CN115337000A - Machine learning method for evaluating brain aging caused by diseases based on brain structure images - Google Patents

Abstract

The invention discloses a machine learning method for evaluating brain aging caused by diseases based on brain structure images, and extracts brain structure features of different brain regions from brain structure magnetic resonance images, including structural features such as thickness and volume of cerebral cortex in different brain regions. Since not all features are helpful for model prediction, the features are screened, and the screened out features that can be generalized on different training subsets and are more concise and effective are used to build a brain age prediction model based on ridge regression. K-fold cross-validation was used to find out the features that were repeatedly recognized in k models, and to locate the structural features of brain regions most relevant to brain age prediction. Finally, the trained model will be predicted on the patient data to assess the degree to which the disease affects brain aging.

Description

A Machine Learning Approach to Assess Disease-Induced Brain Aging Based on Brain Structural Imaging

technical field

The invention relates to the technical field of neuroimage data analysis, and in particular to a machine learning method for evaluating brain aging caused by diseases based on brain structure images.

Background technique

Brain aging is a natural process, but in this process, changes in brain volume, cortical thickness, etc. have obvious individual differences. Brain development follows a pattern during normal aging, which means that a normal age can be predicted based on the pattern of brain development. Brain age prediction not only has important scientific significance, but also has extensive clinical value. Studies have shown that various types of neurological diseases, metabolic diseases such as schizophrenia, metabolic disorders, diabetes, and decreased heart function are all related to brain aging. The brain age generated by the brain age prediction model has been well used to assess the correlation between these diseases and brain aging. If a subject's brain age is greater than the actual biological age, the subject may have an older brain, or a high degree of brain aging, which deviates from the normal aging trajectory and has a high risk of suffering from related diseases. And vice versa, if a subject's brain age is younger than their biological age, the subject probably has a young brain. The difference between a given individual's predicted brain age and actual age is known as the "brain age gap". This value is thought to reflect diffuse, multivariable morphological changes throughout the brain and may be a marker of overall brain health. The degree to which the brain aging trajectory deviates from the average trajectory of healthy brain aging can reflect the individual's future risk of neurodegenerative diseases. Therefore, constructing models based on the characteristic patterns of brain aging contained in neuroimaging data and detecting the aging trajectory of individual brains may provide a new method for studying changes in the brain during aging and how brain diseases affect normal brain aging.

Structural Magnetic Resonance Imaging (sMRI) has a unique ability to study brain structure non-invasively. For healthy and patients, the imaging data will present different structural features. These features can be used to predict the corresponding brain age of the subjects and estimate the biological age of the brain. Since different brain atlases have different basis for dividing the whole brain, different brain atlases have different predictive effects on brain age.

In recent years, with the development of artificial intelligence, predicting the brain age of different individuals based on machine learning using imaging data is one of the key clinical research directions. Provide helpful guidance. At present, various regression models such as support vector regression (SVR), ridge regression, and random forest (random forest, RF) are often used in the brain age prediction research of imaging studies to predict brain age. Ridge regression is a biased estimation regression method dedicated to collinear data analysis. It is essentially an improved least squares estimation method. By giving up the unbiasedness of the least squares method, it is at the cost of losing part of the information and reducing accuracy. A more realistic and reliable regression method is obtained to obtain regression coefficients. These studies confirm that brain age prediction algorithms based on MRI brain structure information are of great help to clinical diagnosis. However, features associated with brain age, such as cortical thickness, cortical area, curvature, etc., are extracted from structural MRI images. Combining these features will result in high-dimensional features, and feature selection is required for high-dimensional features. At present, principal component analysis, partial least squares, etc. are often used for dimensionality reduction in brain age research in imaging, but these dimensionality reduction methods are ambiguous and poorly interpretable, so how to choose a feature selection algorithm is crucial for brain age prediction . In addition, the current brain age prediction models based on machine learning rarely locate the characteristics of brain regions related to brain age. Locating the key brain regions that affect brain age is of great significance for model interpretability and clinical diagnosis. At the same time, measuring the degree of brain aging through the difference between brain age and actual age is helpful for early detection and identification of diseases. Therefore, a new prediction method is urgently needed to solve the above problems.

Contents of the invention

The purpose of the present invention is to address the deficiencies in the prior art and propose a machine learning method for evaluating brain aging caused by diseases based on brain structural images. This method can screen out generalizable and more concise and effective features on different training subsets, and can also locate the structural features of brain regions that contribute most to predicting brain age. It is best to predict the trained model on patient data, which can assess the degree to which the disease affects brain aging.

The present invention is achieved through the following technical solutions: a machine learning method for assessing brain aging caused by diseases based on brain structure images, comprising the following steps:

S1: Preprocessing and feature extraction of brain structure MRI images;

S2: performing feature screening on the brain structural features extracted in step S1 using the Bootstrap self-expanding method;

S3: Construct a ridge regression brain age prediction model according to the screening results of S2;

S4: Use k-fold cross-validation to locate the brain region that contributes the most to predicting brain age;

S5: Train and test the constructed brain age prediction model;

S6: Testing with an independent patient cohort dataset to assess the extent to which disease affects brain aging.

Preferably, said step S1 includes the following sub-steps:

S1.1: Divide the collected clinical data into type 0 patients, type 1 patients and healthy people according to clinical manifestations, and remove the non-brain structure images in the brain structure magnetic resonance images of these patients;

S1.2: Extract the magnetic resonance image of the target tissue according to the anatomical structure of the brain;

S1.3: Use the image segmentation algorithm to segment the magnetic resonance image of the target tissue into 3 different tissues according to the structure of gray matter, white matter and cerebrospinal fluid;

S1.4: Use the FreeSurfer software package to reconstruct the cortex of the segmented tissue, quantify the function, connection and structural properties of the human brain, perform three-dimensional reconstruction of the structural image, generate flattened or flattened images, and use different brain atlases to obtain images of different brain regions Anatomical parameters of cortical thickness, curvature, area, gray matter volume.

Preferably, said step S2 includes the following sub-steps:

S2.1: For the healthy group data set, set the proportion of the sampling subset to the healthy group data set;

S2.2: Set the sampling frequency and sampling method, that is, how many times to perform sampling with replacement;

S2.3: Sampling is performed according to the ratio set in S2.1 and the number and method set in S2.2, and the Pearson correlation between cortical thickness, cortical surface area and other characteristics and brain age is calculated for the sampled subset, and a significant correlation is retained The characteristic structural features are used as candidate features, and the frequency of these features appearing in the sampling is counted;

S2.4: Set the frequency value of the features extracted by the sampling subset, and use the filtered features as the final features of the model according to the frequency value.

Preferably, said step S3 includes the following sub-steps:

S3.1: Randomly divide the healthy group data set into training set and test set according to the proportion;

S3.2: Standardize the characteristic data of the divided training set and test set, unify the data from different sources into one reference system for comparison, and speed up the convergence when the positive-guaranteed program is running, and the convergence speed of most models after normalization will speed up;

S3.3: Define the value range of the alpha parameter of the ridge regression model;

S3.4: Define the evaluation index of cross-validation as R2, and use k-fold cross-validation to find the optimal parameters within the value range of the alpha parameter of the ridge regression model, even the parameter with the highest model accuracy;

S3.5: Use the optimal parameters as the model parameters under the brain template.

Preferably, said step S4 includes the following sub-steps:

S4.1: Perform k-fold cross-validation on the training set, and set the cross-validation evaluation index to R2;

S4.2: Obtain the feature weights of k models through the coef_ parameter of the ridge regression model, sort the weights of the k models from large to small, and obtain the structural features corresponding to the first h feature weights of the k models respectively ;

S4.3: Identify the features that recur in these k models, and locate the structural features of the brain regions that contribute most to the prediction of brain age.

Preferably, said step S5 includes the following sub-steps:

S5.1: Select the model with the best test score in the k-fold cross-validation under n brain templates, so that it can be retrained on the entire training set;

S5.2: Perform model testing on the test set to obtain the brain age of the model test;

S5.3: Calculate the MAE, R2, Pearson correlation coefficient and average error between the real age and the predicted brain age, as the evaluation index of the model, and finally select the brain age prediction model established by the Brainnetome brain atlas as the optimal model.

Preferably, said step S6 includes the following sub-steps:

S6.1: Normalize the data of the independent patient group, load the trained brain age prediction model, use the trained optimal model to test the patient test set, and obtain the brain age predicted by the model;

S6.2: Calculate the average error between the real age and the predicted brain age, and compare it with the average error of the healthy test set. If the average error of the patient test set is higher than that of the healthy group, it indicates that the disease will cause the patient's brain to age;

S6.3: Generate the fitting line between the real value and the predicted value of the healthy group data set and the patient group data set, and prove that the patient's brain deviates from the healthy brain aging trajectory by comparing the slopes of the two fitting lines.

Preferably, the brain atlas used in the step S1.5 includes AAL, DKT, Destrieux and Brainnetome.

The present invention proposes a machine learning method for assessing brain aging caused by diseases based on brain structural images, and extracts brain structural features of different brain regions from magnetic resonance images of brain structures, including structural features such as thickness and volume of cerebral cortex in different brain regions. Since not all features are helpful for model prediction, the features are screened, and the screened out features that can be generalized on different training subsets and are more concise and effective are used to construct a brain age prediction model based on ridge regression. K-fold cross-validation was used to find the features that were repeatedly recognized in k models, and to locate the structural features of brain regions most relevant to brain age prediction. Finally, the trained model will be predicted on the patient data to assess the extent to which the disease affects brain aging. This method can locate the structural features of the brain region that contribute the most to the prediction of brain age, and at the same time predict the trained model on patient data, and can evaluate the degree to which diseases affect brain aging.

Description of drawings

Figure 1 is a schematic flow chart of the machine learning method for evaluating brain aging caused by diseases based on brain structural images;

Figure 2 is a schematic diagram of the feature screening of the Bootstrap self-expanding method;

Figure 3 is a graph of the test set results of the brain age prediction model.

Detailed ways

The present invention will be further described below in conjunction with the accompanying drawings. In order to enable those skilled in the art to better understand the technical solution in the application, the present invention will be further described below in conjunction with the accompanying drawings. But these are only some embodiments of the present application, not all embodiments. Based on the specific embodiments described in this application, other embodiments obtained by other persons in the art without making creative efforts should fall within the scope of the present invention.

A machine learning method for assessing brain aging caused by diseases based on brain structural images of the present invention comprises the following steps:

S1: Preprocessing and feature extraction of brain structure MRI images;

S2: performing feature screening on the brain structural features extracted in step S1 using the Bootstrap self-expanding method;

S3: Construct a ridge regression brain age prediction model according to the screening results of S2;

S4: Use k-fold cross-validation to locate the brain region that contributes the most to predicting brain age;

S5: Train and test the constructed brain age prediction model;

S6: Testing with an independent patient cohort dataset to assess the extent to which disease affects brain aging.

In the step S1, the brain structure magnetic resonance image preprocessing and feature extraction include the following steps: first, since the original structure images of the magnetic resonance data all contain some non-brain structures, such as skulls and the like. Because the skull signal is not used in the subsequent analysis and the signal-to-noise ratio at the edge of the image is poor, it is necessary to remove non-brain structures such as the skull in the image in the image preprocessing operation. Then, during MRI image processing, sometimes only the state of certain specific regions is concerned, which requires the extraction of tissues from the target site according to the anatomical structure of the brain. In the preprocessing process, the brain image is divided into three different tissues according to the structure of gray matter, white matter and cerebrospinal fluid. This is because these three tissues have different functions in the brain, so this step requires the use of image segmentation algorithms . Finally, cortical reconstruction is carried out through the FreeSurfer software package to quantify the function, connection and structural properties of the human brain, perform three-dimensional reconstruction of the structural image, generate a flattened or expanded image, and obtain anatomical parameters such as cortical thickness, curvature, area, and gray matter volume . Four commonly used brain atlases were used to obtain brain structural features, including AAL, DKT, Destrieux, and Brainnetome. It specifically includes the following sub-steps:

S1.1: Divide the collected clinical data into type 0 patients, type 1 patients and healthy people according to clinical manifestations, and remove the non-brain structure images in the brain structure magnetic resonance images of these patients;

S1.2: Extract the magnetic resonance image of the target tissue according to the anatomical structure of the brain;

S1.3: Use the image segmentation algorithm to segment the magnetic resonance image of the target tissue into 3 different tissues according to the structure of gray matter, white matter and cerebrospinal fluid;

S1.4: Use the FreeSurfer software package to reconstruct the cortex of the segmented tissue, quantify the function, connection and structural properties of the human brain, perform three-dimensional reconstruction of the structural image, generate flattened or flattened images, and use different brain atlases to obtain images of different brain regions Anatomical parameters of cortical thickness, curvature, area, gray matter volume.

In the step S2, feature screening using the Bootstrap self-expanding method includes the following steps: first, for the healthy group data set, set the ratio of the sampling subset to the data set as r%, that is, r% samples are used to construct the sample for sampling secondly, set the sampling frequency and method, that is, perform s sampling without replacement; then, calculate the Pearson correlation between each feature and brain age, and keep the structural features with significant correlation (p<0.05) as candidate features, And count the frequency of these features appearing in s samples; finally, set the frequency of the features extracted by the sampling subset to t, that is, the features that are screened out t times in s samples are used as the final features of the model. The final size of the feature set is m*n, where m is the number of samples, and n is the dimensionality of the structural features of each sample. These filtered features are generalizable and more parsimonious and effective on different training subsets. It specifically includes the following sub-steps:

S2.1: For the healthy group data set, set the proportion of the sampling subset to the healthy group data set;

S2.2: Set the sampling frequency and sampling method, that is, how many times to perform sampling with replacement;

S2.3: Sampling is performed according to the ratio set in S2.1 and the number and method set in S2.2, and the Pearson correlation between cortical thickness, cortical surface area and other characteristics and brain age is calculated for the sampled subset, and a significant correlation is retained The characteristic structural features are used as candidate features, and the frequency of these features appearing in the sampling is counted;

S2.4: Set the frequency value of the features extracted by the sampling subset, and use the filtered features as the final features of the model according to the frequency value.

In the step S3, constructing the ridge regression brain age prediction model includes the following steps: first, the healthy group data set is randomly divided into a training set and a test set according to the ratio of a:b; secondly, the divided training set and test set are completed. The characteristic data is standardized, and the mean value of the standardized data is 0, and the standard deviation is 1; then, define the value range of the alpha parameter of the ridge regression model (0.01,0.1,1,3...60), and define the evaluation index of cross-validation As R2, k-fold cross-validation is used to find the optimal parameters within the value range; finally, the optimal parameters are used as the model parameters under the 4 brain templates. It specifically includes the following sub-steps:

S3.1: Randomly divide the healthy group data set into training set and test set according to the proportion;

S3.2: Standardize the characteristic data of the divided training set and test set, unify the data from different sources into one reference system for comparison, and speed up the convergence when the positive-guaranteed program is running, and the convergence speed of most models after normalization will speed up;

S3.3: Define the value range of the alpha parameter of the ridge regression model;

S3.4: Define the evaluation index of cross-validation as R2, and use k-fold cross-validation to find the optimal parameters within the value range of the alpha parameter of the ridge regression model, even the parameter with the highest model accuracy;

S3.5: Use the optimal parameters as the model parameters under the brain template.

In the step S4, using k-fold cross-validation to locate the brain region that contributes the most to the prediction of brain age includes the following steps: first, perform k-fold cross-validation on the training set, and set the cross-validation evaluation index to R2; then, through ridge regression The coef_ parameter of the model obtains the feature weights of k models, sorts the weights of these k models from large to small, and respectively obtains the structural features corresponding to the first h feature weights of the k models; finally, recognizes the k features recurring in each model, locating the structural features of brain regions most relevant to brain age prediction. It specifically includes the following sub-steps:

S4.1: Perform k-fold cross-validation on the training set, and set the cross-validation evaluation index to R2;

S4.2: Obtain the feature weights of k models through the coef_ parameter of the ridge regression model, sort the weights of the k models from large to small, and obtain the structural features corresponding to the first h feature weights of the k models respectively ;

S4.3: Identify the features that recur in these k models, and locate the structural features of the brain regions that contribute the most to the prediction of brain age.

In the step S5, training and testing the brain age prediction model includes the following steps: First, select the model with the best test score in the k-fold cross-validation under n brain templates, so that it can be re-tested on the entire training set. Training; secondly, test the model on the test set to obtain the brain age of the model test; finally, calculate the MAE, R2, Pearson correlation coefficient and average error between the real age and the predicted brain age, as the evaluation index of the model, and finally The brain age prediction model established by Brainnetome was selected as the optimal model. It specifically includes the following sub-steps:

S5.1: Select the model with the best test score in the k-fold cross-validation under n brain templates, so that it can be retrained on the entire training set;

S5.2: Perform model testing on the test set to obtain the brain age of the model test;

S5.3: Calculate the MAE, R2, Pearson correlation coefficient and average error between the real age and the predicted brain age, as the evaluation index of the model, and finally select the brain age prediction model established by the Brainnetome brain atlas as the optimal model.

In the step S6, using an independent patient group data set for testing to evaluate the extent to which the disease affects brain aging includes the following steps: First, the size of the patient group data set is p*n, p represents the number of patient samples, and n is each sample Structural feature dimension, normalize the patient data, load the trained brain age prediction model, use the trained optimal model to test the patient test set, and obtain the brain age predicted by the model; then, calculate the real age Compared with the average error of predicting brain age and the average error of the healthy test set, the average error of the patient test set is higher than that of the healthy group, indicating that the disease will cause the patient's brain to age; finally, in order to further verify that the brain of the patient group will appear Aging phenomenon, generate a fitting line between the real value and the predicted value of the healthy group and the patient group, and prove that the patient's brain deviates from the healthy brain aging trajectory by comparing the slopes of the two fitting lines. It specifically includes the following sub-steps:

S6.1: Normalize the data of the independent patient group, load the trained brain age prediction model, use the trained optimal model to test the patient test set, and obtain the brain age predicted by the model;

S6.2: Calculate the average error between the real age and the predicted brain age, and compare it with the average error of the healthy test set. If the average error of the patient test set is higher than that of the healthy group, it indicates that the disease will cause the patient's brain to age;

S6.3: Generate the fitting line between the real value and the predicted value of the healthy group data set and the patient group data set, and prove that the patient's brain deviates from the healthy brain aging trajectory by comparing the slopes of the two fitting lines.

In general, the present invention proposes a machine learning method for assessing disease-induced brain aging based on brain structural images. This method can locate the structural features of the brain region that contribute the most to the prediction of brain age, and at the same time predict the trained model on patient data, and can evaluate the degree to which diseases affect brain aging. The overall method flow is shown in Figure 1. First, the structural MRI data are preprocessed and feature extracted. The process includes: removing the skull and non-brain tissues of the brain image, segmenting the gray matter, white matter, and cerebrospinal fluid of the image, and performing structural image segmentation. Cortical reconstruction, generating a flattened image. After the preprocessing is completed, the structural characteristics such as cortical thickness, area, curvature, and volume of different brain regions are obtained statistically, and the combined brain region characteristics are constructed. Secondly, based on the Bootstrap bootstrap method, the combined features based on different brain regions are screened, that is, the features that are selected t times in the s sampling are used as the final features of the model, which can be generalized on different training subsets and are more concise and effective. . Next, build a ridge regression brain age prediction model, and obtain the optimal alpha parameter of ridge regression through k-fold cross-validation. Again, k-fold cross-validation was used to identify the recurring features in these k models, and to locate the structural features of brain regions most relevant to brain age prediction. Then, the model is trained and tested on the normal group dataset. Finally, based on the trained brain age prediction model, an independent patient group data set was tested to prove that the patient's brain has aging symptoms.

Example 1

This example uses the clinical data collected by the hospital, and divides the data into type 0 patients, type 1 patients and healthy people according to clinical manifestations. After sorting out the data and controlling the quality, 138 type 0 patients, 94 type 1 patients and 109 healthy people were finally enrolled.

The concrete implementation process of this method comprises the following steps:

(1) Brain structure magnetic resonance image preprocessing and feature extraction: First, because the original structural images of magnetic resonance data contain some non-brain structures, such as skulls. Because the skull signal is not used in the subsequent analysis and the signal-to-noise ratio at the edge of the image is poor, it is necessary to remove non-brain structures such as the skull in the image in the image preprocessing operation. Then, during MRI image processing, sometimes only the state of certain specific regions is concerned, which requires the extraction of tissues from the target site according to the anatomical structure of the brain. In the preprocessing process, the brain image is divided into three different tissues according to the structure of gray matter, white matter and cerebrospinal fluid. This is because these three tissues have different functions in the brain, so this step requires the use of image segmentation algorithms . Finally, cortical reconstruction is carried out through the FreeSurfer software package to quantify the function, connection and structural properties of the human brain, perform three-dimensional reconstruction of the structural image, generate a flattened or expanded image, and obtain anatomical parameters such as cortical thickness, curvature, area, and gray matter volume . Four commonly used brain atlases were used to obtain brain structural features, including AAL, DKT, Destrieux, and Brainnetome.

(2) As shown in Figure 2, the Bootstrap bootstrap method is used for feature screening: first, for the healthy group data set, set the proportion of the sampling subset to 80% of the data set, that is, r% samples are used to construct the sampling sample secondly, set the sampling frequency and method, that is, perform s sampling without replacement; then, calculate the Pearson correlation between each feature and brain age, and keep the structural features with significant correlation (p<0.05) as candidate features, And count the frequency of these features appearing in s samples; finally, set the frequency of the features extracted by the sampling subset to t, that is, the features that are screened out t times in s samples are used as the final features of the model. The final size of the feature set is m*n, where m is the number of samples, and n is the dimensionality of the structural features of each sample. These filtered features are generalizable and more parsimonious and effective on different training subsets.

(3) Construct the ridge regression brain age prediction model: firstly, the healthy group data set is randomly divided into training set and test set according to the ratio of a:b; secondly, the characteristic data of the divided training set and test set are standardized and standardized The mean value of the final data is 0, and the standard deviation is 1; then, define the alpha parameter range of the ridge regression model (0.01,0.1,1,3...60), define the evaluation index of cross-validation as R2, and use k-fold Cross-validation finds the optimal parameters within the range of values; finally, the optimal parameters are used as model parameters under the 4 brain templates. The results of ridge regression brain age prediction are shown in Figure 3.

(4) Use k-fold cross-validation to locate the brain region that contributes the most to predicting brain age: first, perform k-fold cross-validation on the training set, that is, divide the training set into k parts, and take k-1 of them as training data in turn. 1 copy is used as the test data, conduct experiments, and set the cross-validation evaluation index to R2; then, obtain the feature weights of the k models through the coef_ parameter of the ridge regression model, and sort the weights of the k models from large to small, respectively Obtain the structural features of different brain regions corresponding to the first h feature weights of the k models; finally, identify the features that appear repeatedly in the k models, locate the structural features of the brain regions that contribute greatly to brain age prediction, and feature weights The larger the value, the greater the contribution of the structural characteristics of the brain area to the prediction of brain age.

(5) Train and test the brain age prediction model: first, select the model with the best test score in the k-fold cross-validation under the 4 brain templates, so that it can be retrained on the entire training set; secondly, in the test The model test is performed on the set to obtain the brain age of the model test; finally, the MAE, R2, Pearson correlation coefficient and average error between the real age and the predicted brain age are calculated as the evaluation index of the model, and finally the Brainnetome brain atlas is selected. The brain age prediction model is used as the optimal model.

(6) Use an independent patient group data set for testing to evaluate the extent to which the disease affects brain aging: First, the size of the patient group data set is p*n, where p represents the number of patient samples, and n is the structural feature dimension of each sample. Normalize the patient data, load the trained brain age prediction model, use the trained optimal model to test the patient test set, and obtain the brain age predicted by the model; then, calculate the average of the real age and the predicted brain age The error is compared with the average error of the healthy test set. If the average error of the patient test set is higher than that of the healthy group, it indicates that the disease will cause the brain aging of the patient; finally, in order to further verify that the brain of the patient group will age, a healthy group is generated and the fitting line between the true value and the predicted value of the patient group. By comparing the slopes of the two fitting lines, it is proved that the patient's brain deviates from the healthy brain's aging trajectory.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims (8)

1. A machine learning method for evaluating brain aging caused by diseases based on brain structure images is characterized by comprising the following steps:

s1: preprocessing a brain structure magnetic resonance image and extracting characteristics of the brain structure magnetic resonance image;

s2: screening the structural features of the brain extracted in the step S1 by adopting a self-development method;

s3: constructing a ridge regression brain age prediction model according to the screening result of the S2;

s4: positioning a brain region which has the maximum contribution to the predicted brain age by adopting k-fold cross validation;

s5: training and testing the constructed brain age prediction model;

s6: tests were conducted using independent patient data sets to assess the extent to which disease affects brain aging.

2. The method for machine learning to evaluate brain aging caused by diseases based on brain structure image as claimed in claim 1, wherein the step S1 comprises the following sub-steps:

s1.1: dividing the acquired clinical data into a type 0 patient, a type 1 patient and a healthy person according to clinical manifestations, and removing non-brain structure images in brain structure magnetic resonance images of the patients;

s1.2: extracting the magnetic resonance image of the target tissue according to the brain planning structure;

s1.3: adopting an image segmentation algorithm to segment the magnetic resonance image of the target tissue into 3 different tissues according to the structures of the grey brain matter, the white brain matter and the cerebrospinal fluid;

s1.4: performing cortical reconstruction on the segmented tissue through a FreeSharfer software package, quantifying the function, connection and structural attributes of the human brain, performing three-dimensional reconstruction on the structural image to generate a flattened or flatwise image, and obtaining anatomical parameters of cortex thickness, curvature, area and gray matter volume of different brain areas by using different brain maps.

3. The method of claim 1, wherein the step S2 comprises the following sub-steps:

s2.1: setting the proportion of the sampling subset to the health group data set aiming at the health group data set;

s2.2: setting sampling times and sampling modes;

s2.3: sampling according to the proportion set in the S2.1 and the times and modes set in the S2.2, calculating the correlation between the cortex thickness and the cortex surface area characteristic of the sampled subset and the pilsner of the brain age, reserving the structural characteristic with obvious correlation as a candidate characteristic, and counting the frequency of the characteristic appearing in the sampling;

s2.4: and setting the frequency value of the feature extracted from the sampling subset, and taking the screened feature as the final feature of the model according to the frequency value.

4. The method for machine learning to evaluate brain aging caused by diseases based on brain structure image as claimed in claim 1, wherein the step S3 comprises the following sub-steps:

s3.1: randomly dividing a health group data set into a training set and a testing set according to a proportion;

s3.2: standardizing the characteristic data of the divided training set and test set;

s3.3: defining the value range of alpha parameters of the ridge regression model;

s3.4: defining the evaluation index of cross validation as R2, and searching for an optimal parameter in the value range of alpha parameters of the ridge regression model by using k-fold cross validation, namely the parameter with the highest model accuracy;

s3.5: and taking the optimal parameters as model parameters under the brain template.

5. The method for machine learning to assess brain aging caused by diseases based on brain structure images as claimed in claim 1, wherein the step S4 comprises the following sub-steps:

s4.1: performing k-fold cross validation on the training set, and setting a cross validation evaluation index as R2;

s4.2: obtaining the feature weights of k models through coef _ parameters of ridge regression models, sequencing the weights of the k models from large to small, and respectively obtaining the structural features corresponding to the first h feature weights of the k models;

s4.3: features that recur in these k models are identified, and structural features of the brain region that contribute most to the predicted brain age are located.

6. The method for machine learning to assess brain aging caused by diseases based on brain structure images as claimed in claim 1, wherein the step S5 comprises the following sub-steps:

s5.1: selecting a model with the best test score in the k-fold cross validation under the n brain templates, and retraining the model on the whole training set;

s5.2: performing model test on the test set to obtain the brain age of the model test;

s5.3: and calculating MAE, R2, pearson correlation coefficient and average error between the real age and the predicted brain age, taking the MAE, R2, pearson correlation coefficient and average error as evaluation indexes of the model, and finally selecting a brain age prediction model established by a brain atlas as an optimal model.

7. The method of claim 1, wherein the step S6 comprises the following sub-steps:

s6.1: carrying out normalization processing on independent patient group data, loading a trained brain age prediction model, and testing a patient test set by using the trained optimal model to obtain the brain age predicted by the model;

s6.2: calculating the average error between the real age and the predicted brain age, and comparing the average error with the average error tested by the health test set, wherein if the average error of the patient test set is higher than that of the health group, the disease can cause the brain aging of the patient;

s6.3: and generating a fit line between the real value and the predicted value of the health group data set and the patient group data set, and comparing the slopes of the two fit lines to prove that the brain of the patient deviates from the aging track of the healthy brain.

8. The method of claim 2, wherein the brain atlas used in step S1.4 includes AAL, DKT, desrieux and Brainneme.

Images