency of the model in different age groups can be checked to ensure that the corrected model can maintain consistent prediction effects in young and older groups. It can also improve disease identification capabilities. Through experimental verification, the ability of the corrected biological age prediction model in detecting unhealthy physiological aging individuals can be evaluated to ensure the effectiveness of the corrected biological age prediction model in health risk assessment.

[0071] In this paper, a method for correcting the deviation of the biological age prediction model is proposed by introducing relative individual risk differences, aiming to improve the prediction accuracy and consistency of the model in different age groups. Introducing relative individual risk differences into the model construction and deviation correction process overcomes the age deviation problem of the traditional physiological age prediction model in young and elderly groups, thereby improving the accuracy of health risk assessment and the stability of the model.

[0072] Specifically, relative individual risk differences can effectively reflect individual health risk differences, thereby achieving accurate risk prediction for different age groups. This helps to compensate for the bias caused by the existing prediction model that only uses actual age as a label, uses relative individual risk differences as a new indicator to correct model bias, improves the accuracy of physiological age prediction, and enables the corrected biological age prediction model to better reflect the actual health status of individuals. By screening healthy samples with relative individual risk differences ≤3 and creating standardized physiological age group samples, the age bias of different age groups is effectively eliminated, the consistency of the corrected biological age prediction model in young and elderly groups is ensured, and the stability and applicability of the prediction are improved. The corrected biological age prediction model optimized by relative individual risk differences shows stronger recognition ability in detecting unhealthy physiological aging individuals, which helps to more accurately screen out individuals with health risks, effectively improves the disease screening ability of the model, and provides more accurate risk assessment tools for clinicians to achieve early warning and intervention.

[0073] The corrected biological age prediction model established by the present invention by introducing relative individual risk differences can reduce prediction errors and improve the accuracy of the model. In the corrected biological age prediction model, the age deviation phenomenon is reduced, the prediction results are more closely matched with the individual's real age and health status, and the root mean square error and other indicators are significantly reduced, so that the prediction errors of the corrected biological age prediction model in different age groups are effectively controlled.

[0074] In addition, the model optimized relative to individual risk differences achieves stronger generalization ability through standardized samples, making it suitable for a wider range of health management scenarios and adaptable to individuals of different ages and health conditions, with higher potential for promotion and application.

[0075] It is understandable that as an alternative, data standardization or bias correction methods can be applied. During data preprocessing, data standardization methods such as Z-Score standardization, mean correction or skewness correction can be used to reduce age bias. A modified loss model can also be introduced. In machine learning and deep learning models, a tilted loss function is used to correct the age bias phenomenon generated by different age stages.

[0076] See also Figure 2 As shown, in another embodiment, a system for correcting deviations in a physiological age prediction model is provided, which is applicable to a method for correcting deviations in a biological age prediction model.

[0077] The system for correcting the deviation of the physiological age prediction model includes: a data acquisition and preprocessing module, which is used to collect physiological health data, standardize the physiological health data, remove abnormal values ​​and missing values, and form a sample set.

[0078] The sample screening module is used to divide the data in the sample set into a healthy group and an unhealthy group. The data set formed by the samples in the healthy group is recorded as a healthy physiological aging sample set. The data in the healthy physiological aging sample set is purified to achieve the cleaning of the sample set, and samples are further screened from the healthy physiological aging sample set based on relative individual risk differences to form an approximate absolute healthy physiological aging sample set.

[0079] Model building module: The model building module is used to preliminarily establish a biological age prediction model, build a new prediction model, and establish a corrected biological age prediction model. The preliminarily established biological age prediction model is used to generate the predicted biological age of each individual. The new prediction model is a supervised regression model in machine learning or deep learning. The corrected biological age prediction model is a new prediction model reconstructed by approximating the absolute healthy physiological aging sample set.

[0080] Bias correction module,The bias correction module is used to correct the prediction bias.

[0081] The result verification module is used to verify the corrected biological age prediction model.

[0082] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of application of the present invention should still fall within the scope of the claims of the present invention.

Claims

1. A method for correcting the deviation of a biological age prediction model, characterized in that: include: S1. Collect physiological health data, standardize the physiological health data, remove abnormal values ​​and missing values, and form a sample set; S2. Based on industry standards, the data in the sample set are divided into a healthy group and an unhealthy group. The data set formed by the samples in the healthy group is recorded as a healthy physiological aging sample set. The data in the healthy physiological aging sample set is used to preliminarily establish a biological age prediction model. The preliminarily established biological age prediction model is used to generate a predicted biological age for each individual. S3, purifying the data in the healthy physiological aging sample set to clean the sample set, wherein the purification method is to further screen samples from the healthy physiological aging sample set based on relative individual risk differences to form an approximate absolute healthy physiological aging sample set; The relative individual risk difference is the difference between the predicted biological age of each individual and the corresponding mean biological age of the healthy group, which is used to characterize the health risk difference of the individuals; S4. constructing a new prediction model on the cleaned sample set, denoted as a new prediction model, wherein the new prediction model is a supervised regression model in machine learning or deep learning, and establishing a corrected biological age prediction model through the approximate absolute healthy physiological aging sample set, wherein the corrected biological age prediction model is used to generate a corrected biological age; The corrected biological age prediction model is a newly constructed prediction model; S5. In the process of predicting physiological age, the corrected biological age prediction model corrects the prediction deviation by the relative individual risk difference, wherein the prediction deviation is the difference between the predicted physiological age and the actual age; S6. Preset verification indicators, and verify the corrected biological age prediction model through the verification indicators.

2. A method for correcting the deviation of a biological age prediction model according to claim 1, characterized in that: In S1, the physiological health data includes blood test indicators, DNA methylation data, and brain MRI data.

3. A method for correcting the deviation of a biological age prediction model according to claim 1, characterized in that: In S3, the calculation formula of the relative individual risk difference is: RI RD i =|PA i -MPA i | Among them, RI RD i is the relative individual risk difference value of the i-th sample, PA i is the predicted physiological age of the ith sample, MPA i is the mean physiological age of the group in the healthy physiological aging sample set, MPA i Obtained by fitting PA and chronological age in the healthy group.

4. A method for correcting the deviation of a biological age prediction model according to claim 3, characterized in that: In S3, an evaluation value is preset, and the relative individual risk difference is less than or equal to the evaluation value as a sample screening criterion.

5. A method for correcting the deviation of a biological age prediction model according to claim 4, characterized in that: In S3, the samples whose relative individual risk difference is less than or equal to the evaluation value are used as the approximate absolute healthy physiological aging population to purify the sample data of healthy physiological aging of the healthy group, thereby forming the approximate absolute healthy physiological aging sample set.

6. A method for correcting the deviation of a biological age prediction model according to claim 1, characterized in that: In S5, during the process of correcting the prediction deviation, it is necessary to perform model training on the corrected biological age prediction model.

7. A method for correcting the deviation of a biological age prediction model according to claim 6, characterized in that: In S5, during the model training process, the samples screened out in the previous training level need to be cleaned.

8. A method for correcting the deviation of a biological age prediction model according to claim 1, characterized in that: In S6, the verification index includes a root mean square error. During verification, it is necessary to compare the root mean square error values ​​before and after correction.

9. A system for correcting deviations in a physiological age prediction model, applicable to a method for correcting deviations in a biological age prediction model as claimed in any one of claims 1 to 8, characterized in that: include: A data collection and preprocessing module, which is used to collect physiological health data, standardize the physiological health data, remove abnormal values ​​and missing values, and form a sample set; A sample screening module, the sample screening module is used to divide the data in the sample set into a healthy group and an unhealthy group, the data set formed by the samples in the healthy group is recorded as a healthy physiological aging sample set, the data in the healthy physiological aging sample set is purified to achieve the cleaning of the sample set, and samples are further screened from the healthy physiological aging sample set based on relative individual risk differences to form an approximate absolute healthy physiological aging sample set; A model construction module, wherein the model construction module is used to preliminarily establish a biological age prediction model, construct a new prediction model, and establish a corrected biological age prediction model. The preliminarily established biological age prediction model is used to generate a predicted biological age of each individual. The new prediction model is a supervised regression model in machine learning or deep learning. The corrected biological age prediction model is a new prediction model reconstructed by the approximate absolute healthy physiological aging sample set; A deviation correction module, wherein the deviation correction module is used to correct the prediction deviation; A result verification module is used to verify the corrected biological age prediction model.

Images