Diagnostic System For Depression Using Multi-paradigm Electroencephalography With Machine Learning

Abstract

The present invention relates to a depression diagnosis system using multiparadigm brainwaves and machine learning. According to one embodiment of the present invention, a depression diagnosis system using multiparadigm brainwaves and machine learning may include: a brainwave measurement unit that measures a plurality of paradigm brainwaves from a subject; a brainwave extraction unit that extracts and preprocesses necessary multiparadigm brainwaves from the measured plurality of paradigm brainwaves to minimize noise; a model generation unit that creates a diagnosis model by performing machine learning using the extracted multiparadigm brainwaves; and a diagnosis unit that diagnoses the presence of depression using the generated diagnosis model. As such, according to the present invention, by generating a diagnostic model of a subject using multiparadigm brainwaves and diagnosing the presence of depression, the limitations of conventional diagnostic classification using a single paradigm can be overcome. Furthermore, by generating and applying a model through machine learning using multiple key brainwave biomarker data, the classification performance of depression with psychopathological diversity can be enhanced.

Description

Technology Field

[0001] The present invention relates to a depression diagnosis system using multi-paradigm brainwaves and machine learning, and more specifically, to a depression diagnosis system using multi-paradigm brainwaves and machine learning that can distinguish between a depressive state and a normal state and diagnose whether a subject has depression. Background Technology

[0002] Depression refers to a disorder characterized by a lack of motivation and feelings of sadness as primary symptoms, causing various cognitive and psychosomatic symptoms that lead to a decline in daily functioning. Meanwhile, electroencephalography (EEG) refers to the time-series recording of minute electrical activity within groups of brain cells in the cerebral cortex, induced by attaching electrodes to the scalp. EEGs vary spatiotemporally depending on the subject's condition and brain function at the time of measurement, and typically possess a frequency range of 0–50 Hz and an amplitude of 10–200 µV. Furthermore, EEGs are classified according to their frequency range into alpha waves (α), beta waves (β), delta waves (δ), and theta waves (θ), with distinct characteristics specific to each frequency. While research is underway on technologies to diagnose depression utilizing these characteristics of EEGs, diagnostic methods primarily rely on a single EEG signal as an indicator. Prior art literature

[65535] Published Patent Application No. 10-2022-0094967 (July 6, 2022) The problem to be solved

[0003] As such, according to the present invention, the purpose is to provide a depression diagnosis system using multi-paradigm brainwaves and machine learning that can distinguish between a depressive state and a normal state and diagnose whether a subject has depression by utilizing multi-paradigm brainwaves and machine learning. means of solving the problem

[0004] According to an embodiment of the present invention for achieving such technical challenges, a depression diagnosis system using multiparadigm brainwaves and machine learning may include: a brainwave measurement unit that measures multiple paradigm brainwaves from a subject; a brainwave extraction unit that extracts and preprocesses necessary multiparadigm brainwaves from the measured multiple paradigm brainwaves to minimize noise; a model generation unit that creates a diagnostic model by performing machine learning using the extracted multiparadigm brainwaves; and a diagnosis unit that diagnoses the presence of depression using the generated diagnostic model.

[0005] The above-mentioned brainwave extraction unit can extract and preprocess multiparadigm brainwaves including resting state brainwaves, P3 brainwaves induced by sound stimulation, and LDAEP (Loudness Dependence of Auditory Evoked Potential) brainwaves from multiparadigm brainwaves.

[0006] The above brainwave extraction unit may use a frequency filter that removes brainwave components exceeding ±10 µV, including 1 to 55 Hz in the multiparadigm brainwave measured from the above brainwave measurement unit.

[0007] The above brainwave extraction unit can extract the amplitude value or latent value of the P3 brainwave measured by maintaining the target auditory stimulus relative to the standard auditory stimulus at a preset ratio near the midline of the brain from the above brainwave measurement unit.

[0008] The above brainwave extraction unit can extract a steady-state brainwave by using a frequency filter that removes brainwave components exceeding ±10 μV and includes 1 to 30 Hz from the multiparadigm brainwave measured by the above brainwave measurement unit.

[0009] The above brainwave extraction unit extracts the amplitude values ​​of the N1 peak and P2 peak from the LDAEP brainwave and can calculate the slope according to the change in the performance of multiple auditory stimuli for each subject using a linear regression slope equation from the extracted amplitude values ​​of the N1 peak and P2 peak.

[0010] The above-described brainwave extraction unit detects eye movements collected from electrodes attached to the subject's left and right temples and above and below the right eye, and can minimize noise caused by eye movements by removing characteristic waveforms of eye movements in the frontal lobe region according to the examiner's input signal, or by outputting a target through a display and having the subject fixate on the target. Effects of the invention

[0011] As such, according to the present invention, by generating a diagnostic model of a subject using multiparadigm brainwaves and diagnosing the presence of depression, the limitations of conventional diagnostic classification using a single paradigm can be overcome. Furthermore, by generating and applying a model through machine learning using multiple key brainwave biomarker data, the classification performance of depression with psychopathological diversity can be enhanced. Brief explanation of the drawing

[0012] Figure 1 is a configuration diagram of a depression diagnosis system using multiparadigm brainwaves and machine learning according to one embodiment of the present invention. FIG. 2 is a diagram illustrating the steady-state brainwaves of a subject measured according to one embodiment of the present invention. FIG. 3 is a diagram illustrating an example of the amplitude and latency of a P3 brainwave measured from a subject according to an embodiment of the present invention. Figure 4 is a diagram illustrating an example of calculating the amplitude value of the P3 brainwave shown in Figure 3. FIG. 5 is a diagram illustrating an example of LDAEP brain waves measured from a subject according to an embodiment of the present invention. FIG. 6 is a diagram illustrating an example of generating a diagnostic model in a model generation unit and evaluating the performance of the diagnostic model according to an embodiment of the present invention. Specific details for implementing the invention

[0013] Preferred embodiments according to the present invention will be described in detail below with reference to the attached drawings. In this process, the thickness of lines or the size of components shown in the drawings may be exaggerated for clarity and convenience of explanation.

[0014] Furthermore, the terms described below are defined in consideration of their functions within the present invention, and these may vary depending on the intent or practice of the user or operator. Therefore, the definitions of these terms should be based on the content throughout this specification.

[0015] Figure 1 is a configuration diagram of a depression diagnosis system using multiparadigm brainwaves and machine learning according to one embodiment of the present invention.

[0016] Referring to FIG. 1, a depression diagnosis system (100) using multi-paradigm brainwaves and machine learning according to one embodiment of the present invention may be configured to include a brainwave measurement unit (110), a brainwave extraction unit (120), a model generation unit (130), and a diagnosis unit (140).

[0017] The brainwave measuring unit (110) can measure multiple paradigm brainwaves from the subject.

[0018] The brainwave extraction unit (120) can minimize noise by extracting and preprocessing the necessary multiparadigm brainwaves from the measured multiple paradigm brainwaves.

[0019] The model generation unit (130) can generate a diagnostic model by performing machine learning using the extracted multiparadigm brainwaves.

[0020] The diagnostic unit (140) can diagnose whether there is depression using the generated diagnostic model.

[0021] A depression diagnosis system (100) using multiparadigm brainwaves and machine learning according to an embodiment of the present invention will be described in more detail below with reference to FIGS. 2 to 7.

[0022] In the embodiments described below, the multiparadigm brainwaves extracted to generate a diagnostic model may include resting-state brainwaves, P3 brainwaves induced by sound stimulation, and LDAEP (Loudness Dependence of Auditory Evoked Potential) brainwaves. In addition, the subject's brainwaves are measured using a NeuroScan SynAmps amplifier (Compumedics USA, El Paso, TX, USA) including a head cap equipped with AgCl electrodes according to the International Extended 10-20 System, and 62 electrodes are placed on the head (e.g., FP1, FPz, FP2, AF3, AF4, F7, F5, F3, F1, Fz, F2, F4, F6, F8, FT7, FC5, FC3, FC1, FCz, FC2, FC4, FC6, FT8, T7, C5, C3, C1, Cz, C2, C4, C6, T8, TP7, CP5, CP3, CP1, CPz, CP2, CP4, CP6, TP8, P7, P5, P3, P1, Pz, P2, P4, P6, P8, PO7, PO5, PO3, POz, PO4, PO6, PO8, A specific example is described in which the subject's brainwaves are measured by applying band-pass filtering from 1 to 100 Hz at a sampling rate of 1,000 Hz and maintaining an impedance of 5 kΩ or less. However, the present invention is not necessarily limited thereto, and the depression diagnosis system (100) using multiparadigm brainwaves and machine learning according to the present invention can be similarly applied to diagnose various mental illnesses, including depression, using various types of brainwaves.

[0023] The brainwave measuring unit (110) can measure a multiparadigm brainwave including a resting state brainwave, a P3 brainwave induced by sound stimulation, and an LDAEP brainwave from the subject.

[0024] Specifically, resting-state brainwaves are brainwaves measured when a subject is awake with their eyes closed and without any external stimuli, and can serve as an indicator of the brain's intrinsic activity.

[0025] In the case of P3 brainwaves, the brainwave measuring unit (110) presents a low-frequency, irregular target auditory stimulus to the subject at least once, and the brainwaves that are generated accordingly may be an indicator of the brain's cognitive function. At this time, the brainwave measuring unit (110) may present multiple auditory stimuli to the subject according to a pre-set rule (e.g., random repetition method), and the interval between stimulus presentations may be approximately 2,000 ms.

[0026] Specifically, the P3 brainwave may include an N1 peak and have the most negative amplitude occurring 50 to 150 ms after the target auditory stimulus is performed, or the most positive amplitude occurring 250 to 500 ms after the auditory stimulus is performed.

[0027] Additionally, the brainwave measuring unit (110) may present a standard auditory stimulus and a target auditory stimulus to measure a constant P3 brainwave amplitude and latency value (i.e., time information representing the potential of a preset value representing the latency) near the midline of the brain, which serves as a reference for dividing the left and right sides of the subject's frontal lobe and intermediate lobe. In this case, the target auditory stimulus relative to the standard auditory stimulus may maintain a preset ratio (e.g., 8 to 2), and the standard auditory stimulus may be presented with a preset first decibel and frequency (e.g., 85 dB and 1,000 Hz intensity) and the target auditory stimulus with a preset second decibel and frequency (e.g., 85 dB and 1,500 Hz). The P3 brainwave is induced by the target auditory stimulus.

[0028] In the case of LDAEP brainwaves, the brainwave measuring unit (110) presents sound stimuli of multiple decibels (Decibel, dB) (e.g., 60 dB, 70 dB, 80 dB, 90 dB, 100 dB, etc.) multiple times (e.g., 100 times, etc.) and measures the changes accordingly. It may be a sensory function indicator that detects quantitative changes according to the loudness of the sound. At this time, the sound frequency may be a preset value (e.g., 1,000 Hz).

[0029] FIG. 2 is a diagram illustrating a steady-state brainwave of a subject measured according to an embodiment of the present invention, FIG. 3 is a diagram illustrating an example of the amplitude and latency of a P3 brainwave measured from a subject according to an embodiment of the present invention, FIG. 4 is a diagram illustrating an example of calculating the amplitude value of the P3 brainwave shown in FIG. 3, and FIG. 5 is a diagram illustrating an example of an LDAEP brainwave measured from a subject according to an embodiment of the present invention.

[0030] Referring to FIGS. 2 to 5, the brainwave extraction unit (120) can minimize noise by extracting and preprocessing multiparadigm brainwaves, including a stable state brainwave, P3 brainwave, and LDAEP brainwave, among a plurality of brainwaves measured from a subject.

[0031] To elaborate, in order to extract and preprocess a stable brainwave, the brainwave extraction unit (120) can extract a stable brainwave with minimized noise by using a frequency filter that includes 1 to 55 Hz and removes brainwave components exceeding ±10 µV.

[0032] The brainwave extraction unit (120) can extract the maximum amplitude value during the measurement time or extract all amplitude values ​​in the 250 to 500 ms interval to extract the amplitude value of the P3 brainwave, thereby minimizing the inaccuracy of diagnostic distinction due to individual differences and increasing diagnostic ability.

[0033] In the case of LDAEP brainwaves, the brainwave extraction unit (120) may include 1 to 30 Hz and use a frequency filter that removes brainwave components exceeding ±10 μV, and may extract LDAEP brainwaves including N1 peak and P2 peak through a method similar to the method for extracting P3 brainwaves. At this time, the N1 peak may refer to a maximum negative amplitude value with an amplitude between 80 and 160 ms after stimulus presentation, and the P2 peak may be defined as a maximum positive amplitude value between 130 and 280 ms. Additionally, the slope according to the change in the performance of multiple auditory stimuli in each subject may be calculated using a linear regression slope equation from the amplitude values ​​of the extracted N1 peak and P2 peak.

[0034] The brainwave extraction unit (120) detects eye movements collected from electrodes attached to the left and right temples and above and below the right eye of the subject to minimize noise caused by the subject's eye movements, and can minimize noise caused by eye movements by removing characteristic waveforms of eye movements in the frontal lobe region according to the examiner's input signal, or by outputting a preset target (e.g., a cross) through a display and having the subject fixate on the target.

[0035] FIG. 6 is a diagram illustrating an example of generating a diagnostic model in a model generation unit and evaluating the performance of the diagnostic model according to an embodiment of the present invention.

[0036] Referring to FIG. 6, the model generation unit (130) can train by sequentially accumulating extracted and preprocessed brainwaves through a machine learning t-test-based feature selection method.

[0037] Specifically, the model generation unit (130) can generate a diagnostic model using linear discriminant analysis (LDA) or a support vector machine (SVM). Additionally, the model generation unit (130) can evaluate the diagnostic performance of the diagnostic model using 10×10 cross-validation by configuring the brainwaves extracted and preprocessed at a preset ratio (e.g., 90%) from a plurality of brainwaves from a normal group and a patient group suffering from depression as training data, and configuring the remaining ratio of brainwaves as validation data.

[0038] As such, according to an embodiment of the present invention, optimal depression diagnostic performance can be achieved by using multi-paradigm brainwaves as biomarkers, compared to the conventional technology of recording and analyzing using single-paradigm brainwave indicators.

[0039] The present invention has been described with reference to embodiments illustrated in the drawings, but this is merely illustrative, and those skilled in the art will understand that various modifications and equivalent alternative embodiments are possible therefrom. Accordingly, the true technical scope of protection of the present invention should be determined by the technical spirit of the following claims. Explanation of the symbols

[0040] 100: Depression Diagnostic System 110: Brainwave measurement unit 120: Brainwave Extraction Unit 130: Model Creation Section 140: Diagnostic Department

Claims

Claim 1 A depression diagnosis system using machine learning and multiparadigm brainwaves, comprising: a brainwave measurement unit for measuring multiple paradigm brainwaves from a subject; a frequency filter for extracting a frequency within a preset range from the measured multiple paradigm brainwaves or a brainwave extraction unit for extracting multiparadigm brainwaves using a preset amplitude value of the brainwaves and preprocessing the extracted multiparadigm brainwaves to minimize noise; and a diagnosis unit for diagnosing the presence of depression by inputting the preprocessed multiparadigm brainwaves into a pre-trained diagnostic model, wherein the multiparadigm brainwaves include a resting state brainwave, a P3 brainwave, and an LDAEP (Loudness Dependence of Auditory Evoked Potential) brainwave. Claim 2 A depression diagnosis system using multiparadigm brainwaves and machine learning according to claim 1, wherein a predetermined ratio of multiparadigm brainwaves extracted from a normal group and a group of patients suffering from depression is configured as training data, and the remaining ratio of multiparadigm brainwaves is used as verification data, and a diagnostic model is generated through machine learning to diagnose the presence of depression when multiparadigm brainwaves are input using said training data. Claim 3 A depression diagnosis system using multiparadigm brainwaves and machine learning, wherein the brainwave extraction unit detects eye movements collected from electrodes attached to the left and right temples and the upper and lower right eye of the subject, removes waveforms corresponding to the eye movements, or outputs a target through a display and causes the subject to fixate on the target to minimize noise caused by eye movements. Claim 4 In paragraph 3, the brainwave extraction unit extracts a steady-state brainwave using a frequency filter that removes brainwave components exceeding ±10 μV and includes 1 to 55 Hz from the multiparadigm brainwave measured by the brainwave measurement unit, and is a depression diagnosis system using multiparadigm brainwaves and machine learning. Claim 5 In claim 1, the brainwave extraction unit is a depression diagnosis system using multiparadigm brainwaves and machine learning, wherein the brainwave extraction unit extracts the amplitude value or latency value of a P3 brainwave measured by maintaining a target auditory stimulus relative to a standard auditory stimulus at a preset ratio near the midline of the brain from the brainwave measurement unit. Claim 6 A depression diagnosis system using multiparadigm brainwaves and machine learning, wherein the brainwave extraction unit extracts amplitude values ​​of the N1 peak and P2 peak from the LDAEP brainwave, and calculates slopes according to changes in the performance of multiple auditory stimuli by each subject using a linear regression slope equation from the extracted amplitude values ​​of the N1 peak and P2 peak.

Images