Brain age prediction method, system and storage medium

Abstract

The present invention belongs to the technical field of brain age prediction, and specifically discloses a brain age prediction method, system, and storage medium. The method comprises: collecting EEG data of a subject in a resting state; inputting the EEG data into three pre-trained models to obtain three predicted ages, namely predicted age I, predicted age II, and predicted age III; judging whether the three predicted ages all belong to the same age group, and processing them according to the situation. The present invention integrates the prediction results of the three models and is committed to combining the advantages of the three models to make the prediction results more accurate.

Description

Technical Field

[0001] The present invention belongs to the technical field of brain age prediction, and in particular, relates to a brain age prediction method and system, and a computer-readable storage medium. Background Art

[0002] A person's age can be divided into chronological age (calculated by life span), physiological age (the age of the functions of other important organs besides the brain), and brain age (BA), which is used to assess the degree of brain aging. The brain gradually shrinks and declines in function with age, and the speed and trajectory of this change vary significantly between brain regions and individuals. The concept of differential brain aging, for example, that a 50-year-old can "have a 20-year-old heart" or that a 30-year-old smoker's lungs "work like an 80-year-old", is clinically demonstrated and is gaining scientific traction. Similarly, while a decline in brain function is a "normal" part of aging, some people clearly age faster than others. The degree to which an individual's brain aging trajectory deviates from the typical brain aging trajectory may help with diagnosis, treatment decisions, prognosis, and early identification of brain-based diseases, especially neurodegenerative diseases.

[0003] At present, there is no standard method or device for measuring brain age in clinical practice. Its judgment is mainly based on the doctor's subjective judgment of the brain function of individual subjects, including cognitive function (logical thinking function, recognition function, etc.), motor function, sensory function (hearing, vision, etc.). Secondly, some intelligence scales are also used in clinical practice to judge the cognitive function of individual subjects. Although this method quantifies the standard, it can only reflect one aspect of brain function and cannot fully represent brain age. Recently, magnetic resonance imaging (MRI) studies have identified structural biomarkers of aging, such as axonal demyelination and reduction in gray and white matter volume, which help estimate the brain age index of general nervous system health and function. However, there are still two limitations in the practical application of structural MRI to predict brain age. One limitation is that biomarkers of brain structural aging do not always correlate well with cognitive function. This is exemplified by the little-known phenomenon of cognitive reserve and neural reserve, that is, brains that show Alzheimer's disease-related amyloid pathology and structural damage can also maintain a high level of function. Another limitation is cost. It is well known that MRI is expensive, and it would be very expensive to conduct extensive early and repeated screening and it is unlikely to be covered by insurance companies.

[0004] EEG signals, as an important aspect that objectively reflects the functional state of the brain, are closely related to age. EEG can observe the development and changes of brain waves with age. Brain waves can show obvious differences at different age stages. For example, the EEG of infants and young children reacts more significantly to the influence of various internal and external factors than adults, and is prone to obvious brain wave abnormalities, and the range of abnormalities is also wider, but the corresponding disappearance is also faster than that of adults. By adulthood, brain waves gradually stabilize, and after middle age, as brain function gradually declines, brain waves undergo corresponding changes. In addition, EEG images not brain structure, but brain function, and has been proven to be effective in capturing the contribution of cognitive reserve to brain health and cognitive function. The hardware cost of EEG is relatively affordable, which is convenient for extensive screening and multi-scenario use. Summary of the invention

[0005] The purpose of the present invention is to provide a brain age prediction method and system, and a computer-readable storage medium, which partially solve or alleviate the above-mentioned deficiencies in the prior art, and integrate the prediction results of the three models so that the final predicted age is more consistent with the actual brain age.

[0006] In order to solve the above-mentioned technical problems, the present invention specifically adopts the following technical solutions: a method for predicting brain age, comprising:

[0007] Collect EEG data of the subjects in resting state;

[0008] The EEG data is input into three pre-trained models to obtain three predicted ages, namely predicted age I, predicted age II, and predicted age III;

[0009] When the three predicted ages all belong to the same age group, data processing method I is used to obtain the final predicted value;

[0010] In the case where only two of the three predicted ages belonged to the same age group, data processing method II was used to obtain the predicted brain age;

[0011] The data processing method I comprises:

[0012] When the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold I, the formula is used:

[0013] A r =(A I +A II +A III ) / 3;

[0014] Calculate the final predicted age;

[0015] When the difference between the maximum and minimum values ​​of the three predicted ages is greater than the age difference threshold I, the credibility of the three predicted ages is calculated and weights are assigned to the three predicted ages according to the credibility;

[0016] Using the formula:

[0017] A r =k1*A I +k2A II +k3*A III ;

[0018] Calculate the final predicted age;

[0019] The data processing method II comprises:

[0020] When the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold II, the formula is used:

[0021] A r =(A I +A II +A III ) / 3;

[0022] Calculate the final predicted age; otherwise, re-collect the resting EEG data of the subject for re-prediction;

[0023] Among them, A r is the final predicted value, A I , A II , A III are predicted age I, predicted age II, and predicted age III respectively, k1, k2, and k3 are the weights of the three predicted ages respectively, and age difference threshold II is less than age difference threshold I.

[0024] As an improvement, the method for calculating the reliability of the predicted age includes using the formula:

[0025] d=(s t -s p ) / s t ;

[0026] Calculate the reliability of the predicted age; where d is the reliability of the current predicted age, s t is the sum of the absolute values ​​of the differences between the three predicted ages, s p It is the sum of the absolute values ​​of the differences between the current predicted age and the other two predicted ages.

[0027] As an improvement, a method of using confidence to weight the predicted age includes using the formula:

[0028] k=d / d t ;

[0029] Calculate the weight of the predicted age, where k is the weight of the current predicted age, d is the credibility of the current predicted age, and d t is the sum of the three predicted age reliabilities.

[0030] As an improvement, when collecting EEG data of the subject in a resting state, a mobile wireless EEG headset is used for collection; the mobile wireless EEG headset uses 14 salt-soaked felt sensors and 2 reference sensors to collect EEG data, and records using a 0.2-45.0 Hz bandpass filter and a 50 Hz or 60 Hz notch filter.

[0031] As an improvement, the positions of the 14 salt-impregnated felt sensors are AF3 / 4, F3 / 4, F7 / 8, FC5 / 6, T7 / 8, P7 / 8 and O1 / 2 respectively; the position of the 2 reference sensors is P3 / 4.

[0032] As an improvement, when collecting EEG data, the subjects alternated between opening and closing their eyes at 1-minute intervals; and three groups of 4 minutes were collected in total.

[0033] As an improvement, the three models are a radial basis kernel support vector regression model, an elastic net regression model, and a radial basis function kernel Gaussian process model.

[0034] As an improvement, the EEG data of children under 12 years old, patients with psychiatric and brain diseases, and people with bad habits were excluded from the dataset when training the model.

[0035] As an improvement, the age groups are divided into 13 to 20 years old, 21 to 30 years old, 31 to 60 years old, and 61 years old and above.

[0036] The present invention also provides a brain age prediction system, comprising:

[0037] An EEG data acquisition module is used to acquire EEG data of the subject in a resting state;

[0038] A prediction module is used to input EEG data into three pre-trained models to obtain three predicted ages, namely predicted age I, predicted age II, and predicted age III;

[0039] The final predicted age acquisition module is used to obtain the final predicted value by using data processing method I when the three predicted ages all belong to the same age group;

[0040] In the case where only two of the three predicted ages belonged to the same age group, data processing method II was used to obtain the predicted brain age;

[0041] The data processing method I comprises:

[0042] When the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold I, the formula is used:

[0043] A r =(A I +A II +A III ) / 3;

[0044] Calculate the final predicted age;

[0045] When the difference between the maximum and minimum values ​​of the three predicted ages is greater than the age difference threshold I, the credibility of the three predicted ages is calculated and weights are assigned to the three predicted ages according to the credibility;

[0046] Using the formula:

[0047] A r =k1*A I +k2A II +k3*A III ;

[0048] Calculate the final predicted age;

[0049] The data processing method II comprises:

[0050] When the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold II, the formula is used:

[0051] A r =(A I +A II +A III ) / 3;

[0052] Calculate the final predicted age; otherwise, re-collect the resting EEG data of the subject for re-prediction;

[0053] Among them, A r is the final predicted value, A I , A II , A III are predicted age I, predicted age II, and predicted age III respectively, k1, k2, and k3 are the weights of the three predicted ages respectively, and age difference threshold II is less than age difference threshold I.

[0054] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which implements the above-mentioned brain age prediction method when executed by a processor.

[0055] The present invention is beneficial in that:

[0056] The advantages of EEG-based brain age measurement are: (1) EEG-based brain age can reflect functional changes rather than structural changes; (2) the data is more objective and convincing; (3) EEG is more participant-friendly, has fewer costs and contraindications, and can be measured in principle through home devices; (4) EEG-based brain age can facilitate repeated measurements of participants to evaluate the effectiveness of interventions, such as drugs or brain stimulation aimed at maintaining and improving brain function and lifespan. Therefore, in measuring brain age, machine learning can be used to provide an easily accessible and low-cost biomarker of brain health.

[0057] On the basis of the above, the present invention combines the prediction results of the three models and processes the three predicted ages through technical means to obtain a predicted age with higher credibility.

[0058] First, determine whether the three predicted ages are in the same age group, and then handle them according to the situation. Brain age is divided into 4 stages, namely 13 to 20 years old, 21 to 30 years old, 31 to 60 years old, and 61 years old and above. The EEG data of the above four age groups will be more obvious. By judging whether they are in the same age group, the three predicted ages can be preliminarily judged for their credibility.

[0059] If the three predicted ages are not in the same age range, it means that there is a problem when collecting EEG data, and it is necessary to re-enter the backup EEG data or re-collect.

[0060] When the three predicted ages are in the same age range, it means that the predictions of the three models are roughly accurate, and only need to be processed according to data processing method I. Data processing method I has two cases. When the data is relatively compact, the average of the three is used as the final result. When the data is relatively loose, the credibility of the three predicted ages is calculated, and the credibility is used to assign weights to the three predicted ages. Finally, the proportion of the three predicted ages in the final result is adjusted by the weight, so as to obtain more scientific and accurate prediction results.

[0061] As for the case where two of the three predicted ages are in the same age group, since one of the predicted ages has already crossed the age group, we should be more cautious in this case and only adopt it when the data is more compact.

[0062] The present invention uses the method of first determining the age group and then performing case-by-case processing, comprehensively utilizing the three prediction results of the three models, and ultimately obtaining a more accurate prediction result. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions 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. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual scale. Obviously, the drawings described below are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without paying creative labor.

[0064] Figure 1 is a flow chart of the present invention;

[0065] Figure 2 This is a schematic diagram of the location for collecting EEG data;

[0066] Figure 3 It is a structural diagram of the present invention. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0068] Herein, suffixes such as "module", "component" or "unit" used to represent elements are only used to facilitate the description of the present invention, and have no specific meanings by themselves. Therefore, "module", "component" or "unit" can be used mixedly.

[0069] In this document, the terms "upper", "lower", "inner", "outer", "front", "back", "one end", "the other end" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance.

[0070] In this document, unless otherwise clearly specified and limited, the terms "installed", "provided with", "connected", etc. should be understood in a broad sense. For example, "connected" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection, a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0071] Herein "and / or" includes any and all combinations of one or more of the associated listed items.

[0072] Herein, "plurality" means two or more than two, ie, it includes two, three, four, five, etc.

[0073] Embodiment 1

[0074] Prior art Chinese patent application CN201910701028.0 discloses an age group classification method based on resting-state EEG data, comprising the following steps: S1. Collecting original resting-state EEG data corresponding to each age group; S2. Preprocessing the original resting-state EEG data to obtain resting-state EEG data with artifacts removed; S3. Constructing a convolutional neural network, and inputting the resting-state EEG data with artifacts removed into the convolutional neural network, training and testing the convolutional neural network to obtain a trained convolutional neural network; S4. Using the trained convolutional neural network to classify the actual resting-state EEG data into age groups.

[0075] In the above-mentioned prior art, the convolution model is trained by collecting resting EEG data of various age groups as a training set, and the convolution model is used to predict the test samples to obtain the patient's brain age. However, in actual use, due to the limitations of the model itself, the prediction results will be biased. On the other hand, as the age increases, the gap between brain age and actual age will become larger and larger, and the limitations of a single model will be further magnified.

[0076] To solve this problem, Figure 1 As shown, the present invention provides a brain age prediction method, which integrates the prediction results of the three models and is committed to combining the advantages of the three models to make the prediction results more reliable. The specific steps include:

[0077] S1 collects EEG data of subjects in resting state.

[0078] Figure 2The location where the subject collects EEG data is marked. In this embodiment, when collecting EEG data of the subject in a resting state, a mobile wireless EEG headset is used for collection; the mobile wireless EEG headset uses 14 saline felt sensors and 2 reference sensors to collect EEG data, and uses a 0.2-45.0 Hz bandpass filter and a 50 Hz or 60 Hz notch filter for recording.

[0079] More specifically, the positions of the 14 salt-impregnated felt sensors are AF3 / 4, F3 / 4, F7 / 8, FC5 / 6, T7 / 8, P7 / 8 and O1 / 2; the position of the 2 reference sensors is P3 / 4.

[0080] In addition, when collecting EEG data, the subjects alternated between opening and closing their eyes at 1-minute intervals, and collected three sets of 4 minutes for a total of 12 minutes, including 6 minutes of eyes open and 6 minutes of eyes closed. A short rest was allowed between the two collections to reduce drowsiness.

[0081] Since EEG data is seriously affected by short-term interference from the outside world, in this embodiment, multiple groups are collected in advance to avoid repeated collection.

[0082] S2 inputs the EEG data into three pre-trained models to obtain three predicted ages, namely predicted age I, predicted age II, and predicted age III.

[0083] In this embodiment, the three models used to provide predictions are a radial basis kernel support vector regression model, an elastic net regression model, and a radial basis function kernel Gaussian process model.

[0084] The three models have their own advantages and disadvantages when modeling and predicting:

[0085] 1. Radial basis kernel support vector regression (SVR).

[0086] advantage:

[0087] Able to handle nonlinear relationships and suitable for complex data.

[0088] It has good performance for high-dimensional data.

[0089] Usually more robust to noisy data.

[0090] shortcoming:

[0091] Choosing appropriate kernel functions and parameters (such as C and γ) is relatively complex.

[0092] Training time is long, especially on large datasets.

[0093] 2. Elastic Net Regression.

[0094] advantage:

[0095] Combining L1 (Lasso) and L2 (Ridge) regularization helps with feature selection and reducing overfitting.

[0096] Easy to use and computationally efficient.

[0097] Effective for multicollinearity problems.

[0098] shortcoming:

[0099] Suitable for linear relationships, but limited ability to model nonlinear relationships.

[0100] Selecting regularization parameters requires cross-validation, which is a cumbersome process.

[0101] 3. Radial basis function kernel Gaussian process (GPR).

[0102] advantage:

[0103] It has a good theoretical basis and can provide uncertainty estimates.

[0104] It is suitable for small data sets and can model complex nonlinear relationships.

[0105] The kernel function can be chosen flexibly.

[0106] shortcoming:

[0107] The computational complexity is high, especially when the amount of data is large, and the training and inference speeds are slow.

[0108] For large-scale datasets, the storage and computational overheads are very high.

[0109] In this embodiment, when training the above three models, the process of collecting EEG data as a data set is similar to step S1, and will not be repeated in this step. It is worth noting that when collecting data sets for training models, the EEG data of children under 12 years old, patients with psychiatric and brain diseases, and people with bad habits such as smoking, alcoholism, drug dependence, etc. need to be excluded from the data set.

[0110] For normal people, we can assume that their brain age is basically consistent with their actual age. In other words, a normal person's brain age will be within one standard deviation of their actual age. According to existing data, the standard deviation is ±7.6 years. That is, a normal 30-year-old subject's brain age will be between 22.4 and 37.6.

[0111] For normal people, the EEG data of brains of different brain ages in a resting state have certain characteristics to follow. The three models in the present invention are based on this principle to predict brain age.

[0112] It is understandable that the three groups of people excluded from the data set cannot be used as training sets and validation sets to train the model because their brain age does not match their actual age due to reasons such as immature brain development, lesions, and drugs.

[0113] Inputting the EEG data of the subjects into the above three models will result in three predicted ages, namely predicted age I, predicted age II, and predicted age III. Ideally, the three predicted ages should be consistent. But in reality, since the three models have their own advantages and disadvantages, the predicted brain ages will also be biased. Therefore, it is necessary to process the three predicted ages through technical means in the subsequent steps to obtain a predicted age with higher credibility.

[0114] S3: When the three predicted ages belong to the same age group, data processing method I is used to obtain the final predicted value;

[0115] In the case where only two of the three predicted ages belonged to the same age group, data processing method II was used to obtain the predicted brain age;

[0116] The data processing method I comprises:

[0117] When the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold I, the formula is used:

[0118] A r =(A I +A II +A III ) / 3;

[0119] Calculate the final predicted age;

[0120] When the difference between the maximum and minimum values ​​of the three predicted ages is greater than the age difference threshold I, the credibility of the three predicted ages is calculated and weights are assigned to the three predicted ages according to the credibility;

[0121] Using the formula:

[0122] A r =k1*A I +k2A II +k3*A III ;

[0123] Calculate the final predicted age;

[0124] The data processing method II comprises:

[0125] When the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold II, the formula is used:

[0126] A r =(A I +A II +A III ) / 3;

[0127] Calculate the final predicted age; otherwise, re-collect the resting EEG data of the subject for re-prediction;

[0128] Among them, A r is the final predicted value, A I , A II , A III are predicted age I, predicted age II, and predicted age III respectively, k1, k2, and k3 are the weights of the three predicted ages respectively, and age difference threshold II is less than age difference threshold I.

[0129] In the present invention, the brain age is first segmented, namely, 13 to 20 years old, 21 to 30 years old, 31 to 60 years old, and 61 years old and above.

[0130] There will be obvious differences in the EEG data of the above four age groups.

[0131] Adolescence 13-20 years old: The rhythm in the occipital region is between 8 and 12 Hz, with an average of about 10 Hz. The amplitude gradually decreases to 50-75 μV. The insertion slow wave activity in the posterior head gradually decreases, and generally the slow waves in the low-medium amplitude theta and delta frequency bands do not exceed 10%-15%. In normal adolescents, the distribution of slow waves is often more extensive, rather than mainly appearing in the posterior head. The fast wave activity at this age is more and more extensive during the awake period, with the largest number in the frontal area, and can also be seen in the central area and the topic area. μ rhythm is more common, but gradually decreases after the age of 15. Occipital lambda waves are also more common. Theta activity of 6-7 Hz in the frontal head gradually decreases after the age of 13-15. There is no low voltage in normal EEG waves before the age of 13. However, after puberty, as the EEG development matures, the amplitude of basic EEG activity decreases, of which about 5% is manifested as low voltage, often dominated by low-amplitude fast waves in the beta frequency band, which is more common in women and may be genetic.

[0132] Young people aged 21-30: After the age of 20, the brain wave development is basically mature, the occipital alpha rhythm is stable between 9 and 10Hz, the frequency range of the normal whole head alpha wave does not exceed 2Hz, and the frequency difference between the corresponding areas of the two hemispheres does not exceed 0.5Hz. The amplitude is further reduced, between 10 and 100μV, with an average of 50 to 70uV, and the amplitude fluctuates in a shuttle shape, with a relatively good amplitude modulation phenomenon. The individual differences in the alpha index are very large, ranging from 10% to 90%, generally between 60% and 70%. The low-amplitude beta activity of 14 to 30Hz during the awake period increases with age, mainly distributed in the frontal, central, anterior and middle temporal areas, and the occipital area only appears when the alpha rhythm is blocked. The low-amplitude theta activity of 4 to 7Hz is mainly distributed in the frontal and central areas. In addition, a small amount of theta activity can also be seen in the top and temporal areas, with an index of 4% to 5%. There is little slow wave activity in the 8-band, with an amplitude below 20μV, scattered in the frontal area, and an index of no more than 5%. μ rhythm can be seen in the Rolandic area, mainly distributed in the central area, sometimes also in the parietal area, and is seen in 19% of healthy adults. K (Kappa) rhythm is seen in 30% of normal adults, with a frequency of 6-12Hz and an amplitude of 10-40μV, mainly appearing in the frontal and temporal areas during mental activities. Lambda waves can be seen in the occipital area when scanning complex figures.

[0133] Middle-aged people aged 31-60: After entering middle age, brain waves are still in a relatively stable state. However, most middle-aged people are in a state of "sub-health" to varying degrees, and this characteristic is also reflected in brain waves to varying degrees. In middle-aged people with good basic health and no obvious neurological or systemic diseases, brain waves basically maintain the characteristics of normal brain waves of young people. However, when there are other organic or functional lesions of the non-neurological system, some marginal or mildly abnormal brain wave manifestations may appear, such as a decrease in the α index, a slight increase in slow wave activity, poor regulation of the α rhythm, and a slight increase in temporal 0 activity. A small number of middle-aged and elderly people over 50 years old may have a sharp θ rhythm in the temporal region, which is more common in the sleep and light sleep periods. A small number of adults can see widely distributed 5-7Hz sharp or sinusoidal waves during the wakefulness and sleep periods. They are most prominent in the top and temporal regions and are often bilateral, but can appear asymmetrically, lasting for seconds to minutes, and can appear and stop suddenly, or gradually appear and disappear. This clinical rhythmic discharge occurs in 0.5% of people over 50 years old and is a benign EEG pattern.

[0134] Elderly people aged 60 and above: Even for elderly people who are basically healthy, their nervous systems will show varying degrees of degenerative changes with aging, including a selective decrease in the number of neurons, an enlarged ventricular system, a reduction in brain weight, and visible neuronal shrinkage, lipofuscin or amyloid deposition, etc. These changes are most obvious in the frontal and temporal lobes. In addition, elderly people who are generally in good health may have certain hidden lesions, such as cerebral arteriosclerosis or hidden diabetes. These changes will affect changes in cerebral blood flow and cerebral metabolism, thereby affecting brain waves to varying degrees.

[0135] Changes in α rhythm: Elderly people often show a slower α rhythm, lower amplitude, and a reduced α index, among which the fast α of 10-13Hz is more significantly reduced than the slow α of 8-9Hz. The α rhythm of people over 60 years old is often mainly 8.5-9.5Hz, the fast α rhythm of 10-13Hz is reduced, and the basic α rhythm is 1Hz slower than that of young people on average. After entering the elderly period, the α wave slows down further with age. According to a continuous follow-up study, the frequency of the α rhythm decreases by an average of 0.08Hz per year after the age of 60. If the α rhythm in the occipital region of the elderly is lower than 8.5Hz in a fully awake state, it should be considered abnormal. In addition, with aging, the amplitude of the α rhythm decreases, mostly to about 30μV, and becomes more fragmented and discontinuous. The distribution range of the α wave in the elderly is expanded, and it can be extended to the frontal and temporal regions, so that the difference between the front and back of the head disappears. Sometimes it can be seen that the synchronization trend of the whole head is enhanced. Some people believe that this generalization of the α rhythm may also be related to reduced alertness or mild sleepiness during recording.

[0136] Changes in β activity: Normally, from birth to adulthood, β activity gradually increases with age. Before the age of 50, β activity accounts for about 20% of the awake brain waves, mainly in the frontal and central areas; after the age of 50, it accounts for about 30%. After the age of 60, β activity still increases slowly, but after the age of 80, β activity tends to decrease. The β / α ratio of the elderly increases with age and can be used as an indicator to evaluate aging.

[0137] Therefore, in the present invention, it is first determined whether the predicted ages predicted by the three models are in the same age group. Therefore, there will be three situations: all three predicted ages are in the same age group, only two predicted ages are in the same age group, and all three predicted ages are not in the same age group.

[0138] If the three predicted ages are not in the same age range, it means that there is a problem when collecting EEG data, and it is necessary to re-enter the backup EEG data or re-collect.

[0139] When the three predicted ages are in the same age range, it means that the predictions of the three models are roughly accurate and only need to be processed according to data processing method I.

[0140] In data processing method I, there are two cases:

[0141] First, the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold I. In this embodiment, the age difference threshold I is preset to 6, that is, the difference between the maximum and minimum values ​​of the predicted ages of the three models does not exceed 6, such as 20, 21, and 25.

[0142] In the above case, the three predicted ages are relatively close to each other, so it can be considered that the credibility of the three predicted ages is high. Therefore, the average of the three predicted ages is taken as the final predicted age, that is, A r =(A I +A II +A III ) / 3.

[0143] Second, the difference between the maximum and minimum values ​​of the three predicted ages is greater than the age difference threshold I. That is to say, the difference between the maximum and minimum values ​​of the predicted ages of the three models has exceeded 6, such as 20, 23, and 28.

[0144] In the above case, the three predicted ages are relatively loose, so it is necessary to calculate the credibility of the three predicted ages and assign weights to the three predicted ages according to the credibility. A high credibility has a high weight, and a low credibility has a low weight. Specifically, the formula used in this embodiment is:

[0145] A r =k1*A I +k2A II +k3*A III ; Calculate the final predicted age; The method for calculating the reliability of the predicted age includes using the formula:

[0146] d=(s t -s p ) / s t ;

[0147] Calculate the reliability of the predicted age; where d is the reliability of the current predicted age, s t is the sum of the absolute values ​​of the differences between the three predicted ages, s p It is the sum of the absolute values ​​of the differences between the current predicted age and the other two predicted ages.

[0148] For example, the predicted ages I, II, and III are 20, 23, and 28 respectively. Then the absolute values ​​of the differences between the predicted age I and the other predicted ages are 3 and 8 respectively, the absolute values ​​of the differences between the predicted age II and the other predicted ages are 3 and 5 respectively, and the absolute values ​​of the differences between the predicted age III and the other predicted ages are 8 and 5 respectively. In other words, the sum of the absolute values ​​of the differences between the three predicted ages is s t3+5+8=16. Then the reliability of predicting age I is (16-11) / 16=5 / 16, the reliability of predicting age II is 8 / 16, and the reliability of predicting age III is 3 / 16.

[0149] Then use the formula:

[0150] k=d / d t ;

[0151] Calculate the weight of the predicted age, where k is the weight of the current predicted age, d is the credibility of the current predicted age, and d t is the sum of the three predicted age reliabilities.

[0152] The final predicted value can be calculated by substituting the weights of the three predicted ages into the above formula for calculating the final predicted value.

[0153] In the case where two of the three predicted ages are in the same age group, since one of the predicted ages has already crossed the age group, this situation should be treated with more caution and should be adopted when the data is more compact. That is, the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold II, which is preset to 4, smaller than the age difference threshold I, thereby ensuring that the three predicted ages are within a smaller range.

[0154] If the conditions are met, the average of the three predicted ages is taken as the final predicted value. Otherwise, it is considered that the collected EEG data is incorrect, and it is necessary to input backup data for re-prediction or re-collect EEG data for re-prediction.

[0155] Example 2

[0156] like Figure 3 As shown, the present invention also provides a brain age prediction system, comprising:

[0157] An EEG data acquisition module is used to acquire EEG data of the subject in a resting state;

[0158] A prediction module is used to input EEG data into three pre-trained models to obtain three predicted ages, namely predicted age I, predicted age II, and predicted age III;

[0159] The final predicted age acquisition module is used to obtain the final predicted value by using data processing method I when the three predicted ages all belong to the same age group;

[0160] In the case where only two of the three predicted ages belonged to the same age group, data processing method II was used to obtain the predicted brain age;

[0161] The data processing method I comprises:

[0162] When the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold I, the formula is used:

[0163] A r =(A I +A II +A III ) / 3;

[0164] Calculate the final predicted age;

[0165] When the difference between the maximum and minimum values ​​of the three predicted ages is greater than the age difference threshold I, the credibility of the three predicted ages is calculated and weights are assigned to the three predicted ages according to the credibility;

[0166] Using the formula:

[0167] A r =k1*A I +k2A II +k3*A III ;

[0168] Calculate the final predicted age;

[0169] The data processing method II comprises:

[0170] When the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold II, the formula is used:

[0171] A r =(A I +A II +A III ) / 3;

[0172] Calculate the final predicted age; otherwise, re-collect the resting EEG data of the subject for re-prediction;

[0173] Among them, A r is the final predicted value, A I , A II , A III are predicted age I, predicted age II, and predicted age III respectively, k1, k2, and k3 are the weights of the three predicted ages respectively, and age difference threshold II is less than age difference threshold I.

[0174] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0175] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a computer terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0176] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A method for predicting brain age, characterized in that: include: Collect EEG data of the subjects in resting state; Input the EEG data into three pre-trained models to obtain three predicted ages, namely predicted age I, predicted age II, and predicted age III; the three models are radial basis kernel support vector regression model, elastic net regression model, and radial basis kernel Gaussian process model; When the three predicted ages all belong to the same age group, data processing method I is used to obtain the final predicted value; the age groups are 13 to 20 years old, 21 to 30 years old, 31 to 60 years old, and 61 years old and above; In the case where only two of the three predicted ages belonged to the same age group, data processing method II was used to obtain the predicted brain age; The data processing method I comprises: When the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold I, the formula is used: A r =(A I +A II +A III ) / 3, calculate the final predicted age; When the difference between the maximum and minimum values ​​of the three predicted ages is greater than the age difference threshold I, the credibility d of the three predicted ages is calculated, where the credibility d of the current predicted age = (s t -s p ) / s t ,s t is the sum of the absolute values ​​of the differences between the three predicted ages, s p The sum of the absolute values ​​of the differences between the current predicted age and the other two predicted ages; After assigning weights to the three predicted ages according to their credibility, use the formula: A r =k1*A I+ k2A II +k3*A III , calculate the final predicted age; Among them, the method of using credibility to assign weights to predicted age includes using the formula: k=d / d t , calculate the weight of the predicted age, where k is the weight of the current predicted age, d is the credibility of the current predicted age, and d t is the sum of the three predicted age reliability; The data processing method II comprises: When the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold II, the formula is used: A r =(A I +A II +A III ) / 3, calculate the final predicted age; otherwise, re-collect the resting EEG data of the subject for re-prediction; Among them, A r is the final predicted value, A I , A II , A III are predicted age I, predicted age II, and predicted age III respectively, k1, k2, and k3 are the weights of the three predicted ages respectively, and age difference threshold II is less than age difference threshold I.

2. A method for predicting brain age according to claim 1, characterized in that: When collecting EEG data of the subjects in a resting state, a mobile wireless EEG headset was used for collection; the mobile wireless EEG headset used 14 saline-soaked felt sensors and 2 reference sensors to collect EEG data, and recorded using a 0.2-45.0 Hz bandpass filter and a 50 Hz or 60 Hz notch filter.

3. A method for predicting brain age according to claim 2, characterized in that: The positions of the 14 salt-impregnated felt sensors are AF3 / 4, F3 / 4, F7 / 8, FC5 / 6, T7 / 8, P7 / 8 and O1 / 2 respectively; the position of the 2 reference sensors is P3 / 4.

4. A method for predicting brain age according to claim 2, characterized in that: When collecting EEG data, the subjects alternated between opening and closing their eyes at 1-minute intervals; and three sets of 4 minutes were collected in total.

5. A method for predicting brain age according to claim 1, characterized in that: When training the model, the EEG data of children under 12 years old, patients with psychiatric and brain diseases, and people with bad habits were excluded from the dataset.

6. A brain age prediction system, characterized in that: include: An EEG data acquisition module is used to acquire EEG data of the subject in a resting state; A prediction module is used to input EEG data into three pre-trained models to obtain three predicted ages, namely predicted age I, predicted age II, and predicted age III; the three models are radial basis kernel support vector regression model, elastic net regression model, and radial basis function kernel Gaussian process model; The final predicted age acquisition module is used to obtain the final predicted value by using data processing method I when the three predicted ages all belong to the same age group; the age groups are divided into 13 to 20 years old, 21 to 30 years old, 31 to 60 years old, and 61 years old and above; In the case where only two of the three predicted ages belonged to the same age group, data processing method II was used to obtain the predicted brain age; The data processing method I comprises: When the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold I, the formula is used: A r =(AI+A II +A III ) / 3, calculate the final predicted age; When the difference between the maximum and minimum values ​​of the three predicted ages is greater than the age difference threshold I, the credibility d of the three predicted ages is calculated, where the credibility d of the current predicted age = (s t -s p ) / s t ,s t is the sum of the absolute values ​​of the differences between the three predicted ages, s p The sum of the absolute values ​​of the differences between the current predicted age and the other two predicted ages; After assigning weights to the three predicted ages according to their credibility, use the formula: A r =k1*A I +k2A II +k3*A III , calculate the final predicted age; wherein, the method of using credibility to assign weights to the predicted age includes using the formula: k=d / d t , calculate the weight of the predicted age, where k is the weight of the current predicted age, d is the credibility of the current predicted age, and d t is the sum of the three predicted age reliability; The data processing method II comprises: When the difference between the maximum and minimum values ​​of the three predicted ages is less than or equal to the age difference threshold II, the formula is used: A r =(A I +A II +A III ) / 3, calculate the final predicted age; otherwise, re-collect the resting EEG data of the subject for re-prediction; Among them, A r is the final predicted value, A I , A II , A III are predicted age I, predicted age II, and predicted age III respectively, k1, k2, and k3 are the weights of the three predicted ages respectively, and age difference threshold II is less than age difference threshold I.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the brain age prediction method as described in any one of claims 1 to 5 is implemented.

Images