AD risk assessment method based on multistage iterative regression training

Abstract

The invention belongs to the technical field of diagnostic model construction, and particularly relates to an AD risk assessment method based on multistage iterative regression training. According to the method, image analysis and data extraction are carried out on normal and abnormal AD data collected by the public resource library, grading and iterative analysis are carried out, normal person and AD patient machine models are created, and machine intelligent diagnosis is carried out on the basis, so that the accuracy and preciseness of AD auxiliary diagnosis are improved, and the possibility of missed diagnosis and misdiagnosis is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of diagnostic model construction, and specifically relates to an AD risk assessment method based on multi-level iterative regression training. Background Art

[0002] Alzheimer's disease (AD) is a complex neurodegenerative disease, and its diagnosis currently relies mainly on the patient's clinical manifestations. In the early stages of the disease, patients may show symptoms such as memory loss, weakened language skills, disorientation in time and place, and decreased comprehension and execution abilities. As the disease progresses, patients in the middle and late stages may experience more serious symptoms such as continued memory loss, a significant decline in cognitive abilities, loss of self-care ability, emotional and mental state fluctuations, and abnormal physical functions.

[0003] However, existing diagnostic methods have limitations because they mainly rely on the patient's clinical manifestations and lack direct evidence of pathological changes to support early judgment. This diagnostic method that relies on clinical manifestations may increase the risk of missed diagnosis and misdiagnosis, especially in the early stages of the disease, when symptoms may be relatively mild and easily confused with other diseases.

[0004] To overcome this challenge, the present invention proposes an AD risk assessment method based on multi-level iterative regression training. Summary of the invention

[0005] In order to overcome the problems in the prior art, the present invention proposes an AD risk assessment method based on multi-level iterative regression training.

[0006] The technical solution of the present invention to solve the above technical problems is as follows:

[0007] The present invention provides an AD risk assessment method based on multi-level iterative regression training, comprising the following steps:

[0008] Constructing an abnormal brain structure model, including: calculating the probability of each abnormal brain structure appearing in all confirmed AD patients; selecting representative abnormal brain structure types of AD patients according to the probability value of each abnormal brain structure appearing in all confirmed AD patients and the credibility of pre-statistics; sorting the abnormal brain structure types of AD patients and giving scores; obtaining the entropy of the abnormal brain structure types of AD patients, and selecting the number of abnormal brain structure types according to the entropy value of the abnormal brain structure types of AD patients; obtaining an abnormal brain structure model through multiple trainings;

[0009] AD risk is assessed based on abnormal brain structure models.

[0010] Furthermore, the probability of each abnormal brain structure appearing in all confirmed AD patients is calculated using the formula:

[0011] P(x i )=x i / k;

[0012] In the above formula, P(x i ) is the probability of each abnormal brain structure appearing in all confirmed AD patients; k is the number of confirmed AD patients; x i The number of abnormal brain structures of AD patients confirmed as i, x i ≤n; n represents the maximum number of abnormal brain structures confirmed in AD patients.

[0013] Furthermore, the entropy of abnormal brain structure types of AD patients is:

[0014]

[0015] In the above formula, H(X) is the entropy of abnormal brain structure types in AD patients.

[0016] Furthermore, before constructing the abnormal brain structure model, it also includes: constructing a brain structure model for healthy people: using the LinearRegression regression model, performing regression calculations on each normal brain structure data, and establishing a brain structure model for healthy people through continuous training. The brain structure model for healthy people gives the relative relationship between brain structure-related features, which is used to compare and identify the abnormal brain structure of AD patients.

[0017] Furthermore, the AD risk assessment includes:

[0018] Collect the subjects' brain structure data and perform preprocessing;

[0019] Compare the MRI data of the subjects with the brain structure model of healthy people to determine whether there are abnormal brain structures;

[0020] If there is an abnormal brain structure, the type of abnormal brain structure is obtained based on the brain structure parameter model of AD patients, and its score is calculated. If the score exceeds the preset credibility allowable value, it is determined to be an AD patient with a certain credibility.

[0021] Compared with the prior art, the present invention has the following technical effects:

[0022] The present invention constructs abnormal brain structure models and brain structure models of healthy people, optimizes model parameters using multi-level iterative regression training, and introduces entropy value evaluation and ranking scoring mechanism to achieve a more accurate, scientific and convenient assessment of Alzheimer's disease risk, providing strong support for the early detection, prevention and intervention of AD. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0025] In order to further explain the technical means and effects taken by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the technical solutions proposed by the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments. The specific features, structures or characteristics in one or more embodiments may be combined in any suitable form. Unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by technicians in the technical field of the present invention.

[0026] The present invention integrates normal and abnormal Alzheimer's disease (AD) magnetic resonance imaging (MRI) data from public resource libraries, conducts in-depth image analysis and data extraction, and then constructs two core models: a brain structure model of healthy people and a brain structure parameter model of AD patients. Both models are optimized based on a multi-level iterative regression training method, which significantly improves the accuracy and rigor of AD auxiliary diagnosis.

[0027] In the process of constructing the brain structure model of healthy people, the present invention first preprocesses the MRI data and corrects the parameters to ensure the consistency and accuracy of the data. Subsequently, the linear regression model is used to train the brain structure characteristics of healthy people, providing a detailed standard for the MRI performance of the normal brain and a solid comparative basis for the subsequent analysis of the abnormal brain structure of AD patients.

[0028] When constructing the brain structure parameter model of AD patients, the probability of each abnormal brain structure appearing in all confirmed AD patients was calculated, and representative abnormal brain structure types were selected based on these probabilities and statistical credibility. By sorting and scoring these abnormal brain structure types, it is possible to focus more accurately on brain structure abnormalities directly related to AD, thereby improving the accuracy of AD risk assessment. In addition, the concept of entropy was introduced to evaluate the distribution characteristics of abnormal brain structure types in AD patients, providing a scientific basis for determining the number of abnormal brain structure types ultimately used for risk assessment.

[0029] Based on the two constructed models, the subjects are objectively and quantitatively diagnosed with AD. By comparing with the brain structure model of healthy people, it can be determined whether the subjects have abnormal brain structure. If there is an abnormality, its score is further calculated based on the brain structure parameter model of AD patients and compared with the preset credibility permission value to determine whether the subject is an AD patient with a certain credibility. This process not only reduces the possibility of missed diagnosis and misdiagnosis, but also provides strong support for the early detection, personalized treatment and in-depth research of AD.

[0030] In one embodiment of the present invention, referring to Figure 1 , provides an AD risk assessment method based on multi-level iterative regression training, including the following steps: constructing a brain structure model of healthy people; constructing a brain structure parameter model of AD patients; and diagnosing AD in the subjects based on the established brain structure parameter model of AD patients.

[0031] The following is a detailed explanation of each of the above steps:

[0032] Step 100: Construct a brain structure model of healthy people.

[0033] As an example, this step may include the following steps:

[0034] Step 101: MRI (magnetic resonance imaging) data acquisition and integration;

[0035] Obtain normal brain structure data from around the world from the MRI public resource library and unify the data format for subsequent processing;

[0036] The collected MRI data integrates the public resource library data and the collected MRI data into a unified database.

[0037] Step 102: preprocessing the integrated MRI data;

[0038] The collected MRI data are preprocessed, including denoising, standardization, etc. Gaussian filtering, median filtering, etc. can be used to remove noise in the image, and a registration algorithm (such as affine transformation, nonlinear transformation) can be used to align the images to the same space.

[0039] Step 103: performing parameter correction on the preprocessed MRI data to obtain a corrected data set;

[0040] Use image processing technology to extract brain structural features, such as volume and surface area, and compare the collected human brain structural features with data in the MRI (magnetic resonance imaging) public resource library to find out the differences;

[0041] Adjust the parameters of the MRI image, such as zooming and translation, according to the acquired structural features of the human brain.

[0042] Step 104: constructing a brain structure model of healthy people;

[0043] Features related to brain structure are extracted from MRI images; the modified dataset is divided into training set, validation set and test set.

[0044] Use the training set data to train the LinearRegression model and calculate the contribution of each feature to the regression result; use the validation set data for cross-validation and adjust the model parameters (such as regularization parameters) to prevent overfitting; use the test set data to evaluate the performance of the model and calculate the mean square error (MSE) and R 2 The model structure and parameters are adjusted according to the evaluation results until satisfactory performance is achieved.

[0045] LinearRegression model, also known as linear regression model, is a statistical method used to predict the linear relationship between one or more independent variables (features) and dependent variables (target variables). In machine learning and data analysis, linear regression is a basic and widely used regression analysis method. When establishing a brain structure model for healthy people, the linear regression model can be used as a basic model to explore the relationship between independent variables (brain structure characteristics) and dependent variables (brain function indicators).

[0046] The brain structure model of healthy people provides detailed standards for the MRI manifestations of the normal brain. These standards include the normal size, shape, position of various parts of the brain and their relative relationships. These benchmark data are the basis for subsequent analysis to compare and identify abnormal brain structures in AD patients.

[0047] The brain structure model of healthy people provides a basis for setting the credibility threshold. By analyzing the MRI data of healthy people, the normal range and degree of variation of each brain structure feature can be calculated. According to the standard range of brain structure of healthy people and the scores of the subjects, a reasonable credibility threshold can be set. When the scores of the subjects exceed this threshold, it can be determined that they have a certain risk of AD.

[0048] Step 200: Construct a brain structure parameter model for AD patients, use the LinearRegression regression model to perform regression calculations on each normal brain structure data, and establish a brain structure model for healthy people through continuous training. The brain structure model for healthy people provides the relative relationship between brain structure-related features for comparing and identifying abnormal brain structures in AD patients.

[0049] As an example, this step may include the following steps:

[0050] Step 201: Calculate the probability of each abnormal brain structure appearing in all confirmed AD patients.

[0051] P(x i )=x i / k;

[0052] In the above formula, P(x i ) is the probability of each abnormal brain structure appearing in all confirmed AD patients; k is the number of confirmed AD patients; x i The number of abnormal brain structures of AD patients confirmed as i, x i ≤n; n represents the maximum number of abnormal brain structures confirmed in AD patients.

[0053] Step 202: According to P(x i ) values ​​and statistical credibility to select the types of abnormal brain structures in AD patients.

[0054] Set a statistical credibility threshold (for example, 0.05) and select only P(x i ) is greater than the threshold value for abnormal brain structure types.

[0055] Step 203: According to P(x i ) Sort the abnormal brain structures of AD patients in descending order and give scores in reverse order.

[0056] The selected abnormal brain structure types are sorted according to P(x i ) values ​​are sorted from high to low.

[0057] Assuming there are m selected abnormal brain structures m≤n, the sorted structures are a1, a2, …, a m , where P(a1)≥P(a2)≥…≥P(a m );

[0058] Then, the scores are given in reverse order, for example: a1 score: 10×P(a1); a2 score: 10×P(a2)…am score: 10×P(a m ).

[0059] By calculating the probability of each abnormal brain structure appearing in all confirmed AD patients and selecting representative abnormal brain structures based on these probabilities and pre-statistical credibility, this method can more accurately focus on brain structural abnormalities directly related to AD and reduce the impact of noise data. Sorting and scoring the types of abnormal brain structures helps to identify the brain structural features that have the greatest impact on AD risk assessment, thereby improving the accuracy of the assessment.

[0060] Step 204: construct the entropy H(X) of the abnormal brain structure types of AD patients, and select the number of abnormal brain structure types according to the value of H(X).

[0061] Entropy H(X) of abnormal brain structure types in AD patients:

[0062]

[0063] The number of abnormal brain structure types finally selected can be determined according to the value of H(X) and actual needs. Through multiple trainings, an abnormal brain structure model is obtained.

[0064] The concept of entropy is introduced to evaluate the distribution characteristics of abnormal brain structure types in AD patients, which not only provides information about the diversity of abnormal brain structures, but also helps to determine the number of abnormal brain structure types ultimately used for risk assessment, thereby enhancing the scientificity and rationality of the model.

[0065] Step 300: Based on the constructed brain structure model of healthy people and the brain structure parameter model of AD patients, AD is diagnosed on the subject.

[0066] As an example, this step may include the following steps:

[0067] Collect the subjects' brain structure data and perform preprocessing;

[0068] Compare the MRI data of the subjects with the brain structure model of healthy people to determine whether there are abnormal brain structures;

[0069] If there is an abnormal brain structure, the type of abnormal brain structure is obtained based on the brain structure parameter model of AD patients, and its score is calculated. If the score exceeds the preset credibility allowable value, it is determined to be an AD patient with a certain credibility.

[0070] The present invention realizes objective and quantitative diagnosis of AD patients by constructing a brain structure parameter model of AD patients, providing strong support for early detection, personalized treatment and in-depth research of AD.

[0071] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A method for AD risk assessment based on multi-level iterative regression training, characterized in that: The following steps are involved: Constructing an abnormal brain structure model, including: calculating the probability of each abnormal brain structure appearing in all confirmed AD patients; selecting representative abnormal brain structure types of AD patients according to the probability value of each abnormal brain structure appearing in all confirmed AD patients and the credibility of pre-statistics; sorting the abnormal brain structure types of AD patients and giving scores; obtaining the entropy of the abnormal brain structure types of AD patients, and selecting the number of abnormal brain structure types according to the entropy value of the abnormal brain structure types of AD patients; obtaining an abnormal brain structure model through multiple trainings; AD risk is assessed based on abnormal brain structure models.

2. The AD risk assessment method based on multi-level iterative regression training according to claim 1, characterized in that: The probability of each abnormal brain structure appearing in all confirmed AD patients is calculated using the formula: P(x i )=x i / k; In the above formula, P(x i ) is the probability of each abnormal brain structure appearing in all confirmed AD patients; k is the number of confirmed AD patients; x i The number of abnormal brain structures of AD patients confirmed as i, x i ≤n; n represents the maximum number of abnormal brain structures confirmed in AD patients.

3. The AD risk assessment method based on multi-level iterative regression training according to claim 2, characterized in that: Entropy of abnormal brain structure types in AD patients: In the above formula, H(X) is the entropy of abnormal brain structure types in AD patients.

4. The AD risk assessment method based on multi-level iterative regression training according to claim 1, characterized in that: Before constructing the abnormal brain structure model, it also includes: constructing a brain structure model for healthy people: using the LinearRegression regression model, performing regression calculations on each normal brain structure data, and establishing a brain structure model for healthy people through continuous training. The brain structure model for healthy people gives the relative relationship between brain structure-related features, which is used to compare and identify the abnormal brain structure of AD patients.

5. The AD risk assessment method based on multi-level iterative regression training according to claim 4, characterized in that: The AD risk assessment includes: Collect the subjects' brain structure data and perform preprocessing; Compare the MRI data of the subjects with the brain structure model of healthy people to determine whether there are abnormal brain structures; If there is an abnormal brain structure, the type of abnormal brain structure is obtained based on the brain structure parameter model of AD patients, and its score is calculated. If the score exceeds the preset credibility allowable value, it is determined to be an AD patient with a certain credibility.

Images