TWI597683B - a migraine attack detection system included building and prediction method - Google Patents

Description

Establishment and prediction of migraine attack prediction system

The invention relates to a method for predicting migraine, in particular to a method for establishing and predicting a migraine attack prediction system in a resting state.

Migraine is an unexplained headache. The characteristics of headache are obvious. There are no studies to confirm the exact cause of migraine. From the perspective of migraine cycle, it can be divided into interictal period (between two headaches), pre-seizure period (pre-headache period), seizure period (headache period) and late stage (late headache), and the physiological state of patients in different periods is also It will be different, for example, the amplitude of brain waves is different, the degree of habituation is different, and the degree of activation of the cerebral cortex is different.

When the patient is in the early stage of headache, there are usually some reactions, such as muscle stiffness, emotional changes, sensitivity to sound or smell, etc., but sometimes patients may not be aware of their own changes, so when the patient takes the medicine, it is often in the attack period, headache It is unbearable to take painkillers, and everyone reacts differently to drugs. Some people will suffer pain after a period of time to slow down slowly, which affects the quality of life.

It is currently predicted that migraine is usually stimulated by light or sound, and the brain wave changes after stimulation are used to judge whether it is in the early stage of headache. However, the patient may feel dizzy and uncomfortable, which may prompt the patient to have a migraine attack. . Therefore, the present invention proposes a method for establishing and predicting a migraine attack prediction system in a resting state, and the specific architecture and its implementation manner will be described in detail below:

The main purpose of the present invention is to provide a premature system for predicting migraine attacks. The method is to obtain the brain wave signal of the resting state of the migraine patient, and then analyze the brain wave signals of different brain regions (every two channels) to determine the different brain regions (every two channels). The degree of correlation between the two, resulting in multi-channel consistency eigenvalues. Then through the machine learning algorithm, the relevant consistent feature values are input or substituted into the machine learning algorithm. In order to learn, classify, and establish a predictable migraine period judgment model, after the establishment of the migraine development prediction system, when the new migraine patient receives a brain wave signal at rest, they can use the same Sexual eigenvalues and migraine period judgment models predict which headache period the migraine patient is at the moment of measuring brain waves.

Another object of the present invention is to provide a method for establishing and predicting a migraine attack prediction system, which only needs to extract brain waves of the patient's resting state (open eyes and closed eyes) without giving any stimulation or reaction to the patient. Measuring brain waves at rest can avoid evoked migraine attacks and comfort for migraine patients.

A further object of the present invention is to provide a method for establishing and predicting a migraine attack prediction system, which classifies brainwave signals into a headache interval and a pre-headache period to assist the patient in determining the timing of taking the medication (timely in the early stage of the headache) and slowing down The pain of the patient.

To achieve the above object, the present invention provides a method for establishing and predicting a migraine attack prediction system, comprising the steps of: capturing a brain wave signal of a patient at rest; using a migraine attack prediction system for brain wave signals Perform brain area consistency analysis and obtain multi-channel consistency eigenvalues; migraine attack prediction system performs machine learning based on consistent eigenvalues obtained from analysis; in machine learning, when learning the consistency of patients belonging to different periods After the eigenvalues, the different migraine periods are correctly classified by the consistency eigenvalues to establish a migraine period judgment model; and the brainwave signals of a migraine patient's resting state are taken, and the migraine attack prediction system is utilized. The consistency eigenvalue and the migraine period judgment model determine from the brain wave signal which period of the migraine period the migraine patient currently belongs to.

The purpose, technical content, features and effects achieved by the present invention will be more readily understood by the detailed description of the embodiments.

Figure 1 is a flow chart of the method for establishing and predicting a migraine attack prediction system of the present invention.

Fig. 2 is a flow chart showing the steps of brain brain consistency analysis in the establishment and prediction method of the migraine attack prediction system of the present invention.

Figures 3A and 3B are schematic diagrams of brainwave signal channels for the inter-headache period and the early headache period, respectively.

The invention provides a method for establishing and predicting a migraine attack prediction system, and provides a personalized classification of interictal and preictal in a migraine patient, assisting drug treatment, and slowing the pain and treatment of the patient. . Using the coherence of migraine patients during different migraine periods, find indicators of consistency changes to establish prediction and expert systems.

Fig. 1 is a flow chart showing the method for establishing and predicting a migraine attack prediction system according to the present invention, wherein steps S10 to S16 are methods for establishing a migraine attack prediction system, and steps S18 to S20 are for predicting the migraine attack. The system predicts the current migraine period of a migraine patient. First, step S10 captures a multi-channel brain wave signal of a patient in a resting state (eye or closed eyes), which allows the patient to wear an acceptable brainwave. The instrument can extract brain wave signals ranging from 17 channels to 64 channels. In the resting state, the patient only needs to do the action of opening or closing the eyes; then, step S12 uses a migraine attack prediction system. The brain wave signal performs brain region consistency analysis to obtain a complex consistency feature value, and the migraine attack prediction system is a program installed in at least one host to analyze the collected brain wave signals, and the brain region consistency The analysis is to see the consistency of every two channels of the brain wave. If the consistency is high, it means that the energy in the two channels has a high similarity, and vice versa, the similarity is low; in step S14, the partiality is Pain attack prediction system for machine learning based on the consistency of the resulting eigenvalue analysis After learning the consistency characteristic values of the patients belonging to different periods, in step S16, the migraine attack prediction system can classify the consistency feature values into different migraine periods to establish a migraine period judgment model. For example, the brainwave signal of the patient is classified as a headache interval or a pre-headache period; after that, by taking the brainwave signal of any migraine patient, the consistency characteristic value and the migraine period in the migraine attack prediction system can be utilized. The model is judged from the brain wave signal to determine which of the migraine patients is currently in the migraine period, as described in steps S18 to S20.

For example, in the case of the eye opening and the closing eye movement in step S10, one eye can be closed for one minute after one minute of eye opening, and several rounds are alternately performed for 5 to 10 minutes to collect brain wave signals during this time.

Referring to FIG. 2, in step S12 of FIG. 1, the brain region consistency step further includes two steps. After step S10 captures the brain wave signal of the patient, step S122 first filters out the low frequency in the brain wave signal. And after the high frequency part, extracting part of the brain wave information; then performing brain area consistency analysis on the brain wave information in step S124, analyzing the feature consistency of each of the two channels in the brain wave information, and obtaining consistency characteristics In the present invention, the comparison method uses a dimensionality reduction algorithm to extract the feature information of the brain wave information, and obtains the consistency feature value by weighted scoring.

For example, if the patient's brainwave signal is 17 channels, the consistency between each of the 17 channels will be calculated. = 153 consistent values, which are consistent eigenvalues, and all consistent eigenvalues (eg 153 contiguous values) use the dimensionality reduction algorithm to preserve important characteristics, and the final consistency is obtained by weighted scoring Eigenvalues.

In the present invention, the migraine attack prediction system uses a supervised machine learning algorithm recognizer or an unsupervised machine learning algorithm recognizer to learn the consistency feature value, and obtains a migraine period judgment model. However, it is not limited to supervising machine learning algorithm recognizers or unsupervised machine learning algorithm recognizers, and other algorithms or recognizers having learning modes can also be applied to the present invention.

The migraine period judgment model is a feature of the consistency of the migraine attack prediction system The judgment mechanism learned by the value, and then the brainwave information of the new migraine patient, can be correctly judged whether the patient is in the headache period or the early headache period. Please refer to Figure 3A and Figure 3B. Figure 3A is a schematic diagram of the brainwave signal channel during the headache interval. Figure 3B is a schematic diagram of the brainwave signal channel in the early stage of headache. It can be seen from the figure that the brainwave signal at different times is Differentiated, the difference in the consistency of the eigenvalues will be different, which affects the classification of the brainwave signals of the migraine period judgment model. The ultimate goal of the migraine period judgment model in the present invention is to judge that the patient is currently in a headache. Period or early headache, to assist patients in selecting appropriate drugs at the right time.

Suppose there is a migraine patient who has not yet had an episode, but is in a headache or pre-headache. Usually, the patient wears a brainwave signal twice a day to detect whether he has entered the early stage of headache. If he is in the early stage of headache, he should take the medicine immediately to suppress the occurrence of migraine. If he is still in the headache period, that is, two migraine headaches. During the period between the two, it is only necessary to keep the brainwave signal check every day.

In summary, the method for establishing and predicting a migraine attack prediction system provided by the present invention first acquires the brain wave signal of a migraine patient, and uses a consistency between each two channels of the brain wave to give a consistency characteristic. Value, according to the consistency characteristic value, the patient's brain wave signal is classified into pre-headache or headache period. The migraine attack prediction system can learn a set of judgment mechanism, and the migraine attack will be detected whenever the patient's brain wave is detected. The predictive system can correctly determine that the patient is at the time of headache or pre-headache, and because the migraine will be predicted in advance before the migraine attack, the patient can take appropriate drugs to suppress migraine. Occurs, slows down the suffering of the patient. The invention is suitable for home monitoring and care for long-term migraine patients.

The above is only the preferred embodiment of the present invention and is not intended to limit the scope of the present invention. Therefore, any changes or modifications of the features and spirits of the present invention should be included in the scope of the present invention.

no

Claims (5)

A method for establishing and predicting a migraine attack prediction system, which collects brainwave signals and uses at least one host to perform migraine prediction, the method comprising the steps of: establishing a migraine attack prediction system, comprising the steps of: taking a patient Multi-channel complex brainwave signals at rest; performing brain region consistency analysis, using the migraine attack prediction system to analyze the consistency of the brainwave signals between different brain regions, and filtering out the brainwave signals After the low frequency and high frequency part, extract some of the brain wave information, analyze the characteristic consistency of each two channels in the brain wave information, and then use a dimensionality reduction algorithm to perform the brain wave information. Feature extraction, weighted scoring to obtain multi-channel complex consistency feature values; and the migraine attack prediction system substitutes the consistency feature values into a machine learning algorithm for machine learning, when learning patients belonging to different periods After the consistency feature values, different migraine periods are classified by the consistency feature values to establish a migraine period judgment model; And predicting a migraine period in a migraine patient, including: extracting the multi-channel brainwave signals of the migraine patient at rest, using the consistency feature values in the migraine attack prediction system And the migraine period judgment model, from which the brain wave signals are used to determine which of the migraine periods the migraine patient currently belongs to. The method for establishing and predicting a migraine attack prediction system according to claim 1, wherein the brain wave signals are the brain wave signals when the patient performs an alternating action of opening and closing eyes in a resting state, and recording the Wait for the brain wave signal for 5~10 minutes. The method for establishing and predicting a migraine attack prediction system according to claim 1, wherein the bias The headache period includes an interictal and a preictal. The method for establishing and predicting a migraine attack prediction system according to claim 1, wherein the migraine attack prediction system uses a supervised machine learning algorithm identifier or an unsupervised machine learning algorithm identifier to learn the agreement. The sexual characteristic value is obtained by the migraine period judgment model. The method for establishing and predicting a migraine attack prediction system according to claim 1, wherein the migraine attack prediction system is installed in the host.