CN110265143B - Intelligent auxiliary diagnosis system based on EEG - Google Patents

Abstract

The invention relates to an intelligent auxiliary diagnosis system based on electroencephalogram, which comprises electroencephalogram preprocessing, waveform model construction, waveform identification, appearance model construction, appearance mode identification, distribution model construction, distribution mode identification, state model construction, and patient status It consists of eleven subsystems for identifying, diagnosing model building and diagnosing. The five subsystems of waveform model construction, appearance model construction, distribution model construction, state model construction and diagnosis model construction use different methods to build models respectively; the EEG preprocessing subsystem filters, artifact removal, annotation and reference of brain power signals After preprocessing such as transformation, the waveform identification subsystem, the appearance mode identification subsystem, the distribution mode identification subsystem, the patient state identification subsystem and the diagnosis subsystem successively use the corresponding models to process the EEG data to form diagnostic opinions, which are provided to the patients. doctor as a reference. The invention can improve the efficiency and quality of EEG medical diagnosis.

Description

Intelligent auxiliary diagnosis system based on EEG

technical field

The invention relates to an intelligent auxiliary diagnosis system based on electroencephalogram.

Background technique

At present, the diagnosis of brain diseases relies on medical personnel to manually check the EEG, and the diagnosis results are limited by the experience level and knowledge accumulation of medical personnel. Different doctors give different diagnoses for the same EEG, which affects the treatment of patients. In addition, in order to capture effective information, EEG-based medical diagnosis needs to record EEG signals for a long time. In the face of long-term EEG signals, doctors need to read screen by screen to make a diagnosis, which is very time-consuming and difficult to ensure the quality of reading.

SUMMARY OF THE INVENTION

The purpose of the present invention is to solve the problem of the existing intelligent auxiliary diagnosis based on electroencephalogram, and propose an intelligent auxiliary diagnosis system based on electroencephalogram. The diagnostic system is applied to other patients to reduce the adverse influence of doctors on the grasp of the disease due to personal limitations, and to improve the efficiency of disease diagnosis.

In order to achieve the above purpose, the technical scheme of the present invention is: an EEG-based intelligent auxiliary diagnosis system, including an EEG preprocessing subsystem, a waveform model building subsystem, a waveform identification subsystem, an appearance model building subsystem, and an appearance model building subsystem. Mode identification subsystem, distributed model construction subsystem, distributed mode identification subsystem, state model construction subsystem, patient state identification subsystem, diagnosis model construction subsystem and eleven subsystems of diagnosis subsystem;

The EEG preprocessing subsystem receives the brain power signal as input, and outputs the preprocessed EEG data after filtering, artifact removal, labeling and reference transformation; the EEG preprocessing subsystem removes the noise in the EEG signal by filtering, The signal segment affected by the artifact is removed by thresholding, and the artifact mixed on each lead is removed by independent component analysis; the labeling function of the EEG preprocessing subsystem allows users to additionally label unlabeled events in the source data; EEG The labeling function of the preprocessing subsystem also allows users to label the EEG waveform; the reference transformation function of the EEG preprocessing subsystem can switch the source signal, that is, the source signal can be switched to the EEG signal of other reference modes through calculation;

The waveform model building subsystem extracts samples of each EEG waveform from the preprocessed EEG data marked with EEG waveforms, and uses the extracted samples to construct a training set for each EEG waveform. Methods A waveform model is constructed for each EEG waveform, and the waveform model is output to the waveform identification subsystem; the waveform model construction subsystem has the functions of continuously accumulating and optimizing the EEG waveform samples, expanding the training set, and reconstructing and improving the optimized waveform model;

The waveform identification subsystem receives the preprocessed EEG data to be diagnosed output by the EEG preprocessing subsystem, and uses the waveform model provided by the waveform model construction subsystem to identify all the EEG waveforms from the EEG data. Label the identified EEG waveforms, the spatial and temporal positions of the EEG waveforms in the data, and output the labeled EEG data;

Appearance model building subsystem, distribution model building subsystem, state model building subsystem, and diagnosis model building subsystem receive the doctor's empirical rules by interacting with the user, and express them with a normalized data structure, and use logical reasoning to exclude the rules between them. It uses a set of optimized rules to build models and output them; the appearance model building subsystem receives the empirical rules for doctors to judge the appearance of EEG waveforms in time, and outputs the appearance models that identify the appearance patterns; The distribution model building subsystem receives the empirical rules used by doctors to determine the spatial distribution of EEG waveforms, and outputs a distribution model that identifies the distribution patterns; The EEG waveform distribution method is used to determine the empirical rules of the patient's state, and the state model used to identify the patient's state is output; the diagnosis model building subsystem receives the doctor's EEG waveform, the appearance mode of the EEG waveform and the EEG waveform according to the different states of the patient. Empirical rules for diagnosis in a distributed manner, and output diagnostic models for diagnosing diseases;

The appearance mode identification subsystem receives the EEG waveform data output from the waveform identification subsystem, uses the appearance model provided by the appearance model building subsystem to identify the EEG waveform appearance mode, marks the identified appearance mode in the EEG data, and uses the appearance model provided by the appearance model construction subsystem to identify the EEG waveform appearance mode. Annotated EEG data output;

The distribution pattern identification subsystem receives the EEG waveform data output from the waveform identification subsystem, uses the distribution model provided by the distribution model building subsystem to identify the EEG waveform distribution pattern, marks the identified distribution pattern in the EEG data, and uses Annotated EEG data output;

The patient state identification subsystem receives the EEG data marked with the EEG waveform, appearance mode and distribution mode, and uses the state model provided by the state model to build the subsystem to process, identify the state of the patient at different time periods, and add the EEG data to the EEG data. After the status is marked, it will be output;

The diagnosis subsystem receives the EEG data marked with the EEG waveform, appearance, distribution and patient status, uses the diagnostic model provided by the diagnostic model to build the subsystem to process the EEG data, and gives a diagnosis opinion for the patient; the diagnosis opinion is determined by qualitative Opinions, descriptions about EEG, and EEG data indexes corresponding to EEG descriptions; each qualitative opinion is supported by a set of EEG descriptions, and each EEG description corresponds to a set of EEG data ; The diagnostic subsystem supports users in the retrieval of EEG data from qualitative opinions to EEG descriptions.

Compared with the prior art, the present invention has the following beneficial effects: an EEG-based intelligent auxiliary diagnosis system provided by the present invention uses the idea of intelligent technology to construct a diagnostic system for various brain diseases according to the collected data set. The system is applied to other patients to reduce the adverse influence of doctors on the disease grasp due to personal limitations, and to improve the efficiency of disease diagnosis.

Description of drawings

FIG. 1 is a block diagram of an EEG-based intelligent auxiliary diagnosis system of the present invention.

Detailed ways

The technical solutions of the present invention will be described in detail below with reference to the accompanying drawings.

It should be noted that the following detailed description is exemplary and intended to provide further explanation of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

It should be noted that the terminology used herein is for the purpose of describing specific embodiments only, and is not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural as well, furthermore, it is to be understood that when the terms "comprising" and/or "including" are used in this specification, it indicates that There are features, steps, operations, devices, components and/or combinations thereof.

As shown in FIG. 1 , this embodiment provides an EEG-based intelligent auxiliary diagnosis system, including an EEG preprocessing subsystem, a waveform model construction subsystem, a waveform identification subsystem, an appearance model construction subsystem, and an appearance method. Identification subsystem, distributed model construction subsystem, distributed mode identification subsystem, state model construction subsystem, patient state identification subsystem, diagnosis model construction subsystem and eleven subsystems of diagnosis;

The EEG preprocessing subsystem receives the brain power signal as input, and outputs the preprocessed EEG data after filtering, artifact removal, labeling and reference transformation; the EEG preprocessing subsystem removes the noise in the EEG signal by filtering, Threshold limit removes signal segments affected by artifacts, and independent component analysis removes artifacts mixed in each lead; the labeling function of the EEG preprocessing subsystem allows users to additionally label unlabeled events in the source data; EEG preprocessing The labeling function of the subsystem also allows users to label the EEG waveform; the reference transformation function of the EEG preprocessing subsystem can switch the source signal to the EEG signal of other reference modes through calculation; Annotated events include known epileptic seizures, eye-opening and closing tests, eye-state sensitivity tests, hyperventilation tests, flash stimulation, sleep-inducing, drug-inducing, and events that doctors believe should be labeled during EEG recordings; EEG waveforms include Sine wave, alpha rhythm, beta rhythm, gamma rhythm, delta rhythm, theta rhythm, bow wave, notch wave, biphasic wave, triphasic wave, polyphasic wave, spike wave, sharp wave, spike slow wave, polyspike wave, polyspike and slow wave, complex wave, K complex, polymorphic wave; other reference modes include ipsilateral ear reference, binaural reference, average reference and bipolar lead;

The waveform model building subsystem extracts samples of each EEG waveform from the preprocessed EEG data marked with EEG waveforms, and uses the extracted samples to construct a training set for each EEG waveform. Methods A waveform model is constructed for each EEG waveform, and the waveform model is output to the waveform identification subsystem; the waveform model construction subsystem has the functions of continuously accumulating and optimizing the EEG waveform samples, expanding the training set, and reconstructing and improving the optimized waveform model;

The waveform identification subsystem receives the preprocessed EEG data to be diagnosed output by the EEG preprocessing subsystem, and uses the waveform model provided by the waveform model construction subsystem to identify all the EEG waveforms from the EEG data. Label the identified EEG waveforms, the spatial and temporal positions of the EEG waveforms in the data, and output the labeled EEG data;

Appearance model building subsystem, distribution model building subsystem, state model building subsystem, and diagnosis model building subsystem receive the doctor's empirical rules through interaction with users, represent them with a normalized data structure, and use logical reasoning to exclude rules It uses a set of optimized rules to build models and output them; the appearance model building subsystem receives the empirical rules for doctors to judge the appearance of EEG waveforms in time, and outputs the appearance models that identify the appearance patterns; The distribution model building subsystem receives the empirical rules used by doctors to judge the spatial distribution of EEG waveforms, and outputs a distribution model that identifies the distribution patterns; the state model building subsystem receives the doctor’s data based on the EEG waveform, the appearance mode of the EEG waveform, and the distribution model. The EEG waveform distribution method is used to determine the empirical rules of the patient's state, and the state model used to identify the patient's state is output; the diagnostic model building subsystem receives the doctor's EEG waveform, the appearance mode of the EEG waveform and the EEG waveform according to the different states of the patient. Empirical rules for diagnosis in a distributed manner, and output diagnostic models for diagnosing diseases;

The appearance mode identification subsystem receives the EEG waveform data output from the waveform identification subsystem, uses the appearance model provided by the appearance model building subsystem to identify the EEG waveform appearance mode, marks the identified appearance mode in the EEG data, and uses the appearance model provided by the appearance model construction subsystem to identify the EEG waveform appearance mode. The marked EEG data output; in this embodiment, the EEG waveform appearance modes include activity, rhythm, outbreak, burst, periodicity, sporadic, transient, and synchronicity;

The distribution pattern identification subsystem receives the EEG waveform data output from the waveform identification subsystem, uses the distribution model provided by the distribution model building subsystem to identify the EEG waveform distribution pattern, marks the identified distribution pattern in the EEG data, and uses The marked EEG data is output; in this embodiment, the EEG waveform distribution modes include extensive, diffuse, one-sided, localized, multifocal, migratory, and symmetrical;

The patient state identification subsystem receives the EEG data marked with the EEG waveform, appearance mode and distribution mode, and uses the state model provided by the state model to build the subsystem to process, identify the state of the patient at different time periods, and add the EEG data to the EEG data. The state is marked and then output; in this embodiment, the patient state includes normal wakefulness, normal sleep, sleep cycle, abnormal background activity, and abnormal formation;

The diagnosis subsystem receives the EEG data marked with the EEG waveform, appearance, distribution and patient status, uses the diagnostic model provided by the diagnostic model to build the subsystem to process the EEG data, and gives a diagnosis opinion for the patient; the diagnosis opinion is determined by qualitative Opinions, descriptions about EEG, and EEG data indexes corresponding to EEG descriptions; each qualitative opinion is supported by a set of EEG descriptions, and each EEG description corresponds to a set of EEG data ; The diagnosis subsystem supports users to search from qualitative opinions to EEG descriptions and then to EEG data; in this embodiment, qualitative opinions are divided into normal, normal range, boundary line, mild abnormality, moderate abnormality and severe abnormality grading; descriptions of EEG include background activity, evoked tests, sleep, abnormal waves, clinical seizures, drug effects, and interference waves;

The above are only preferred embodiments of the present invention, and are not intended to limit the present invention in other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications to equivalent changes. Example. However, any simple modifications, equivalent changes and modifications made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solutions of the present invention still belong to the protection scope of the technical solutions of the present invention.

Claims (1)

1. An electroencephalogram-based intelligent auxiliary diagnosis system is characterized by comprising eleven subsystems, namely an electroencephalogram preprocessing subsystem, a waveform model building subsystem, a waveform identification subsystem, an appearance model building subsystem, an appearance mode identification subsystem, a distribution model building subsystem, a distribution mode identification subsystem, a state model building subsystem, a patient state identification subsystem, a diagnosis model building subsystem and a diagnosis subsystem;

the electroencephalogram preprocessing subsystem receives an electroencephalogram source signal as input, and outputs preprocessed electroencephalogram data after filtering, artifact removing, labeling and reference transformation; the electroencephalogram preprocessing subsystem removes noise in an electroencephalogram source signal through filtering, removes signal segments influenced by artifacts through threshold limitation, and removes the artifacts mixed on each lead through independent component analysis; the label of the brain electrical preprocessing subsystem is used for allowing a user to supplement and label events which are not labeled in the brain electrical source signal; the marking of the brain electrical preprocessing subsystem is also used for allowing a user to mark brain electrical waveforms; the reference transformation of the brain electrical pre-processing subsystem is used for switching the brain electrical source signal into brain electrical signals of other reference modes; the unmarked events comprise known epileptic seizure, open-close eye tests, eye state sensitivity tests, hyperventilation tests, flash stimulation, sleep induction and drug induction in the electroencephalogram recording process; the brain wave comprises sine-like wave, alpha rhythm, beta rhythm, gamma rhythm, delta rhythm, theta rhythm, bow wave, incisional wave, diphasic wave, triphase wave, multiposition phasic wave, spike wave, multiple spike wave, complex wave, K complex wave and polymorphism wave; other reference patterns include ipsilateral ear reference, binaural reference, mean reference, and bipolar leads;

the waveform model construction subsystem extracts a sample of each brain waveform from the preprocessed brain electrical data labeled with the brain waveform, constructs a training set for each brain waveform by using the extracted sample, constructs a waveform model for each brain waveform by a method of training on the training set, and outputs the waveform model to the waveform recognition subsystem; the waveform model construction subsystem is used for continuously accumulating brain waveform samples, expanding a training set and reconstructing an improved and optimized waveform model;

the waveform identification subsystem receives the preprocessed electroencephalogram data to be diagnosed, which are output by the electroencephalogram preprocessing subsystem, identifies all electroencephalogram waveforms from the electroencephalogram data by adopting a waveform model provided by the waveform model construction subsystem, marks the identified electroencephalogram waveforms, the spatial positions and the temporal positions of the electroencephalogram waveforms in the electroencephalogram data, and outputs the marked electroencephalogram data;

the appearance model construction subsystem, the distribution model construction subsystem, the state model construction subsystem and the diagnosis model construction subsystem respectively receive experience rules of doctors in a mode of interacting with users, express the experience rules by using a normalized data structure, eliminate logic contradictions among the rules by using logic reasoning, construct corresponding models by using a group of optimized rules and output the models; the appearance model construction subsystem receives an empirical rule that a doctor judges the appearance mode of the brain waveform in time and outputs an appearance model for identifying the appearance mode; the distribution model construction subsystem receives an empirical rule that a doctor judges the spatial distribution mode of the brain waveform and outputs a distribution model for identifying the distribution mode; the state model building subsystem receives an experience rule that a doctor judges the state of a patient according to a brain waveform, a brain waveform occurrence mode and a brain waveform distribution mode, and outputs a state model for identifying the state of the patient; the diagnosis model construction subsystem receives the experience rules of doctors for diagnosing according to the brain waveforms, the brain waveform occurrence modes and the brain waveform distribution modes of patients in different states, and outputs a diagnosis model for diagnosing diseases;

the appearance mode identification subsystem receives the brain waveforms output by the waveform identification subsystem, identifies the appearance modes of the brain waveforms by adopting the appearance models provided by the appearance model construction subsystem, marks the identified appearance modes in the brain electrical data, and outputs the marked brain electrical data; wherein, the appearance modes of the brain wave comprise activity, rhythm, outbreak, paroxysmal, periodicity, sporadic, transient and synchronicity;

the distribution mode identification subsystem receives the brain waveforms output by the waveform identification subsystem, identifies the brain waveform distribution mode by adopting a distribution model provided by the distribution model construction subsystem, marks the identified distribution mode in the brain electrical data and outputs the marked brain electrical data; wherein, the brain waveform distribution mode comprises universality, diffusivity, sidedness, locality, multifocal, wandering and symmetry;

the patient state identification subsystem receives the electroencephalogram data marked with the electroencephalogram waveform, the appearance mode and the distribution mode, processes the electroencephalogram data by adopting the state model provided by the state model construction subsystem, identifies the states of the patient at different time intervals, and outputs the electroencephalogram data after adding state marks; wherein the patient state comprises normal waking period, normal sleep period, sleep cycle, background activity abnormality, and paroxysmal abnormality;

the diagnosis subsystem receives the electroencephalogram data marked with the electroencephalogram waveform, the appearance mode, the distribution mode and the patient state, processes the electroencephalogram data by using the diagnosis model provided by the diagnosis model construction subsystem, and gives a diagnosis suggestion for the patient; the diagnosis opinions consist of qualitative opinions, description about electroencephalogram and electroencephalogram data indexes corresponding to the electroencephalogram description; each qualitative opinion is supported by a set of descriptions about electroencephalograms, each electroencephalogram description corresponding to a set of electroencephalogram data; the diagnosis subsystem supports the retrieval of the user from qualitative opinions, electroencephalogram description and electroencephalogram data; wherein, the qualitative opinions are divided into the grades of normal, normal range, borderline, mild abnormity, moderate abnormity and severe abnormity; the description about electroencephalogram includes background activity, evoked potential, sleep, abnormal waves, clinical attacks, drug action, and interfering waves.

Images