Method for detecting mild cognitive impairment

Abstract

A method for detecting mild cognitive impairment includes the following steps: obtaining a pre-measured pulse wave from a target subject; obtaining a post-measured pulse wave from the target subject after performing a cognitive intervention activity thereon; performing a Fourier transform on the pre-measured pulse wave to generate multiple pre-measured harmonics; performing a Fourier transform on the post-measured pulse wave to generate multiple post-measured harmonics; determining whether the target subject has mild cognitive impairment based on the variations between the pre-measured and post-measured harmonics.

Description

Methods for detecting mild cognitive impairment The present invention relates to a method for detecting mild cognitive impairment, and more particularly to a method for determining mild cognitive impairment by utilizing changes in harmonics of a pulse wave. As the global aging population intensifies, cognitive impairment is receiving increasing attention. Specifically, cognitive function gradually deteriorates with age, a normal part of aging. When the rate of cognitive decline exceeds that of normal aging, it becomes "cognitive impairment," which is divided into two types: dementia and mild cognitive impairment (MCI). Dementia, commonly known as Alzheimer's disease, is a condition in which patients' memory and cognitive functions deteriorate severely, to the point where they are unable to care for themselves. Mild cognitive impairment, on the other hand, lies between normal aging and dementia, and patients are still able to care for themselves. Mild cognitive impairment (MCI) has many causes. If it's caused by anxiety, depression, sleep disorders, vitamin B12 or folic acid deficiency, or medication side effects, cognitive function often recovers with treatment. However, some patients, especially those with simple amnesia, may develop Alzheimer's dementia. Therefore, early detection of MCI can effectively prevent and slow the progression to dementia. Conventional testing methods involve administering cognitive assessments, such as the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA). However, these tests are time-consuming and exhibit significant fluctuations in recognition sensitivity. Furthermore, due to significant linguistic and cultural differences in these tests, there is a lack of standardized, internationally calibrated norms, which can lead to floor effects. Therefore, how to provide a non-invasive and effective means of detecting cognitive impairment is an urgent problem to be solved in this field. One aspect of the present invention provides a method for detecting mild cognitive impairment, comprising the following steps: obtaining a pre-measured pulse wave of a target subject; performing a cognitive intervention activity on the target subject and then obtaining a post-measured pulse wave of the target subject; generating a plurality of pre-measured harmonics by Fourier transforming the pre-measured pulse wave; generating a plurality of post-measured harmonics by Fourier transforming the post-measured pulse wave; and determining whether the target subject has mild cognitive impairment based on changes in the pre-measured harmonics and the post-measured harmonics. According to one embodiment of the present invention, the step of determining whether the target subject has mild cognitive impairment based on the changes in the pre-test harmonics and the post-test harmonics includes determining whether the target subject has mild cognitive impairment based on the difference between the amplitude ratio of the pre-test harmonics and the amplitude ratio of the post-test harmonics. According to one embodiment of the present invention, the step of determining whether the target subject has mild cognitive impairment based on the difference between the amplitude ratios of the pre-test harmonics and the amplitude ratios of the post-test harmonics includes: determining whether the target subject has mild cognitive impairment based on a first difference between the amplitude ratios of a first-order harmonic in the pre-test harmonics and the amplitude ratios of a first-order harmonic in the post-test harmonics, a second difference between the amplitude ratios of a second-order harmonic in the pre-test harmonics and the amplitude ratios of a second-order harmonic in the post-test harmonics, a third difference between the amplitude ratios of a third-order harmonic in the pre-test harmonics and the amplitude ratios of a third-order harmonic in the post-test harmonics, and a fourth difference between the amplitude ratios of a fourth-order harmonic in the pre-test harmonics and the amplitude ratios of a fourth-order harmonic in the post-test harmonics. According to one embodiment of the present invention, the step of determining whether the target subject has mild cognitive impairment based on the difference between the amplitude ratio of the pre-test harmonics and the amplitude ratio of the post-test harmonics further includes: inputting the first difference value, the second difference value, the third difference value, and the fourth difference value as input data into a pre-established mild cognitive impairment determination model for determination, and the mild cognitive impairment determination model generates an output result to determine whether the target subject has mild cognitive impairment. According to one embodiment of the present invention, the mild cognitive impairment judgment model is established using a multi-layer perceptron architecture. According to one embodiment of the present invention, the judgment model for mild cognitive impairment is established by performing machine learning using a plurality of first difference values, a plurality of second difference values, a plurality of third difference values, and a plurality of fourth difference values ​​generated by a plurality of the pre-measured pulse waves and a plurality of the post-measured pulse waves of a plurality of sampled subjects as a training data set, wherein the sampled subjects include a plurality of known healthy persons and a plurality of patients known to have mild cognitive impairment. According to one embodiment of the present invention, the cognitive intervention activity is an activity that allows the target subject to utilize the prefrontal lobe of the target subject. According to one embodiment of the present invention, the cognitive intervention activity is an activity that allows the target subject to perform logical reasoning of mathematical operations. According to one embodiment of the present invention, the measurement frequencies of the pre-measured pulse wave and the post-measured pulse wave are set to be the same. In summary, by analyzing the harmonic characteristics of the pulse waves converted from the target subjects before and after the cognitive intervention activities, key data for distinguishing between the mild cognitive impairment group and the healthy group was provided. Next, the machine learning model was used to effectively distinguish between the mild cognitive impairment group and the healthy group with an accuracy rate of 83%, which has application potential in the early identification of mild cognitive impairment. The model was subjected to three-fold cross-validation and hold-out analysis, showing that it has high sensitivity and specificity. Therefore, the detection method system of the present invention is not only non-invasive and fast (compared to the assessment scale, the cognitive intervention activities provided by the present invention are simple mathematical logic arithmetic, which takes only a few minutes), but also objective, suitable for large-scale early screening, and can effectively solve the problems of complex operation, high cost and reliance on subjective judgment in the existing technology, and promote early intervention and prevention of cognitive function deterioration. In order to make the above and other objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments of the present invention will be specifically cited below and described in detail with reference to the accompanying drawings. Please refer to FIG. 1 , which is a flow chart of a method 100 for detecting mild cognitive impairment according to one embodiment of the present invention. First, in step S110 , a pre-measured pulse wave of a target subject is obtained. Then, in step S130 , a post-measured pulse wave of the target subject is obtained after a cognitive intervention activity is performed on the target subject. Then, in step S150 , a plurality of pre-measured harmonics are generated by Fourier transforming the pre-measured pulse wave, and a plurality of post-measured harmonics are generated by Fourier transforming the post-measured pulse wave. Then, in step S170 , whether the target subject has mild cognitive impairment is determined based on the changes in the pre-measured harmonics and the post-measured harmonics. Specifically, in the detection method 100 of mild cognitive impairment proposed in the present invention, the harmonics of the subject's pulse wave are mainly used to determine mild cognitive impairment. According to research, the pulse wave of the human body can reflect the condition of the human organs. For example, Chinese medicine diagnoses people through pulse diagnosis. By performing Fourier transform on the pulse wave, the pulse wave signal in the time domain can be converted into multiple harmonic signals in the frequency domain. In one embodiment, the harmonics obtained by Fourier transforming the pulse wave have eleven orders, among which the zero-order harmonic (C0) (the part with zero frequency, also called the DC component) is related to the pericardium meridian, the first-order harmonic (C1) (also called the fundamental frequency) is related to the liver meridian, the second-order harmonic (C2) (the frequency is twice the fundamental frequency) is related to the kidney meridian, the third-order harmonic (C3) (the frequency is three times the fundamental frequency) is related to the spleen meridian, and the fourth-order harmonic (C4) (the frequency is four times the fundamental frequency) is related to the lung meridian. The fifth harmonic (C5) (frequency is five times the fundamental frequency) is related to the stomach meridian, the sixth harmonic (C6) (frequency is six times the fundamental frequency) is related to the gallbladder meridian, the seventh harmonic (C7) (frequency is seven times the fundamental frequency) is related to the bladder meridian, the eighth harmonic (C8) (frequency is eight times the fundamental frequency) is related to the large intestine meridian, the ninth harmonic (C9) (frequency is nine times the fundamental frequency) is related to the triple burner meridian, the tenth harmonic (C10) (frequency is ten times the fundamental frequency) is related to the small intestine meridian, and the eleventh harmonic (C11) (frequency is eleven times the fundamental frequency) is related to the heart meridian. Although these harmonics can reflect the condition of the relevant organs of the human body, they cannot directly reflect the quality of cognitive function, that is, there is no obvious key harmonic that indicates a direct correlation with cognitive function. However, in the detection method 100 of mild cognitive impairment of the present invention, the pulse wave is measured before and after the subject undergoes cognitive intervention activities, and the harmonic components therein are analyzed. It can be found that after the cognitive intervention activities, the first-order harmonics, second-order harmonics, third-order harmonics and fourth-order harmonics in the harmonics converted from the pulse wave have changed. For healthy subjects, after undergoing cognitive intervention activities, the energy of the second-order harmonics and the third-order harmonics decreased significantly; however, for patients with mild cognitive impairment, after undergoing cognitive intervention activities, there was no obvious change in the energy of the second-order harmonics and the third-order harmonics. Therefore, in the mild cognitive impairment detection method 100 of the present invention, the changes in the harmonics of the corresponding pulse wave before and after the cognitive intervention activity of the subject are mainly used to determine whether the subject has mild cognitive impairment. In one embodiment, cognitive intervention activities involve engaging the subject's prefrontal cortex. The brain's prefrontal cortex is primarily responsible for thinking, memory, generating ideas, controlling emotions, making judgments, and applying knowledge. Therefore, when prefrontal cortex activity is induced, cognitive behavior can be considered to be stimulated. In one embodiment, the cognitive intervention activities used in the present invention involve engaging the subject in mathematical operations and logical reasoning. For example, by presenting a series of simple mathematical equations (e.g., continuous subtraction) and asking the subject to determine whether the equations hold (e.g., 100 - 7 = 93? 93 - 7 = 85?), the subject can be encouraged to think and make judgments, thereby stimulating cognitive behavior. In one embodiment, the measurement frequency of the pulse wave measured on the target subject before the cognitive intervention activity (also referred to as the pre-measured pulse wave in this case) and the pulse wave measured on the target subject after the cognitive intervention activity (also referred to as the post-measured pulse wave in this case) are set to be the same, for example, both are 1000 Hz. Please also refer to FIG. 2 , which is a flow chart illustrating step S170 according to the mild cognitive impairment detection method 100 of FIG. 1 . In one embodiment, step S170 further includes steps S171 and S173 . In step S171 , a first difference value between the amplitude ratio of the first-order harmonic (C11) in the pre-measured harmonic and the amplitude ratio of the first-order harmonic (C21) in the post-measured harmonic is obtained; a second difference value between the amplitude ratio of the second-order harmonic (C12) in the pre-measured harmonic and the amplitude ratio of the second-order harmonic (C22) in the post-measured harmonic is obtained; a third difference value between the amplitude ratio of the third-order harmonic (C13) in the pre-measured harmonic and the amplitude ratio of the third-order harmonic (C23) in the post-measured harmonic is obtained; and a fourth difference value between the amplitude ratio of the fourth-order harmonic (C14) in the pre-measured harmonic and the amplitude ratio of the fourth-order harmonic (C24) in the post-measured harmonic is obtained. As previously mentioned, for healthy subjects, changes in the first to fourth order harmonics (particularly the second and third order harmonics) occur before and after cognitive intervention. However, for subjects with mild cognitive impairment, changes in the first to fourth order harmonics are less pronounced. Therefore, the mild cognitive impairment detection method 100 of the present invention primarily determines whether a subject has mild cognitive impairment based on changes in the first to fourth order harmonics before and after cognitive intervention. In this embodiment, the magnitude of the change is determined by the difference in amplitude ratios. The amplitude ratio of the Nth-order harmonic refers to the ratio of the amplitude of the Nth-order harmonic to the amplitude of the first-order harmonic. A larger amplitude ratio indicates that the Nth-order harmonic accounts for a greater proportion of the energy in the pulse wave. Next, in step S173, the first difference value, the second difference value, the third difference value, and the fourth difference value are input as input data into a pre-established mild cognitive impairment judgment model for judgment, and the mild cognitive impairment judgment model generates an output result to determine whether the subject has mild cognitive impairment. Similarly, the judgment model for mild cognitive impairment is established by machine learning using the first difference value, second difference value, third difference value, and fourth difference value generated by the above steps of the pre-measured pulse waves and post-measured pulse waves obtained from multiple sampled subjects before and after undergoing cognitive intervention activities (for example, the logical judgment of the series of subtraction operations mentioned above) as a training data set. These sampled subjects include known healthy people of different age groups and multiple patients of different age groups who are known to have mild cognitive impairment. For example, the sampled subjects may be 100 people. In one example, 200 pre-measured pulse waves and 200 post-measured pulse waves may be measured for each sampled subject, resulting in a total of 20,000 pre-measured pulse waves and 20,000 post-measured pulse waves. Next, for each sampled subject, the 200 pre-measured pulse waves and 200 post-measured pulse waves are Fourier transformed to obtain 200 pre-measured harmonics and 200 post-measured harmonics. Next, the amplitude ratios of the first-order harmonics in all pre-test harmonics can be averaged to obtain a pre-test first-order harmonic average amplitude ratio. Then, the amplitude ratio of the first-order harmonic in each post-test harmonic is subtracted from the pre-test first-order harmonic average amplitude ratio to obtain the first-order harmonic amplitude ratio difference value corresponding to the post-test harmonic (in this example, 200 first-order harmonic amplitude ratio difference values ​​can be obtained for each sampled subject) as the first difference value. The amplitude ratio difference values ​​of the second-order harmonic, third-order harmonic, and fourth-order harmonic are also obtained in the same way. These data will all be used as the training data set for training the judgment model of mild cognitive impairment. In this embodiment, the mild cognitive impairment judgment model is established using a multilayer perceptron (MLP) architecture. The input data required by the input layer are the multiple first difference values, second difference values, third difference values, and fourth difference values ​​obtained above, and whether the subject has mild cognitive impairment is used as the prediction target. The ReLU (Rectified Linear Unit) activation function is applied between the hidden layers to ensure nonlinear transformation capabilities. The final classification layer uses SoftMax for classification. In one embodiment, to address the data imbalance between the mild cognitive impairment group and the healthy control group, the Synthetic Minority Oversampling Technique (SMOTE) technique was used to perform synthetic data augmentation to ensure a balanced training dataset. Furthermore, three-fold cross-validation was used to prevent overfitting, and model hyperparameters (such as the learning rate, batch size, and regularization coefficient) were adjusted through random search. A three-fold cross-validation experiment on the mild cognitive impairment diagnosis model revealed an average accuracy of 83%, a sensitivity of 85.14%, a specificity of 80.85%, and an AUC (Area Under Curve) value of 0.82, indicating that the model has high accuracy in distinguishing mild cognitive impairment from healthy controls. The hold-out validation results, performed on a dataset not included in the cross-validation, showed that the model had an accuracy of 81.84%, a sensitivity of 76.05%, a specificity of 87.64%, and an AUC value of 0.82, demonstrating that the model also performs well in practical applications. Based on this, by analyzing the harmonic characteristics of the pulse waves converted from the target subjects before and after the cognitive intervention activities, key data for distinguishing between the mild cognitive impairment group and the healthy group was provided. Then, the machine learning model was used to effectively distinguish between the mild cognitive impairment group and the healthy group with an accuracy rate of 83%, which has application potential in the early identification of mild cognitive impairment. The model was subjected to three-fold cross-validation and hold-out analysis, showing that it has high sensitivity and specificity. Therefore, the detection method system of the present invention is not only non-invasive and fast (compared to the assessment scale, the cognitive intervention activity provided by the present invention is a simple mathematical logic arithmetic that takes only a few minutes), but also objective, suitable for large-scale early screening, and can effectively solve the problems of complex operation, high cost and reliance on subjective judgment in the existing technology, promoting early intervention and prevention of cognitive function deterioration. Although the present invention has been disclosed with reference to preferred embodiments, this is not intended to limit the present invention. Anyone skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be determined by the scope of the appended patent applications. 100: Detection method for mild cognitive impairment S110, S130, S150, S170, S171, S173: Steps FIG. 1 is a flow chart of a method for detecting mild cognitive impairment according to an embodiment of the present invention; FIG. 2 is a flow chart of a step of the method for detecting mild cognitive impairment in FIG. 1 . 100: Detection methods for mild cognitive impairment S110, S130, S150, S170: Steps

Claims

1. A method for detecting mild cognitive impairment, comprising: Acquiring a pre-measured pulse wave of a target subject; performing a cognitive intervention activity on the target subject and then acquiring a post-measured pulse wave of the target subject; generating a plurality of pre-measured harmonics by Fourier transforming the pre-measured pulse wave; generating a plurality of post-measured harmonics by Fourier transforming the post-measured pulse wave; and determining whether the target subject has mild cognitive impairment based on changes in the pre-measured harmonics and the post-measured harmonics.

2. The method for detecting mild cognitive impairment according to claim 1, wherein the step of determining whether the target subject has mild cognitive impairment based on changes in the pre-test harmonics and the post-test harmonics comprises: Whether the target subject has mild cognitive impairment is determined according to the difference between the amplitude ratio of the pre-test harmonics and the amplitude ratio of the post-test harmonics.

3. The method for detecting cognitive impairment according to claim 2, wherein the step of determining whether the target subject has mild cognitive impairment based on the difference between the amplitude ratio of the pre-test harmonics and the amplitude ratio of the post-test harmonics comprises: Whether the target subject has mild cognitive impairment is determined based on a first difference value between the amplitude ratio of a first-order harmonic among the pre-test harmonics and the amplitude ratio of a first-order harmonic among the post-test harmonics, a second difference value between the amplitude ratio of a second-order harmonic among the pre-test harmonics and the amplitude ratio of a second-order harmonic among the post-test harmonics, a third difference value between the amplitude ratio of a third-order harmonic among the pre-test harmonics and the amplitude ratio of a third-order harmonic among the post-test harmonics, and a fourth difference value between the amplitude ratio of a fourth-order harmonic among the pre-test harmonics and the amplitude ratio of a fourth-order harmonic among the post-test harmonics.

4. The method for detecting cognitive impairment according to claim 3, wherein the step of determining whether the target subject has mild cognitive impairment based on the difference between the amplitude ratio of the pre-test harmonics and the amplitude ratio of the post-test harmonics further comprises: The first difference value, the second difference value, the third difference value, and the fourth difference value are input as input data into a pre-established mild cognitive impairment judgment model for judgment, and the mild cognitive impairment judgment model generates an output result to determine whether the target subject has mild cognitive impairment.

5. The method for detecting cognitive impairment according to claim 4, wherein the judgment model for mild cognitive impairment is established using a multi-layer perceptron architecture.

6. A method for detecting cognitive impairment as described in claim 5, wherein the judgment model for mild cognitive impairment is established by performing machine learning using a plurality of first difference values, a plurality of second difference values, a plurality of third difference values, and a plurality of fourth difference values ​​generated by obtaining a plurality of the pre-measured pulse waves and a plurality of the post-measured pulse waves of a plurality of sampled subjects as a training data set, wherein the sampled subjects include a plurality of known healthy persons and a plurality of patients known to have mild cognitive impairment.

7. The method for detecting cognitive impairment according to claim 1, wherein the cognitive intervention activity is an activity that allows the target subject to utilize the prefrontal lobe of the target subject.

8. The method for detecting cognitive impairment as described in claim 7, wherein the cognitive intervention activity is an activity in which the target subject performs logical reasoning of mathematical operations.

9. The method for detecting cognitive impairment according to claim 1, wherein the measurement frequencies of the pre-measured pulse wave and the post-measured pulse wave are set to be the same.

Images