KR102340829B1 - APPARATUS AND METHOD FOR PREDICTING CEREBRAL β-AMYLOID BURDEN BASED ON MACHINE LEARNING - Google Patents

Abstract

A method for predicting amyloid beta (β-Amyloid, Aβ) accumulation includes: learning a predictive model using demographic information, neuropsychological test results, and amyloid tomography (PED) image data pairs of a plurality of subjects; Receiving demographic information and neuropsychological test results, and predicting the degree of amyloid beta (Aβ) accumulation of the subject using the learned predictive model.

Description

Method and apparatus for predicting amyloid beta accumulation based on machine learning

The present invention relates to a method and apparatus for predicting amyloid beta accumulation based on machine learning, and more particularly, to a method and apparatus for predicting amyloid beta accumulation based on machine learning using demographic information and neuropsychological test results will be.

Alzheimer's disease (AD) is the most common cause of dementia and accounts for about 70% of dementia cases. The deposition of amyloid beta (β-Amyloid, Aβ) in the brain is a typical feature of Alzheimer's disease, and begins to accumulate about 10-20 years before the onset of Alzheimer's disease clinically.

Finding the amyloid beta (Aβ) that accumulates in the brain at the presymptomatic stage has very important implications. The reason for this is that these interventions allow for the identification of populations that may need anti-amyloid therapy, thereby providing an opportunity to slow or prevent the onset of Alzheimer's disease.

Until now, conventional methods for detecting amyloid beta (Aβ) deposition in the brain using amyloid positron emission tomography (PET) or finding out through a cerebrospinal fluid test have been used.

However, this conventional method has several problems. For example, the use of amyloid tomography (PET) to confirm amyloid beta (Aβ) deposition is expensive, difficult to use except in hospitals that operate specialized medical centers, such as tertiary hospitals, as well as radiation There are risk factors such as exposure to Cerebrospinal fluid analysis has disadvantages of using invasive lumbar puncture, labor required for work and analysis, and there is a difference in reliability depending on the institution.

The present invention was devised in view of the above-mentioned necessity, and using a neuropsychological test and demographic data using machine learning techniques, amyloid beta (Aβ) deposition level and neuropsychological tests and demographic variables An object of the present invention is to provide a method and apparatus for constructing a predictive model according to the relationship and predicting the accumulation of amyloid beta (Aβ) by using the predictive model.

The technical problems to be achieved by the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those of ordinary skill in the art to which the present invention belongs from the description below. will be able

In order to achieve the above object, according to one aspect of the present invention, as a method for predicting amyloid beta (β-Amyloid, Aβ) accumulation, demographic information of a plurality of subjects, neuropsychological test results, and amyloid tomography (PED) Learning a predictive model using an image data pair, receiving demographic information and neuropsychological test results of the subject, and using the learned predictive model to determine the degree of amyloid beta (Aβ) accumulation of the subject A method for predicting amyloid beta (Aβ) accumulation is provided, comprising the step of predicting.

According to an embodiment of the present invention, the step of learning the predictive model includes constructing a pair of demographic information, neuropsychological test results, and amyloid tomography (PED) image data of each of the plurality of objects, the data The method may include calculating a coefficient using an adaptive LASSO algorithm using a pair, and evaluating the accuracy of a predictive model using the calculated coefficient.

According to an embodiment of the present invention, the demographic information may include one or more of age, gender, number of years of education, and apolipoprotein E (APOE) genotype information.

According to an embodiment of the present invention, the neuropsychological test result may include one or more of an episodic memory test, a visual-spatial test, an attention test, an executive function test, a language test, and a general cognitive function test.

According to another aspect of the present invention, there is provided an apparatus for predicting amyloid beta (Aβ) accumulation, comprising one or more processors, wherein the one or more processors include demographic information, neuropsychological test results, and amyloid tomography of a plurality of subjects. (PED) learning a predictive model using an image data pair, receiving demographic information and neuropsychological test results of the subject, and using the learned predictive model to determine the degree of amyloid beta (Aβ) accumulation of the subject A device for predicting amyloid beta (Aβ) accumulation, characterized in that the prediction is provided.

According to an embodiment of the present invention, the one or more processors configure a pair of demographic information, neuropsychological test result, and amyloid tomography (PED) image data of each of the plurality of objects, and adaptive using the data pair A coefficient may be calculated using the LASSO algorithm, and the accuracy of the prediction model may be evaluated using the calculated coefficient.

According to another aspect of the present invention, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the above-described method for predicting amyloid beta (Aβ) accumulation.

The device and method proposed by the present invention are inexpensive compared to conventional methods, are readily available in general hospitals other than tertiary hospitals, and can confirm cerebral amyloid beta (Aβ) deposition without risk factors such as exposure to radiation. have. In addition, by using a predictive model based on machine learning to predict amyloid beta (Aβ) accumulation in a non-invasive way with demographic information and psychopsychological test results, the labor required for work and analysis is reduced, and the same The difference in reliability due to the use of a predictive model can be reduced. Through this, it is possible to apply the Alzheimer's disease neuropathology early diagnosis technology in primary and secondary clinics using the method and device for predicting whether amyloid positivity by non-invasive amyloid beta (Aβ) accumulation proposed in the present invention is, and false- It is possible to minimize the number of positives, thereby saving in cost, and increasing the possibility of filtering out subjects suitable for clinical trials through simple score combinations. Ultimately, the present invention can be applied to the field of early diagnosis of Alzheimer's disease.

The effects obtainable in the present invention are not limited to the above-mentioned effects, and other effects not mentioned may be clearly understood by those of ordinary skill in the art to which the present invention belongs from the following description. will be.

1 shows a conceptual diagram according to an embodiment of the present invention.
2 is a conceptual diagram illustrating predictive model learning using demographic information and a neuropsychological test according to an embodiment of the present invention.
3 is a flowchart illustrating a learning process of a predictive model using a data set according to an embodiment of the present invention.
4 is a flowchart for predicting amyloid beta accumulation using a learned prediction model according to an embodiment of the present invention.

Advantages and features of the present invention, and a method for achieving them will become apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms, and only these embodiments allow the disclosure of the present invention to be complete, and common knowledge in the technical field to which the present invention belongs It is provided to fully inform the possessor of the scope of the invention, and the present invention is only defined by the scope of the claims. Like reference numerals refer to like elements throughout.

In describing the embodiments of the present invention, if it is determined that a detailed description of a well-known function or configuration may unnecessarily obscure the gist of the present invention, the detailed description will be omitted, and the terms to be described later are used in the embodiment of the present invention. As terms defined in consideration of the function of Therefore, the definition should be made based on the content throughout this specification.

Each block in the accompanying block diagram and combinations of steps in the flowchart may be executed by computer program instructions (execution engine), which are executed by a processor of a general-purpose computer, special-purpose computer, or other programmable data processing equipment. It may be mounted so that the instructions, which are executed by the processor of a computer or other programmable data processing equipment, create means for performing the functions described in each block of the block diagram or in each step of the flowchart.

These computer program instructions may also be stored in a computer usable or computer readable memory which may direct a computer or other programmable data processing equipment to implement a function in a particular manner, and thus the computer usable or computer readable memory. It is also possible to produce an article of manufacture containing instruction means for performing the functions described in each block of the block diagram or each step of the flowchart, the instructions stored in the block diagram.

And since the computer program instructions may be mounted on a computer or other programmable data processing equipment, a series of operational steps is performed on the computer or other programmable data processing equipment to create a computer-executable process to create a computer or other programmable data processing equipment. It is also possible that the instructions for performing the data processing equipment provide steps for carrying out the functions described in each block of the block diagram and each step of the flowchart.

In addition, each block or step may represent a module, segment, or portion of code comprising one or more executable instructions for executing specified logical functions, and in some alternative embodiments the blocks or steps referred to in some alternative embodiments. It should be noted that it is also possible for functions to occur out of sequence. For example, it is possible that two blocks or steps shown one after another may be performed substantially simultaneously, and also the blocks or steps may be performed in the reverse order of the corresponding functions, if necessary.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the embodiments of the present invention illustrated below may be modified in various other forms, and the scope of the present invention is not limited to the embodiments described below. Examples of the present invention are provided to more completely explain the present invention to those of ordinary skill in the art.

Alzheimer's disease (AD) is the most common cause of dementia and accounts for about 70% of dementia cases. The deposition of amyloid beta (Aβ) in the brain is a typical feature of Alzheimer's disease, and it begins to accumulate about 10-20 years before the onset of Alzheimer's disease clinically.

Finding the amyloid beta (Aβ) that accumulates in the brain during symptom onset has very important implications. The reason for this is that these interventions make it possible to identify populations that may need anti-amyloid treatment, thereby providing an opportunity to slow or prevent the onset of Alzheimer's disease.

Until now, conventional methods for detecting amyloid beta (Aβ) deposits in the brain using amyloid positron tomography (PET) or finding out through a cerebrospinal fluid test have been used. However, this conventional method has several problems. For example, the use of amyloid tomography (PET) to confirm amyloid beta (Aβ) deposition is expensive, difficult to use except in hospitals that operate specialized medical centers, such as tertiary hospitals, as well as radiation There are risk factors such as exposure to Cerebrospinal fluid analysis has the disadvantages of using invasive lumbar puncture, requires labor for work and analysis, and there is a difference in reliability depending on the institution. Therefore, the development of a new method that is less invasive, less costly, and readily available to all institutions may help to more effectively detect amyloid beta (Aβ) deposition.

A cost-effective and sensitive neuropsychological test for clinically assessing Alzheimer's disease (AD) will help screen out those with preclinical Alzheimer's disease (AD) from among the stages of dementia. Although the application of neuropsychological tests to the task of screening high-risk groups with Alzheimer's disease (AD) neuropathology has several advantages, the relationship between specific cognitive function and amyloid beta (Aβ) using neuropsychological tests to date has been used. has been confirmed only limitedly, and has not been specifically identified. In addition, the use of neuropsychological test scores to predict cerebral amyloid beta (Aβ) deposition, not for the purpose of comparing the high and low amyloid beta (Aβ) group, is also rarely conducted. It was limited.

Accordingly, the present invention intends to propose a method and apparatus for predicting amyloid beta (Aβ) accumulation using demographic information and neuropsychological tests, which will be described below with reference to FIGS. 1 to 4 .

1 shows a conceptual diagram according to an embodiment of the present invention. In addition, FIG. 2 shows a conceptual diagram of predictive model learning using demographic information and a neuropsychological test according to an embodiment of the present invention. Referring to FIG. 1 , a conceptual diagram for predicting a machine learning-based amyloid beta (Aβ) accumulation state using demographic information and neuropsychological test information is shown. Referring to FIG. 2 , specifically, a conceptual diagram of learning a predictive model by extracting coefficients and significance of a predictive model using demographic information and neuropsychological test information to predict accuracy is shown.

The machine learning-based prediction model proposed in the present invention matches the patient's demographic information and neuropsychological test results with amyloid tomography (PET) test results, and amyloid tomography (PET) results from demographic information and neuropsychological test results. ) can determine the factors affecting the results. A machine learning-based predictive model is trained to predict the amyloid tomography (PET) result, that is, the accumulation state of amyloid beta (Aβ) from these factors.

According to an embodiment of the present invention, demographic information may include one or more of age, gender, education year, and Apolipoportein E (APOE) genotype information.

In addition, neuropsychological test information includes episodic memory test, visual-spatial ability test, attention and executive function test, language test, general cognition It may include one or more of a Global Cognition Function Test and Premorbid IQ. Specifically, episodic memory tests include logical memory (immediate/delayed recall), Rey auditory verbal learning test (RAVLT; total, delayed memory, cognition, list B). etc. The visual-spatial test may be a clock drawing test (clock/copy) The attention and executive function test may be trail makings A/B (TMT A/B) Language tests include category fluency (animals), Boston naming test (BNT), general cognitive function tests, ADAS-cog, mini-mental state examination, MMSE), Montreal cognitive assessment (MoCA), etc. The pre-onset IQ may be the American national adult reading test (ANART).

The amyloid tomography (PET) result is a Flobetapir amyloid tomography (PET) test result, indicating whether amyloid is positive. Amyloid positivity means a pathologically abnormal amyloid accumulation state.

Specifically, as shown in FIG. 2 , an embodiment for collecting data for learning a predictive model of the present invention, calculating coefficients and evaluating accuracy is as follows.

data collection

According to an embodiment of the present invention, a total of 144 Alzheimer's disease (AD), 332 mild cognitive impairment (MCI) of the Alzheimer's disease Neuroimaging Initiative (ANDI) database (http://www.loni.ucla.edu/ADNI/) ) and 286 normal cognition (NC) subjects were investigated. The data information collected for specific machine learning is shown in <Table 1>.

Here, APOE ε4 means apolipoproten, Aβ means amyloid-beta, CN (clinically normal), SMC (subjective memory concerns), EMCI (early mild cognitive impairment), LMCI (late mild cognitive impairment), AD ( Alzheimer's disease). Unless otherwise indicated, it was expressed in the form of mean (standard deviation). For APOE ε4, it indicates the percentage of people who have at least one APOE ε4 gene.

The amyloid beta (Aβ) accumulation (Aβ positivity) of the data set in <Table 1> was visualized using the corresponding Flobetabpir-PET of each subject. For this purpose, semi-quantitative PET results were retrieved from the latest available data set ("UCBERKELEYAV45_11_14_17.csv"). Florbetapir images consist of 4 × 5 min frames acquired 50–70 min post-injection. Images were reconstructed, averaged, adjusted to a common voxel size (1.5 mm), and smoothed to a common resolution of up to 8 mm to half. Structural T1-weighted images acquired concurrently with baseline Florbetapir images are used as structural templates to define regions of interest (ROIs) and reference regions for each subject using FreeSurfer software. Baseline florbetapir scans for each subject are co-registered with baseline T1-weighted MRI images. Afterwards, the standardized uptake value (SUV) is extracted from gray matter within the four cerebral ROIs (frontal, anterior/posterior cingulate, lateral parietal, and lateral temporal) and the average value is obtained. Detailed information is available on the ADNI website (http://www.loni.ucla.edu/ADNI/).

Calculation of amyloid beta (Aβ) accumulation (Aβ positivity)

The cerebral SUV ratio (SUVR) is a value obtained by normalizing the cerebral SUV using the average absorption value compared to the entire cerebellar reference area. Participants were classified as cerebral amyloid beta (Aβ) positive if the Florbetapir SUV ratio was greater than 1.1.

Coefficient calculation, significance calculation, and accuracy calculation using Adaptive LASSO algorithm

As an alternative marker for amyloid beta (Aβ) accumulation, an adaptive Least Absolute Shirinkage and Selection Operator (LASSO) model using demographic information and neuropsychological test scores was used. The adaptive LASSO equation used according to an embodiment of the present invention is as shown in Equation 1 below.

A classification model was created for a data set of 762 people, and the coefficients of each variable were calculated, and then significance was calculated. In addition, 10-fold cross validation was performed to evaluate the classifier produced by measuring the accuracy of amyloid beta (Aβ) accumulation classification. The cross-validation process was repeated 1,000 times in each classifier to avoid selection bias.

Coefficient calculation and significance evaluation

According to an embodiment of the present invention, coefficients calculated using the above-described data set and the adaptive LASSO algorithm may be obtained as shown in <Table 2>.

Among the variables, variables whose coefficient was not 0.00 were significant variables in predicting amyloid beta (Aβ) accumulation, and the degree was expressed as a coefficient value. In other words, variables with a coefficient of 0.00 can be considered to have 0 significance and do not have an effect on predicting amyloid beta (Aβ) accumulation.

Classification Accuracy Assessment

In order to measure the classification accuracy and compare it with the availability of the proposed value, a binary classifier was constructed using the combination of the coefficient values of 20 variables through adaptive LASSO processing. This predictive model performance is obtained as shown in Table 3 below, the left side is the training set application result, and the right side is the verification set application result. Calculation and evaluation of coefficients may be repeated until the predictive model performs above a certain level.

As shown in Table 3, the AUC (are under the curve) of the ROC (receiver operating characteristic) curve was used as an index of the prediction performance. AUC is a value obtained by obtaining the base area of the ROC curve, and may be calculated using Equation (2).

는 positive로 분류된 모든 예들의 rank의 합을 나타내고,

는 데이터에서 positive 값을 지닌 값들의 수(예를 들어, Aβ +)를 의미하며,

는 데이터에서 negative 값을 지닌 값들의 수(예를 들어, Aβ -)를 의미한다. here,

represents the sum of the ranks of all examples classified as positive,

means the number of positive values in the data (eg, Aβ +),

is the number of negative values (eg, Aβ -) in the data.

3 is a flowchart illustrating a learning process of a predictive model using a data set according to an embodiment of the present invention. For example, a flowchart of a process of learning a predictive model of amyloid beta (Aβ) accumulation using demographic information, neuropsychological test information, and each pair of amyloid tomography (PET) image data from a plurality of patients is shown. do. In this process, the operation may be performed by one or more processors in the apparatus for predicting amyloid beta (Aβ) accumulation, the apparatus for predicting Alzheimer's disease (AD), or an application related thereto. The operations of the method for predicting amyloid beta (Aβ) accumulation to be described below are stored in a computer readable storage medium recording a program, and when the instructions are executed by at least one processor, at least one of the operations of the method for predicting amyloid beta (Aβ) accumulation It can be set to perform steps.

Referring to FIG. 3 , the learning process of a predictive model using a data set includes a data set setting step for machine learning (S310), a coefficient calculation step using the Adaptive LASSO algorithm (S320), and the accuracy of the prediction model using the calculated coefficients. It includes an evaluation step (S330).

First, a data set for machine learning is set (S310). For example, a data set is configured by matching demographic information and neuropsychological test results of a plurality of patients or subjects with an amyloid tomography (PET) test result of each patient or subject.

Next, coefficients are calculated using the Adaptive LASSO algorithm (S320). For example, factors affecting demographic information and amyloid tomography (PET) test results among neuropsychological tests can be extracted by using the Adaptive LASSO algorithm as in <Equation 1>. At this time, it is possible to calculate the correlation coefficient between the demographic information and various factors of neuropsychological examination and the amyloid-positive result of the amyloid tomography (PET) test, and a factor whose calculated coefficient is not 0 is judged as a significant factor can do.

Finally, the accuracy of the prediction model using the calculated coefficients is evaluated ( S330 ). For example, the predictive performance of the predictive model can be calculated as shown in <Table 3> using the coefficients calculated in the previous step, and the learning process of coefficient calculation and evaluation may be repeated until the predictive model achieves performance above a certain level. can

4 is a flowchart for predicting amyloid beta (Aβ) accumulation using a learned prediction model according to an embodiment of the present invention. For example, a flowchart of a process for predicting amyloid beta (Aβ) accumulation using demographic information, neuropsychological test information, and their respective pairs of amyloid tomography (PET) image data from a plurality of patients is shown. In this process, the operation may be performed by one or more processors in the apparatus for predicting amyloid beta (Aβ) accumulation, the apparatus for predicting Alzheimer's disease (AD), or an application related thereto. The operations of the method for predicting amyloid beta (Aβ) accumulation are stored in a computer readable storage medium having a program recorded thereon, so that when the instructions are executed by at least one processor, at least one of the operations of the method for predicting amyloid beta (Aβ) accumulation is performed. can be set to perform.

4, the process of predicting amyloid beta (Aβ) accumulation is a predictive model learning step using a data set (S410), demographic information and neuropsychological test result input step (S420), and using the learned predictive model amyloid beta (Aβ) accumulation prediction step (S430).

First, a prediction model using a data set is learned (S410). For example, the coefficients of the predictive model are calculated using a data set configured by matching demographic information and neuropsychological test results of a plurality of patients or subjects with the amyloid tomography (PET) test results of each patient or subject, and Learn by evaluating accuracy. Learning of the predictive model using the data set may be performed through the process illustrated in FIG. 3 described above.

Next, demographic information and neuropsychological test results are input (S420). That is, demographic information and neuropsychological test results of subjects who want to predict amyloid beta (Aβ) accumulation are input. At this time, for example, only significant factors may be input according to the result of calculating the significant factors of the predictive model in the machine learning process of the previous step.

Finally, amyloid beta (Aβ) accumulation is predicted using the learned prediction model (S330). For example, amyloid beta (Aβ) accumulation is predicted from the subject's demographic information and neuropsychological test results using a machine learning-based predictive model. Predicting amyloid beta (Aβ) accumulation may mean predicting whether amyloid positive or negative.

As described above, the invention of Alzheimer's disease (AD) can be predicted through the prediction of amyloid beta (Aβ) accumulation using only non-invasive demographic information and neuropsychological test results, and this can lead to amyloid beta (Aβ) accumulation such as Alzheimer's disease. This can help in early diagnosis of known diseases.

In the specific embodiments described above, elements included in the invention are expressed in the singular or plural according to the specific embodiments presented. However, the singular or plural expression is appropriately selected for the situation presented for convenience of description, and the above-described embodiments are not limited to the singular or plural component, and even if the component is expressed in plural, it is composed of the singular or , even a component expressed in the singular may be composed of a plural.

On the other hand, although specific embodiments have been described in the description of the invention, various modifications are possible without departing from the scope of the technical idea contained in the various embodiments. Therefore, the scope of the present invention should not be limited to the described embodiments and should be defined by the following claims as well as the claims and equivalents.

Claims (9)

A method for predicting amyloid beta (β-Amyloid, Aβ) accumulation, comprising:
learning a predictive model by using a pair of demographic information, neuropsychological test results, and amyloid tomography (PED) image data of a plurality of subjects;
receiving demographic information and neuropsychological test results of the subject; and
Predicting the degree of amyloid beta (Aβ) accumulation in the subject using the learned predictive model,
Learning the predictive model comprises:
constructing pairs of demographic information, neuropsychological test results, and amyloid tomography (PED) image data of each of the plurality of subjects;
calculating coefficients using an adaptive LASSO algorithm using the data pair; and
A method for predicting amyloid beta (Aβ) accumulation, comprising the step of evaluating the accuracy of a prediction model using the calculated coefficient. delete According to claim 1,
The demographic information, a method of predicting amyloid beta (Aβ) accumulation, characterized in that it comprises one or more of age, sex, number of years of education, and APOE (Apolipoprotein E) genotype information. According to claim 1,
The neuropsychological test result, episodic memory test, visual-spatial test, attention test, executive function test, speech test, characterized in that it includes one or more of the overall cognitive function test, amyloid beta (Aβ) accumulation prediction method . A device for predicting amyloid beta (β-Amyloid, Aβ) accumulation, comprising:
one or more processors;
The one or more processors learn a predictive model using demographic information, neuropsychological test results, and amyloid tomography (PED) image data pairs of a plurality of subjects, and obtain demographic information and neuropsychological test results of the subject. Receive input and predict the degree of amyloid beta (Aβ) accumulation of the subject using the learned predictive model,
The one or more processors configure pairs of demographic information, neuropsychological test results, and amyloid tomography (PED) image data of each of the plurality of subjects, and calculate coefficients using an Adaptive LASSO algorithm using the data pairs, , An apparatus for predicting amyloid beta (Aβ) accumulation, characterized in that the accuracy of the prediction model is evaluated using the calculated coefficient. delete 6. The method of claim 5,
The demographic information, age, gender, number of years of education, APOE (Apolipoprotein E) genotype information, characterized in that it comprises one or more, amyloid beta (Aβ) accumulation prediction device. 6. The method of claim 5,
The neuropsychological test results, episodic memory test, visual-spatial test, attention test, executive function test, speech test, characterized in that it includes one or more of the overall cognitive function test, amyloid beta (Aβ) accumulation prediction device . A non-transitory computer-readable storage medium comprising:
A storage medium storing computer instructions for causing the computer to perform the method according to any one of claims 1, 3 and 4.

Images