WO2022012653A1 - 预测婴幼儿的语言发展和交流能力的方法和工具 - Google Patents

预测婴幼儿的语言发展和交流能力的方法和工具 Download PDF

Info

Publication number
WO2022012653A1
WO2022012653A1 PCT/CN2021/106684 CN2021106684W WO2022012653A1 WO 2022012653 A1 WO2022012653 A1 WO 2022012653A1 CN 2021106684 W CN2021106684 W CN 2021106684W WO 2022012653 A1 WO2022012653 A1 WO 2022012653A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
language
quantitative data
eeg
external auditory
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2021/106684
Other languages
English (en)
French (fr)
Inventor
黄俊文
诺维茨基尼古拉
冯刚毅
梁廷勋
林鸿生
马古阿克什·拉吉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chinese University of Hong Kong CUHK
Original Assignee
Chinese University of Hong Kong CUHK
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Chinese University of Hong Kong CUHK filed Critical Chinese University of Hong Kong CUHK
Priority to US18/097,578 priority Critical patent/US20250268510A1/en
Priority to AU2021308673A priority patent/AU2021308673A1/en
Priority to GB2302076.1A priority patent/GB2612256B/en
Publication of WO2022012653A1 publication Critical patent/WO2022012653A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/242—Detecting biomagnetic fields, e.g. magnetic fields produced by bioelectric currents
    • A61B5/245—Detecting biomagnetic fields, e.g. magnetic fields produced by bioelectric currents specially adapted for magnetoencephalographic [MEG] signals
    • A61B5/246—Detecting biomagnetic fields, e.g. magnetic fields produced by bioelectric currents specially adapted for magnetoencephalographic [MEG] signals using evoked responses
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/05—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state
    • A61B5/165—Evaluating the state of mind, e.g. depression, anxiety
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/242—Detecting biomagnetic fields, e.g. magnetic fields produced by bioelectric currents
    • A61B5/245—Detecting biomagnetic fields, e.g. magnetic fields produced by bioelectric currents specially adapted for magnetoencephalographic [MEG] signals
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/369—Electroencephalography [EEG]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/369—Electroencephalography [EEG]
    • A61B5/372—Analysis of electroencephalograms
    • A61B5/374—Detecting the frequency distribution of signals, e.g. detecting delta, theta, alpha, beta or gamma waves
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/369—Electroencephalography [EEG]
    • A61B5/377—Electroencephalography [EEG] using evoked responses
    • A61B5/38—Acoustic or auditory stimuli
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/40—Detecting, measuring or recording for evaluating the nervous system
    • A61B5/4058—Detecting, measuring or recording for evaluating the nervous system for evaluating the central nervous system
    • A61B5/4064—Evaluating the brain
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/40—Detecting, measuring or recording for evaluating the nervous system
    • A61B5/4076—Diagnosing or monitoring particular conditions of the nervous system
    • A61B5/4088—Diagnosing of monitoring cognitive diseases, e.g. Alzheimer, prion diseases or dementia
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/48—Other medical applications
    • A61B5/4803—Speech analysis specially adapted for diagnostic purposes
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7225—Details of analogue processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B2503/00—Evaluating a particular growth phase or type of persons or animals
    • A61B2503/04—Babies, e.g. for SIDS detection
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B2503/00—Evaluating a particular growth phase or type of persons or animals
    • A61B2503/06—Children, e.g. for attention deficit diagnosis

Definitions

  • the present application generally relates to the field of convergence of medicine and artificial intelligence, and more particularly, the present application provides prediction of language and communication in healthy infants or young children based on electroencephalography (EEG) or magnetoencephalography (MEG) detection and artificial intelligence training Methods, tools and systems for differences in developmental levels of competency normality.
  • EEG electroencephalography
  • MEG magnetoencephalography
  • Language and communication skills are fundamental to individual human development and affect all aspects of later childhood and adulthood. According to reports, language and communication skills not only affect children's academic success, but also have significant predictive significance for future employment and salary levels. If the normal development level of language and communication skills can be predicted in the early stage of life, and targeted active intervention and training can be carried out according to the predicted results, it will be beneficial to the development of individuals.
  • the present application provides a method for predicting differences in developmental levels of normality of language and communication skills in healthy infants or young children, the method comprising:
  • EEG electroencephalogram
  • MEG magnetoencephalogram
  • the quantitative data includes measurement index data representing the central nervous system elicited by the external auditory stimulation
  • a predictive score for the difference in developmental level of normality of language and communication skills of the healthy infant or toddler is generated.
  • the method further comprises: obtaining electroencephalogram (EEG) or magnetoencephalography (MEG) waveform data of the healthy infant or toddler in a resting state; and
  • Corresponding quantitative data is extracted from the EEG or MEG waveform of the healthy infant or toddler in a resting state, wherein the quantitative data includes measurement index data indicating that the central nervous system is in a resting state.
  • the external auditory stimuli include a plurality of different stimuli, such as sound stimuli.
  • the quantitative data includes measurement data indicative of the central nervous system response to one or more of the plurality of external auditory stimuli and at rest.
  • the quantitative data includes data characterizing the functional activity of auditory center-related processing pathways.
  • the quantitative data includes data characterizing the functional activity of the inferior colliculus and centers connected to it.
  • the quantitative data includes data characterizing the functional activity of the primary auditory cortex, which includes the Hector's gyrus or centers connected thereto.
  • the external auditory stimuli come from language and speech signals that use pitch patterns to express meaning at the lexical or sentence level.
  • the external auditory stimuli are from Chinese, including Mandarin and Cantonese, among others.
  • the external auditory stimuli come from language and speech signals that use pitch patterns to express meaning at the phrase or sentence level.
  • the external auditory stimuli are from Chinese, including Mandarin and Cantonese, among others.
  • the external auditory stimulus is Hanyu Pinyin of one or more tones.
  • the EEG or MEG is performed while the healthy infant or toddler is in natural sleep or in an awake state.
  • the healthy infant or toddler is under 18 months of age. In some embodiments, regarding the acquisition of the training data set, the plurality of healthy infants or young children are under 18 months of age when the training data set is obtained from the plurality of healthy infants or young children.
  • the machine learning classifier is a support vector machine (SVM) and the predicted difference in developmental level of normality of language and communication skills is high or low, or a continuous measure.
  • SVM support vector machine
  • the machine learning classifier is a support vector regression algorithm (SVR) or a ranking support vector machine (RankSVM), and the predicted differences in developmental levels of normality of language and communication skills are quantified.
  • SVR support vector regression algorithm
  • RankSVM ranking support vector machine
  • analyzing the quantitative data further includes analyzing one or more variables associated with differences in developmental levels of language and communication normality, such as gender and birth data.
  • quantitative data is extracted from the EEG or MEG waveform by a method selected from: Automatic peak detection, Fast Fourier Transform, Autocorrelation, Root Mean Square (Root-Mean-Square, RMS), Morlet Wavelet Transform, Discrete Wavelet Transform, Wavelet Scattering, Stimulus-Response Cross-correlation, Empirical Mode Decomposition processing (Empirical Mode Decomposition), Hilbert-Huang Transform (Hilbert-Huang Transform).
  • a method selected from: Automatic peak detection, Fast Fourier Transform, Autocorrelation, Root Mean Square (Root-Mean-Square, RMS), Morlet Wavelet Transform, Discrete Wavelet Transform, Wavelet Scattering, Stimulus-Response Cross-correlation, Empirical Mode Decomposition processing (Empirical Mode Decomposition), Hilbert-Huang Transform (Hilbert-Huang Transform).
  • the quantitative data includes one or more of the following: time-domain peak amplitude (time-domain peak amplitude), time-domain peak latency (time-domain peak latency), fundamental frequency (F0) , harmonics, signal-to-noise ratio, RMS amplitude, correlation coefficient, inter-trial phase coherence, phase lock coefficient -locking coefficient), response consistency (response consistency), pitch strength (pitch strength), pitch error (pitch error), pitch-tracking accuracy (pitch-tracking accuracy).
  • extracting quantitative data from the EEG waveform includes segmenting the EEG waveform into segments by stimulus onset markers, and converting the EEG waveform into segments using a fast Fourier transform (FFT) in sliding time windows with overlapping windows. Each segment is transformed to the frequency domain.
  • FFT fast Fourier transform
  • quantitative data includes a matrix of T*(E*3)*F tensors for the healthy infant or toddler, where T is the number of time windows and E is each stimulus (eg, a The number of segments of a tone of Hanyu Pinyin), F is the number of frequency bins from the FFT analysis, and the (E*3)*F matrix for each T is first normalized within the row, then in the The columns are normalized to remove the influence of the absolute magnitude of the spectrum and preserve only the frequency-dependent time-varying patterns.
  • the machine learning classifier is a support vector machine (SVM) using parameters including a Gaussian kernel, C, and gamma.
  • the classification by the machine learning classifier is cross-validated, and the result of the cross-validation is the average accuracy, specificity, at a multiple (eg, 5-fold, 10-fold, or more) , Sensitivity, AUC, Parity, Correlation, or a combination of the above.
  • the present application provides a computer-readable medium storing a plurality of instructions that, when executed by a processor of a computer system, control the computer system to perform various operations to implement the method of the first aspect.
  • the application provides a computer system, comprising:
  • One or more processors for executing instructions stored on the computer-readable medium.
  • the application provides a computer system, comprising:
  • a processor communicatively coupled to the memory and configured to perform operations to implement the method of the first aspect.
  • the application provides a system for predicting differences in developmental level of normality of language and communication skills in a healthy infant or young child, comprising means for implementing the method of the first aspect.
  • the present application provides a method of generating a machine learning classifier for predicting differences in developmental levels of normality of language and communication skills in healthy infants or young children, the method comprising:
  • EEG electroencephalogram
  • MEG magnetoencephalography
  • Quantitative data extracted from the EEG or MEG waveform data and the qualitative or quantitative data characterizing language and communication skills are fed into a machine learning classifier for training to generate a model for predicting normal language and communication skills in healthy infants or young children
  • a machine learning classifier for differences in sexual developmental levels are fed into a machine learning classifier for training to generate a model for predicting normal language and communication skills in healthy infants or young children.
  • Figure 1 shows a schematic diagram of the operation to obtain auditory-evoked EEG data from an infant, placing electrodes on the infant's head and delivering auditory stimuli by inserting earphones while the infant is naturally sleeping (or awake).
  • Figure 2 shows the EEG responses to different tones in Chinese in the exemplary scheme of the present application.
  • three tones were embedded in the syllable /ga/: Cantonese tones /ga2/, Mandarin tones /ga3/, and Cantonese tones /ga4/ (infants in this study were raised in Cantonese-speaking families).
  • Early EEG responses truly reflect the time course and spectral composition of auditory stimuli.
  • Panel A shows the time course of auditory stimuli (left column) and their corresponding mean EEG responses (middle column) and their corresponding mean spectra of EEG responses (right column).
  • Panel B shows the spectrum of auditory stimuli (far left column) and the mean EEG spectrum of auditory stimuli in infants at 1, 3, 6, 9 and 15 months of age (five remaining columns). The older the infant, the greater the overlap between auditory and EEG spectral structures.
  • Figure 3 shows the predicted classification results of the EEG responses of three Chinese lexical sounds /ga2/, /ga3/ and /ga4/ by applying a support vector machine (SVM) classifier to infants of different months of age, which can reflect the infant brain’s the ability to discriminate these sounds.
  • SVM support vector machine
  • the EEG data in response to the three tones were input to the SVM classifier, and the data were Fourier transformed every 50ms interval unit.
  • the classification was cross-validated by training on 9/10 of the data and testing on the remaining 1/10.
  • the random level of classification accuracy for 3 tones was 33%.
  • the graph illustrates that the classification accuracy has two peaks along the time course of EEG responses, and that the accuracy increases with the infant's age in months.
  • the figure illustrates classification predictions for individuals. Do the same for each individual individual.
  • Figure 4 shows that the machine learning model (SVM classifier) predicts with greater than 80% accuracy the late gestures (Later Gestures) and Interactive Acts), the machine learning model (SVM classifier) was trained and cross-validated based on EEG data obtained at about 6 months of age.
  • the figure shows the bootstrapped distribution (10,000-fold) of the classification accuracies used to assign the two measures of language and communication abilities of an individual to the lower or upper half (median split) at 15 months of age.
  • the average permutation marker classification accuracy was at the 50% random level.
  • the mean of the true marker classification was higher than 95% of the permuted marker classification in both cases (horizontal dashed line), indicating that the classification was significantly higher than random (p ⁇ 0.05).
  • the present application provides a technical platform that mainly relies on neural data from infancy (such as EEG or MEG extraction data) to predict the development trend of language and communication skills in the future.
  • Machine learning techniques are used to train classifiers to make such predictions.
  • the training data set required for training the classifier it is necessary to obtain a certain amount of neural data of infant samples, then evaluate the language and communication skills, and conduct cross-validation and training on the two parts of the results.
  • the machine learning classifier can be used to predict differences in normal developmental levels of language and communication in other infants.
  • the methods of the present application may include placing surface electrodes around the infant's head while performing EEG recordings while the infant is applying auditory stimuli during natural sleep (see Figure 1); measuring neural responses from the subcortical auditory brainstem , in particular, how the infant's auditory brainstem follows the frequency (pitch) of stimuli (see Figure 2 for exemplary results), thereby exploiting the ability to observe tone-following in the subcortical auditory brainstem and connected neural structures to predict the infant's future developmental level of language and communication skills.
  • the present application provides a method for predicting differences in developmental levels of normality of language and communication skills in healthy infants or young children, the method comprising:
  • EEG electroencephalogram
  • MEG magnetoencephalogram
  • the quantitative data includes measurement index data representing the central nervous system elicited by the external auditory stimulation
  • a predictive score for the difference in developmental level of normality of language and communication skills of the healthy infant or toddler is generated.
  • the method further comprises: obtaining electroencephalogram (EEG) or magnetoencephalography (MEG) waveform data of the healthy infant or toddler in a resting state; and obtaining data from said healthy infant or young child in a resting state
  • EEG electroencephalogram
  • MEG magnetoencephalography
  • Corresponding quantitative data is extracted from EEG or MEG waveforms of healthy infants or young children, wherein the quantitative data includes measurement index data indicating that the central nervous system is in a resting state.
  • the EEG or MEG is performed while the healthy infant or toddler is in natural sleep or in an awake state.
  • the healthy infant or toddler of the present application is under 18 months of age.
  • the plurality of healthy infants or young children are under 18 months of age when the training data set is obtained from a plurality of healthy infants or young children.
  • health refers to a state of health at least in terms of listening, speaking, intelligence, etc. related to language and communication skills, and includes, but is not limited to, a state of complete health.
  • infant and toddler are based on the usual age criteria, where infants are 0-12 months old and young children are 1-3 years old.
  • the “external auditory stimuli” herein refers to auditory stimuli in terms of language, and can be selected accordingly according to the structure and composition of different languages. It should be understood that the language of the external auditory stimuli should match the language to which the test individual will be exposed to in the future growth environment. Typically, a single external auditory stimulus should be a relatively simple unit of language, such as letters/pinyin, words/single words (different tones), phrases, and the like. In some embodiments, the external auditory stimuli come from language and speech signals that use pitch patterns to express meaning at the lexical or sentence level. In some embodiments, the external auditory stimuli are from Chinese, including Mandarin and Cantonese, among others.
  • the external auditory stimuli come from language and speech signals that use pitch patterns to express meaning at the phrase or sentence level.
  • the external auditory stimuli are from Chinese, including Mandarin and Cantonese, among others.
  • Chinese Choinese uses pitch patterns to express meaning at the lexical or sentence meaning level, so it is particularly suitable for this application
  • the applications of the inventions of this application are not limited to Therefore, other languages such as English, French, German, Spanish, Portuguese, Japanese, Korean, etc. are also suitable for this application, since almost all languages use pitch patterns to express meaning at least at the lexical or sentence level.
  • the external auditory stimulus includes a plurality of different stimuli.
  • the quantitative data includes measurement data indicative of the central nervous system response to one or more of the plurality of external auditory stimuli and at rest.
  • the multiple external auditory stimuli may be multiple tones of one Hanyu Pinyin.
  • EEG electroencephalogram
  • MEG magnetoencephalography
  • quantitative data may include data characterizing the functional activity of processing pathways associated with auditory centers; or data characterizing the functional activity of the inferior colliculus and centers connected to it; or data characterizing the functional activity of the primary auditory cortex, which includes Hector's gyrus or the center connected to it.
  • EEG or MEG waveforms There are various methods available for extracting the above quantitative data from EEG or MEG waveforms, including but not limited to, Automatic peak detection, Fast Fourier Transform, Autocorrelation, Root Mean Square (Root Mean Square) -Mean-Square, RMS), Morlet Wavelet Transform, Discrete Wavelet Transform, Wavelet Scattering, Stimulus-Response Cross-correlation, Empirical Mode Decomposition Empirical Mode Decomposition, Hilbert-Huang Transform, or a combination of the above.
  • Automatic peak detection Fast Fourier Transform
  • Autocorrelation Root Mean Square (Root Mean Square) -Mean-Square, RMS)
  • Morlet Wavelet Transform Discrete Wavelet Transform
  • Wavelet Scattering Stimulus-Response Cross-correlation
  • Empirical Mode Decomposition Empirical Mode Decomposition
  • Hilbert-Huang Transform or a combination of the above.
  • Available quantitative data can be extracted from EEG or MEG waveforms in a variety of forms, including, but not limited to, time-domain peak amplitude, time-domain peak latency, fundamental frequency ( fundamental frequency, F0), harmonics, signal-to-noise ratio, RMS amplitude, correlation coefficient, inter-trial phase coherence, phase-locking coefficient, response consistency, pitch strength, pitch error, pitch-tracking accuracy, or a combination of the above .
  • the process of extracting quantitative data from an EEG waveform may include segmenting the EEG waveform into segments by stimulus onset markers, and utilizing Fast Fourier Transform (FFT) in sliding time windows with overlapping windows Transform each segment to the frequency domain.
  • FFT Fast Fourier Transform
  • quantitative data includes a matrix of T*(E*3)*F tensors for the healthy infant or toddler, where T is the number of time windows and E is each stimulus (eg, a The number of segments of a tone of Hanyu Pinyin), F is the number of frequency bins from the FFT analysis, and the (E*3)*F matrix for each T is first normalized within the row, then in the The columns are normalized to remove the influence of the absolute magnitude of the spectrum and preserve only the frequency-dependent time-varying patterns.
  • the quantitative data is analyzed using a machine learning classifier.
  • the general working principle of machine-learning classifiers is known.
  • the training datasets on which the machine learning classifier work of the present application is based are derived from multiple similar individuals belonging to the same population as the test individuals.
  • the "same population” or “same group” referred to herein refers to a collection of individuals at the same linguistic and cultural level. For example, if test individuals will be raised in a Mandarin-speaking environment, a Mandarin-based training data set should be used; similarly, if test individuals will be raised in a Cantonese-speaking environment, a Cantonese-based training data set should be used; And so on.
  • the multiple individuals used to create the training dataset may have undergone the same external auditory stimulation and EEG/MEG quantitative data extraction as the test individuals, and the multiple individuals have also undergone verbal and subsequent (eg, 6-12 months later) Practical assessment of communication skills and use the results to create a training dataset.
  • verbal and subsequent e.g, 6-12 months later
  • the machine learning classifier is a support vector machine (SVM) using parameters including a Gaussian kernel, C, and gamma.
  • SVM support vector machine
  • the machine learning classifier is a support vector machine (SVM) and the predicted difference in developmental level of normality of language and communication skills is high or low, or a continuous measure.
  • SVM support vector machine
  • the machine learning classifier is a support vector regression algorithm (SVR) or a ranking support vector machine (RankSVM), and the predicted differences in developmental levels of normality of language and communication skills are quantified.
  • SVR support vector regression algorithm
  • RankSVM ranking support vector machine
  • the classification by the machine learning classifier is cross-validated, and the result of the cross-validation is the average accuracy, specificity, at a multiple (eg, 5-fold, 10-fold, or more) , Sensitivity, AUC, Parity, Correlation, or a combination of the above.
  • analyzing the quantitative data further includes analyzing one or more variables related to language development, such as gender and birth data.
  • the present application provides a computer-readable medium storing a plurality of instructions that, when executed by a processor of a computer system, control the computer system to perform various operations to implement the method of the first aspect.
  • the application provides a computer system, comprising:
  • One or more processors for executing instructions stored on the computer-readable medium.
  • the application provides a computer system, comprising:
  • a processor communicatively coupled to the memory and configured to perform operations to implement the method of the first aspect.
  • the application provides a system for predicting differences in developmental level of normality of language and communication skills in a healthy infant or young child, comprising means for implementing the method of the first aspect.
  • the present application provides a method of generating a machine learning classifier for predicting differences in developmental levels of normality of language and communication skills in healthy infants or young children, the method comprising:
  • EEG electroencephalogram
  • MEG magnetoencephalography
  • Quantitative data extracted from the EEG or MEG waveform data and the qualitative or quantitative data characterizing language and communication skills are fed into a machine learning classifier for training to generate a model for predicting normal language and communication skills in healthy infants or young children
  • a machine learning classifier for differences in sexual developmental levels are fed into a machine learning classifier for training to generate a model for predicting normal language and communication skills in healthy infants or young children.
  • the technical solutions and technical features of the first aspect are also applicable to the second to sixth aspects under the condition of no contradiction or adaptive adjustment.
  • the invention of the sixth aspect relates to the acquisition of quantitative data in EEG or MEG waveform data and the acquisition of qualitative or quantitative data to characterize language and communication ability later, the implementation of the former may refer to the method of the first aspect;
  • any embodiment of the present invention may be implemented in a modular or integrated manner using hardware (eg, application specific integrated circuits or field programmable gate arrays) and/or in the form of control logic using computer software with a general purpose programmable processor .
  • a processor includes a single-core processor, a multi-core processor on the same integrated chip, or multiple processing units on a single circuit board or networked.
  • Any software components or functions described in this application may be implemented to be implemented by a processor using any suitable computer language such as eg Java, C, C++, C#, object oriented C, Swift or a scripting language (eg using eg software code executed using conventional or object-oriented technologies such as Perl or Python).
  • the software code may be stored as a series of instructions or commands on a computer readable medium for storage and/or transmission, suitable media including random access memory (RAM), read only memory (ROM), magnetic media such as a hard drive or floppy disk) or optical media (such as compact disc (CD) or DVD (digital versatile disc)), flash memory, etc.
  • RAM random access memory
  • ROM read only memory
  • magnetic media such as a hard drive or floppy disk
  • optical media such as compact disc (CD) or DVD (digital versatile disc)
  • flash memory etc.
  • the computer-readable medium can be any combination of such storage or transmission devices.
  • Such programs may also be encoded and transmitted using carrier signals suitable for transmission via wired, optical and/or wireless networks (including the Internet) consistent with various schemes.
  • a computer-readable medium according to one embodiment of the present invention can be produced using a data signal encoded with such a program.
  • Computer-readable media encoded with program code may be packaged with compatible devices or provided separately from other devices (eg, via Internet download). Any such computer-readable medium may reside on or within a single computer product (eg, a hard drive, a CD, or an entire computer system), and may reside on or within different computer products within a system or network.
  • the computer system may include a monitor, printer, or other suitable display for providing any of the results mentioned herein to the user.
  • any of the methods described herein can be performed, in whole or in part, using a computer system that includes one or more processors that can be configured to perform the steps.
  • embodiments may relate to a computer system configured to perform the steps of any method described herein, possibly with different components for performing the respective steps or respective groups of steps.
  • steps of the methods herein are presented as numbered steps, they can be performed simultaneously or in a different order. Additionally, portions of these steps may be used with other steps portions of other methods. Furthermore, all or part of the steps may be optional. Additionally, any step of any method can be performed using modules, circuits, or other means for performing the steps.
  • This study involves the use of neurological measures to construct predictive algorithms for the prediction of developmental-level outcomes of language and communication normality at the individual level.
  • this study used EEG data as a neurological measure.
  • This study mainly requires: 1) an EEG testing procedure, and that this procedure is preferably brief and contains the most informative auditory stimuli, 2) existing data from a group of children who have been tested on this EEG procedure, And these children's subsequent language and communication outcomes are known, and 3) the algorithm used to correlate EEG with actual assessment outcome data for language and communication has high specificity and sensitivity in making predictions.
  • Panel A shows the time course of auditory stimuli (left column) and mean EEG responses (middle column) and the mean spectrogram of EEG responses (right column) overlaid with the contours surrounding the upper quartile of the auditory stimulus spectrum. The spectral amplitude densities of the EEG responses were centered within the upper quartile of the auditory stimulus spectral density.
  • Panel B shows the spectra of auditory stimuli (left column) and their corresponding mean EEG spectra (remaining five columns) in infants at 1, 3, 6, 9 and 15 months of age. The older the infant, the more obvious the overlap between auditory and EEG spectral structures.
  • the collected auditory-evoked neural EEG responses were analyzed according to a series of automated procedures performed in Matlab, such as fast Fourier transform, autocorrelation and root mean square (RMS). Measurements such as fundamental frequency (F0), harmonics, signal-to-noise ratio and RMS amplitude were extracted following the reported procedure.
  • Matlab such as fast Fourier transform, autocorrelation and root mean square (RMS).
  • Measurements such as fundamental frequency (F0), harmonics, signal-to-noise ratio and RMS amplitude were extracted following the reported procedure.
  • the algorithm of the present study used neural EEG measures to classify individual infants into better or poorer language outcome groups several months after EEG recordings were made during infancy. Binomial classification was implemented in this study using a support vector machine (SVM) procedure with a 10-point cross-validation procedure and bootstrap-permutation statistics.
  • SVM support vector machine
  • This study performed a two-step SVM classification: at the individual subject level and at the group level of 88 subjects, of which 30 subjects underwent EEG data collection at approximately 6 months of age and at approximately 1 year of age Conduct language and communication competency (CCDI) assessments and data collection. This study used EEG data to predict the level of language and communication development after 6 months.
  • CCDI Conduct language and communication competency
  • the filtered EEG was segmented into segments by stimulus onset markers.
  • a set of 1000 EEG segments was obtained for each of the three tones.
  • Each segment was transformed to the frequency domain using FFT in a sliding window of 50ms with 50% overlap between windows.
  • the (E*3)*F matrix for each T is normalized first in the rows and then in the columns, thereby removing the effect of the absolute magnitude of the spectrum and leaving only the frequency-dependent modes.
  • the classifications were cross-validated 10-fold, ie, the data was randomly divided into 10 parts, and for each part, the classifier was trained on 90% of the data and tested on 10% of the data.
  • the result of the cross-validation process is the average precision across multiples. As a result, for each subject, a vector of T classification accuracy was obtained.
  • Figure 3 shows the discriminative ability of the infant brain can be reflected by applying a support vector machine (SVM) classifier to the predicted classification results of the EEG responses to three Chinese lexical sounds /ga2/, /ga3/ and /ga4/.
  • SVM support vector machine
  • Fourier transforms of 50ms interval cells along the time course of the EEG responses to the three tones were input to the SVM classifier.
  • the classification was cross-validated by training on 9/10 of the data and testing on the remaining 1/10.
  • the chance level of classification accuracy for 3 tones is 33%.
  • the graph illustrates that the classification accuracy has two peaks along the time course of responses, and that the accuracy increases with the infant's age in months.
  • the figure illustrates the classification across individuals. The same operation is performed in each individual individual.
  • S is the number of subjects with verbal and communication results from EEG recordings and CCDI.
  • S is the number of subjects with verbal and communication results from EEG recordings and CCDI.
  • These matrices were created for each possible combination of EEG recording time and language and communication outcome generation time, and were normalized within each time interval. These combinations cover partially overlapping subsets of the original 83 subjects. Language and communication outcomes within each of these age combinations were transformed into binary variables with median splits. Each predictor matrix is normalized within each time interval.
  • a binary SVM was then performed for each language and communication outcome and for each possible combination of language and communication outcome time and EEG recording time.
  • Confidence intervals for each classification were calculated using a combination of bootstrapping and permutation.
  • the rows of the matrix (subjects) were resampled with permutation, and a 10-point cross-validated SVM classification was performed on the actual and randomly permuted subject labels. This process was repeated 10,000 times to obtain the distribution of actual prediction accuracy and permuted prediction accuracy (see Figure 4).
  • the mean of the true label accuracy is higher than 95% of the distribution of the permuted label accuracy, and the accuracy is considered to be significantly higher than the random level.
  • Example 1 On the basis of Example 1, the inventors of the present application expanded the research group to 118 children (including the group of children in Example 1, keeping the experimental protocol unchanged). The results of the expanded sample study showed that the predictive model constructed in this application was able to classify children as below or above average, or below or above 25% percentile, with an accuracy of up to 0.92 (AUC).

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Molecular Biology (AREA)
  • Surgery (AREA)
  • Biophysics (AREA)
  • Veterinary Medicine (AREA)
  • Psychiatry (AREA)
  • Psychology (AREA)
  • Artificial Intelligence (AREA)
  • Physiology (AREA)
  • Signal Processing (AREA)
  • Neurology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Acoustics & Sound (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Databases & Information Systems (AREA)
  • Epidemiology (AREA)
  • Primary Health Care (AREA)
  • Child & Adolescent Psychology (AREA)
  • Hospice & Palliative Care (AREA)
  • Developmental Disabilities (AREA)
  • Neurosurgery (AREA)
  • Radiology & Medical Imaging (AREA)
  • Power Engineering (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Educational Technology (AREA)
  • Social Psychology (AREA)

Abstract

预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的方法。该方法是基于健康婴儿或幼儿对外部听觉刺激引起的脑电图(EEG)或脑磁图(MEG)波形数据，从中获取表征中枢神经系统对所述外部听觉刺激引起的定量数据，并用此数据集训练机器学习分类器，从而给出预测结果。还提供了相关的工具和集成系统。

Description

预测婴幼儿的语言发展和交流能力的方法和工具
相关申请
本申请要求于2020年7月17日提交的中国专利申请202010692253.5号的优先权，通过援引加入的方式将该申请的全部内容并入本文，用于所有目的。
技术领域
本申请大体涉及医学和人工智能的融合领域，更具而言，本申请提供了基于脑电图(EEG)或脑磁图(MEG)检测以及人工智能训练来预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的方法、工具和系统。
背景技术
语言和交流能力是人类个体发展的基础，影响着儿童后期和成年期的各个方面。据报道，语言和交流能力不仅影响儿童的学业方面的成功，而且也对于日后的就业和薪资水平的也具有显著预测意义。如果能在生命早期对于语言和交流能力正常性的发展水平进行预测，并根据预测结果进行有针对性的主动干预和训练，对于个体的发展会是有利的。
发明概述
第一方面，本申请提供了一种预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的方法，所述方法包括：
获得所述健康婴儿或幼儿对外部听觉刺激引发的脑电图(EEG)或脑磁图(MEG)波形数据；
从所述EEG或MEG波形数据中提取定量数据，其中所述定量数据包括表征中枢神经系统对所述外部听觉刺激所引起的测量指标数据；
使用机器学习分类器分析所述定量数据，所述机器学习分类器被训练为能预测所述健康婴儿或幼儿所属人群的语言和交流能力正常性的发展水平差异的预测评分，其中所述训练是基于从具有已知语言和交流能力正常性的发展水平的多个健康婴儿或幼儿获得的训练数据集的对应定量数据；以及
基于所述机器学习分类器输出的结果，生成所述健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的预测评分。
在一些实施方案中，该方法还包括：获得处于静息状态下的所述健康婴儿或幼儿的脑电图(EEG)或脑磁图(MEG)波形数据；和
从处于静息状态的所述健康婴儿或幼儿的EEG或MEG波形中提取相应的定量数据，其中所述定量数据包括指示中枢神经系统处于静息状态下的测量指标数据。
在一些实施方案中，外部听觉刺激包括多个不同的刺激，例如声音刺激。在一些实施方案中，定量数据包括指示中枢神经系统对所述多个外部听觉刺激中一个或多个的反应与处于静息状态下的测量指标数据。
在一些实施方案中，定量数据包括表征听觉中枢相关的处理通路功能活动的数据。
在一些实施方案中，定量数据包括表征下丘及与其连接的中心的功能活动的数据。
在一些实施方案中，定量数据包括表征初级听觉皮层的功能活动的数据，所述初级听觉皮层包括赫氏脑回或与其连接的中心。
在一些实施方案中，外部听觉刺激来自使用音高模式来在词汇或句意水平表达含义的语言和语音信号。在一些实施方案中，外部听觉刺激来自汉语，包括普通话和粤语等。
在一些实施方案中，外部听觉刺激来自使用音高模式来在短语或句子水平表达含义的语言和语音信号。在一些实施方案中，外部听觉刺激来自汉语，包括普通话和粤语等。
在一些实施方案中，外部听觉刺激是一个或多个声调的汉语拼音。
在一些实施方案中，在所述健康婴儿或幼儿处于自然睡眠时或处于清醒状态下执行所述EEG或MEG。
在一些实施方案中，所述健康婴儿或幼儿的月龄为18个月以下。在一些实施方案中，关于训练数据集的获取，当从所述多个健康婴儿或幼儿获得所述训练数据集时，所述多个健康婴儿或幼儿的月龄为18个月以下。
在一些实施方案中，机器学习分类器是支持向量机(SVM)，并且所述预测的语言和交流能力的正常性的发展水平差异是高水平或低水平，或连续性评价方式。
在一些实施方案中，机器学习分类器是支持向量回归算法(SVR)或排序支持向量机(RankSVM)，并且所预测的语言和交流能力正常性的发展水平差异是量化的。
在一些实施方案中，分析所述定量数据还包括分析与语言和交流能力正常性的发展水平差异相关的一个或多个变量，例如性别和出生数据。
在一些实施方案中，通过选自以下的方法从所述EEG或MEG波形提取定量数据：自动峰值检测(Automatic peak detection)，快速傅立叶变换(Fast Fourier Transform)，自相关(Autocorrelation)，均方根(Root-Mean-Square，RMS),Morlet小波变换(Morlet Wavelet Transform)，离散小波变换(Discrete Wavelet Transform)，小波散射(Wavelet Scattering)，刺激响应互相关(Stimulus-Response Cross-correlation)，经验模式分解的处理(Empirical Mode Decomposition)，希尔伯特-黄变换(Hilbert-Huang Transform)。在一些实施方案中，定量数据包括以下中的一个或多个：时域峰值幅度(time-domain peak amplitude)，时域峰值时延(time-domain peak latency)，基频(fundamental frequency，F0)，谐波(harmonics)，信噪比(signal-to-noise ratio)，RMS幅度(RMS amplitude)，相关系数(correlation coefficient)，试验间相位相干(inter-trial phase coherence)，锁相系数(phase-locking coefficient)，响应一致性(response consistency)，音高强度(pitch strength)，音高误差(pitch error)，音高跟踪精度(pitch-tracking accuracy)。
在一些实施方案中，从所述EEG波形提取定量数据包括：通过刺激起始标记将所述EEG波形分割成节段，并且在具有重叠窗口的滑动时间窗口中，利用快速傅立叶变换(FFT)将每个节段变换到频域(frequency domain)。在一些实施方案中，定量数据包括所述健康婴儿或幼儿的T*(E*3)*F张量表示的矩阵，其中T是所述时间窗口的数量，E是每个刺激(例如，一个汉语拼音的一个声调)的节段的数量，F是来自FFT分析的频率间隔(frequency bin)的数量，并且每个T的(E*3)*F矩阵首先在行内被归一化，然后在列中被归一化，从而消除频谱的绝对幅度的影响并且仅保留频率相关随时间变化的模式。在一些实施方案中，机器学习分类器是使用包括高斯核、C和伽马的参数的支持向量机(SVM)。
在一些实施方案中，由所述机器学习分类器进行的分类是通过交叉验证的，并且交叉验证的结果是在一定倍数(例如5倍、10倍或更多倍)的平均准确度、特异性、灵敏度、AUC、奇偶率、相关系数或以上的组合。
第二方面，本申请提供了一种计算机可读介质，其存储多个指令，所述多个指令当被计算机系统的处理器执行时控制计算机系统执行多种操作以实施第一方面的方法。
第三方面，本申请提供了一种计算机系统，包括：
第二方面所述的计算机可读介质；和
用于执行存储在所述计算机可读介质上的指令一个或多个处理器。
第四方面，本申请提供了一种计算机系统，包括：
存储器；和
处理器，所述处理器与所述存储器可通讯地连接并且被配置为执行多种操作以实施第一方面的方法。
第五方面，本申请提供了一种预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的系统，包括用于实施第一方面的方法的多种装置。
第六方面，本申请提供了一种生成用于预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的机器学习分类器的方法，所述方法包括：
获得一名或多名健康婴儿或幼儿对外部听觉刺激引发的脑电图(EEG)或脑磁图(MEG)波形数据，并且从所述EEG或MEG波形数据中提取定量数据，其中所述定量数据包括表征中枢神经系统对所述外部听觉刺激所引起的测量指标数据；
在一定时间(例如6-12个月)后，对所述一名或多名健康婴儿或幼儿进行语言和交流能力测试，并获得表征语言和交流能力的定性或定量数据；
将从所述EEG或MEG波形数据中提取的定量数据和所述表征语言和交流能力的定性或定量数据输入机器学习分类器进行训练，以生成用于预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的机器学习分类器。
附图简要描述
图1显示了从婴儿获得听觉诱发EEG数据的操作示意图，在婴儿头部上放置电极，在婴儿自然睡眠时(或清醒状态下)通过插入耳机传递听觉刺激。
图2显示了在本申请的示例性方案中，对汉语不同声调的EEG应答。在外部听觉刺激中，将三种声调嵌入音节/ga/：粤语声调/ga2/,普通话声调/ga3/和粤语声调/ga4/(本研究中的婴儿在粤语家庭中成长)。早期EEG反应真实地反映了听觉刺激的时程和频谱组成。图A显示听觉刺激(左栏)和其对应的平均EEG响应(中间栏)的时间进程以及其对应的EEG响应的平均谱图(右栏)。EEG响应的频谱幅度密度集中在听觉刺激谱密度的上四分位位置。图B显示了听觉刺激的频谱(最左栏)和在1、3、6、9和15月龄的婴儿的对听觉刺激的平均EEG谱图(其余五栏)。婴儿月龄越大，听觉频谱结构和EEG频谱结构之间的重叠越明显。
图3显示了通过将支持向量机(SVM)分类器应用于不同月龄婴儿对三个汉语词汇音/ga2/、/ga3/和/ga4/的EEG反应的预测分类结果，可以反映婴儿大脑对这些语音的辨别能力。向SVM分类器输入对三个声调响应的EEG数据，数据每50ms间隔单元经过傅立叶变换。对于随机分割数据的10份中的每一份，通过对9/10的数据进行训练和对其余1/10的数据进行测试，对分类进行交叉验证。对于3个声调的分类精度的随机水平是33％。该图说明分类准确度沿着EEG反应 的时间进程具有两个峰值，并且准确度随着婴儿的月龄增大而增长。该图说明了对个体的分类预测。在每个单独个体中进行相同的操作。
图4显示了机器学习模型(SVM分类器)以高于80％的精确度预测两种语言和交流能力测量(来自汉语沟通发展评定量表(Chinese Communication Development Inventories)的后期姿态(Later Gestures)和交互动作(Interactive Acts))，该机器学习模型(SVM分类器)基于约6月龄取得的EEG数据进行训练并进行交叉验证。图中显示了用于将个体15月龄时的两种语言和交流能力测量归于下半部或上半部(中值分割)的分类准确度的自举分布(bootstrapped distribution)(10000倍)。对真实的(左侧)和置换的个体标记(右侧)进行分类。置换标记分类准确度的平均值为50％的随机水平。真实标记分类的平均值在两种情况下都高于置换标记分类的95％(水平虚线)，这表明分类显著高于随机水平(p<0.05)。
发明详细描述
如前文所述，对于个体而言，语言和交流能力是非常重要的素质，对学业和事业等多个方面会产生显著影响。汉语和英语是世界上使用者最多的2种语言，但是它们具有显著不同的语言结构。一个显著的区别是在汉语中使用词汇声调或词汇意义层面的音高模式，而英语或很多其他语言使用音高模式表达句意。例如，对于一个音节来说，汉语普通话通常会有平上去入4个声调，而在一些汉语方言中的声调可能会更多(例如粤语可以包括6个声调)。因此，人体的神经系统精确地编码音调的能力对于语言发育是至关重要的。已有研究发现，学习汉语的婴儿的听觉神经系统具有显著的追踪音调的能力。
本申请希望解决的问题之一是在婴幼儿时期对个体语言和交流能力正常性的发展水平差异进行预测，根据得到的预测结果可以进行有针对性的主动干预和训练。针对这样的技术问题，本申请提供了一个技术平台，主要依靠来自婴幼儿期的神经数据(例如EEG或MEG提取数据)来预测以后的语言和交流能力的发展趋势。机器学习技术被用于训练分类器以进行这种预测。为了获得训练分类器所需要的训练数据集，需要获得一定量的婴幼儿样本的神经数据，随后进行语言和交流能力的评定，将两部分结果进行交叉验证和训练。一旦训练完成，机器学习分类器就可以被用于预测其他婴幼儿的语言和交流能力的正常性的发展水平差异。
大体而言，本申请的方法可以包括在婴幼儿头部周围放置表面电极同时进行EEG记录，当婴幼儿在自然睡眠中施加听觉刺激(参见图1)；测量来自皮层下听觉脑干的神经响应，特别是婴幼儿的听觉脑干如何跟随刺激的频率(音调)(示例性结果参见图2)，从而利用在皮层下听觉脑干和连接的神经结构中观察音调跟踪的能力来预测婴幼儿未来的语言和交流能力发育水平。
因此，第一方面，本申请提供了一种预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的方法，所述方法包括：
获得所述健康婴儿或幼儿对外部听觉刺激引发的脑电图(EEG)或脑磁图(MEG)波形数据；
从所述EEG或MEG波形数据中提取定量数据，其中所述定量数据包括表征中枢神经系统对所述外部听觉刺激所引起的测量指标数据；
使用机器学习分类器分析所述定量数据，所述机器学习分类器被训练为能预测所述健康婴儿或幼儿所属人群的语言和交流能力正常性的发展水平差异的预 测评分，其中所述训练是基于从具有已知语言和交流能力正常性的发展水平的多个健康婴儿或幼儿获得的训练数据集的对应定量数据；以及
基于所述机器学习分类器输出的结果，生成所述健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的预测评分。
在一些实施方案中，该方法还包括：获得处于静息状态下的所述健康婴儿或幼儿的脑电图(EEG)或脑磁图(MEG)波形数据；和从处于静息状态的所述健康婴儿或幼儿的EEG或MEG波形中提取相应的定量数据，其中所述定量数据包括指示中枢神经系统处于静息状态下的测量指标数据。
在一些实施方案中，在所述健康婴儿或幼儿处于自然睡眠时或处于清醒状态下执行所述EEG或MEG。在一些实施方案中，本申请的健康婴儿或幼儿的月龄为18个月以下。在一些实施方案中，关于训练数据集的获取，当从多个健康婴儿或幼儿获得所述训练数据集时，多个健康婴儿或幼儿的月龄为18个月以下。
本文中的“健康”指至少在听、说、智力等与语言和交流能力方面相关方面的健康状态，并且包括但不限于完全健康状态。
本文所述的“婴儿”和“幼儿”按照通常的年龄标准，婴儿为0-12月龄，幼儿为1-3岁。
本文中的“外部听觉刺激”是指语言方面的听觉刺激，并且根据不同语言的结构和组成可以相应地选择。应当理解，外部听觉刺激的语言应当与测试个体未来成长环境中接触的语言相匹配。通常，单个外部听觉刺激应当为相对简单的语言单元，例如，字母/拼音、单词/单字(不同声调)、短语等。在一些实施方案中，外部听觉刺激来自使用音高模式来在词汇或句意水平表达含义的语言和语音信号。在一些实施方案中，外部听觉刺激来自汉语，包括普通话和粤语等。在一些实施方案中，外部听觉刺激来自使用音高模式来在短语或句意水平表达含义的语言和语音信号。在一些实施方案中，外部听觉刺激来自汉语，包括普通话和粤语等。虽然本申请的研究针对的示例性语言为汉语(汉语使用音高模式在词汇或句意水平都能表达含义，因而特别适用于本申请)，但是本申请的各项发明的应用并不局限于此，诸如英语、法语、德语、西班牙语、葡萄牙语、日语、韩语等其他语言也同样适用于本申请，因为几乎全部的语言都至少使用音高模式在词汇或句意水平表达含义。
在一些实施方案中，外部听觉刺激包括多个不同的刺激。在一些实施方案中，定量数据包括指示中枢神经系统对所述多个外部听觉刺激中一个或多个的反应与处于静息状态下的测量指标数据。作为非限制性的例子，多个外部听觉刺激可以是一个汉语拼音的多个声调。
在获得脑电图(EEG)或脑磁图(MEG)波形之后，需要从中提取表征中枢神经系统对外部听觉刺激的响应的数据。例如，定量数据可以包括表征听觉中枢相关的处理通路功能活动的数据；或表征下丘及与其连接的中心的功能活动的数据；或表征初级听觉皮层的功能活动的数据，所述初级听觉皮层包括赫氏脑回或与其连接的中心。
从EEG或MEG波形提取上述定量数据的可用方法有多种，包括但不限于，自动峰值检测(Automatic peak detection)，快速傅立叶变换(Fast Fourier Transform)，自相关(Autocorrelation)，均方根(Root-Mean-Square，RMS),Morlet小波变换(Morlet Wavelet Transform)，离散小波变换(Discrete Wavelet Transform)，小波散射(Wavelet Scattering)，刺激响应互相关(Stimulus-Response Cross- correlation)，经验模式分解的处理(Empirical Mode Decomposition)，希尔伯特-黄变换(Hilbert-Huang Transform)，或以上方法的组合。
可从EEG或MEG波形提取的可用定量数据的形式有多种，包括但不限于，时域峰值幅度(time-domain peak amplitude)，时域峰值时延(time-domain peak latency)，基频(fundamental frequency，F0)，谐波(harmonics)，信噪比(signal-to-noise ratio)，RMS幅度(RMS amplitude)，相关系数(correlation coefficient)，试验间相位相干(inter-trial phase coherence)，锁相系数(phase-locking coefficient)，响应一致性(response consistency)，音高强度(pitch strength)，音高误差(pitch error)，音高跟踪精度(pitch-tracking accuracy)，或上述数据的组合。
作为非限制性实例，从EEG波形提取定量数据的过程可以包括：通过刺激起始标记将所述EEG波形分割成节段，并且在具有重叠窗口的滑动时间窗口中，利用快速傅立叶变换(FFT)将每个节段变换到频域(frequency domain)。在一些实施方案中，定量数据包括所述健康婴儿或幼儿的T*(E*3)*F张量表示的矩阵，其中T是所述时间窗口的数量，E是每个刺激(例如，一个汉语拼音的一个声调)的节段的数量，F是来自FFT分析的频率间隔(frequency bin)的数量，并且每个T的(E*3)*F矩阵首先在行内被归一化，然后在列中被归一化，从而消除频谱的绝对幅度的影响并且仅保留频率相关随时间变化的模式。
在获得定量数据之后，使用机器学习分类器分析所述定量数据。机器学习分类器(machine-learning classifier)的大体工作原理是已知的。本申请的机器学习分类器工作所基于的训练数据集来自于与测试个体属于相同人群的多个类似个体。本文中所指的“相同人群”或“相同群体”是指处于相同语言文化层面的个体的集合。例如，如果测试个体会在说普通话的环境中生长，则应当使用基于普通话的训练数据集；类似得，如果测试个体会在说粤语的环境中生长，则应当使用基于粤语的训练数据集；以此类推。用于创建训练数据集的多个个体可以与测试个体进行过相同的外部听觉刺激和EEG/MEG定量数据提取，并且所述多个个体还在随后(例如6-12个月后)经过语言和交流能力的实际评价并将结果用于创建训练数据集。评价儿童语言和交流能力的测试有多种，例如汉语沟通发展评定量表(Chinese Communication Development Inventories)中的后期姿态(Later Gestures)和交互动作(Interactive Acts)。
在一些实施方案中，机器学习分类器是使用包括高斯核、C和伽马的参数的支持向量机(SVM)。
在一些实施方案中，机器学习分类器是支持向量机(SVM)，并且所述预测的语言和交流能力正常性的发展水平差异是高水平或低水平，或连续性评价方式。
在一些实施方案中，机器学习分类器是支持向量回归算法(SVR)或排序支持向量机(RankSVM)，并且所预测的语言和交流能力正常性的发展水平差异是量化的。
在一些实施方案中，由所述机器学习分类器进行的分类是通过交叉验证的，并且交叉验证的结果是在一定倍数(例如5倍、10倍或更多倍)的平均准确度、特异性、灵敏度、AUC、奇偶率、相关系数或以上的组合。
在一些实施方案中，分析所述定量数据还包括分析与语言发育相关的一个或多个变量，例如性别和出生数据。
本申请的实施例描述了一种具体的第一方面的方法，应当理解，实施例中的技术特征，在不存在技术矛盾的情况下，也可以适用于上述实施方案。
第二方面，本申请提供了一种计算机可读介质，其存储多个指令，所述多个指令当被计算机系统的处理器执行时控制计算机系统执行多种操作以实施第一方面的方法。
第三方面，本申请提供了一种计算机系统，包括：
第二方面所述的计算机可读介质；和
用于执行存储在所述计算机可读介质上的指令一个或多个处理器。
第四方面，本申请提供了一种计算机系统，包括：
存储器；和
处理器，所述处理器与所述存储器可通讯地连接并且被配置为执行多种操作以实施第一方面的方法。
第五方面，本申请提供了一种预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的系统，包括用于实施第一方面的方法的多种装置。
第六方面，本申请提供了一种生成用于预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的机器学习分类器的方法，所述方法包括：
获得一名或多名健康婴儿或幼儿对外部听觉刺激引发的脑电图(EEG)或脑磁图(MEG)波形数据，并且从所述EEG或MEG波形数据中提取定量数据，其中所述定量数据包括表征中枢神经系统对所述外部听觉刺激所引起的测量指标数据；
在一定时间(例如6-12个月)后，对所述一名或多名健康婴儿或幼儿进行语言和交流能力测试，并获得表征语言和交流能力的定性或定量数据；
将从所述EEG或MEG波形数据中提取的定量数据和所述表征语言和交流能力的定性或定量数据输入机器学习分类器进行训练，以生成用于预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的机器学习分类器。
在不发生矛盾或进行适应性调整的情况下，第一方面的各个技术方案和技术特征也适用于第二至第六方面。还应当理解，第六方面的发明涉及EEG或MEG波形数据中的定量数据的获取和后期表征语言和交流能力的定性或定量数据的获取，前者的实施可以参照第一方面的方法；后者的实施具有多种方式，例如本文示例性采用的汉语沟通发展评定量表(Chinese Communication Development Inventories)中的多种与语言和沟通能力相关的模块，无论是获得定性还是定量的结果，均可以将其进行适当的数据变换应用于机器学习分类器进行训练。
应理解，本发明的任何实施方式可以以模块化或集成方式使用硬件(例如专用集成电路或现场可编程门阵列)和/或使用具有通用可编程处理器的计算机软件以控制逻辑的形式来实施。如本文所用，处理器包括单核处理器、在同一集成芯片上的多核处理器，或在单个电路板上或网络化的多个处理单元。基于本公开和本文所提供的教示，本领域的普通技术人员将知道并且理解使用硬件和硬件与软件的组合来实施本发明实施方式的其它方式和/或方法。
在本申请中所述的任何软件组件或功能可以实施为以待通过处理器使用任何合适的计算机语言如例如Java、C、C++、C#、面向对象的C语言、Swift或脚本语言(如使用例如常规或面向对象技术的Perl或Python)而执行的软件代码。软件代码可以存储为用于存储和/或传输的计算机可读介质上的一系列指令或命令，合适的介质包括随机存取存储器(RAM)、只读存储器(ROM)、磁性介质(如硬盘驱动器或软性磁盘)或光学介质(如光盘(CD)或DVD(数字通用光盘))、快闪存储器等。计算机可读介质可以是这类存储或传输装置的任何组合。
还可以使用适合于经由符合多种方案的有线、光学和/或无线网络(包括因特网)传输的载波信号来编码和传输这类程序。因此，根据本发明一个实施方式的计算机可读介质可以使用利用这类程序编码的数据信号产生。利用程序代码编码的计算机可读介质可以与兼容装置一起封装或与其它装置分开提供(例如经由因特网下载)。任何这类计算机可读介质可驻存于单个计算机产品(例如硬盘驱动器、CD或整个计算机系统)上或其内部，并且可以存在于系统或网络内的不同计算机产品上或其内部。。计算机系统可包括用于向用户提供本文所提及的任何结果的监视器、打印机或其它合适的显示器。
本文中所述的任何方法可完全或部分用计算机系统执行，所述计算机系统包括一个或多个处理器，所述处理器可以经配置以执行所述步骤。因此，实施方式可以涉及经配置以执行本文所述的任何方法的步骤的计算机系统，可能用不同组件执行相应步骤或相应步骤群组。尽管本文中方法的步骤以经编号步骤的形式呈现，但其可以同时或以不同顺序执行。另外，这些步骤的部分可以与其它方法的其它步骤部分一起使用。此外，步骤的全部或部分可以是任选的。另外，任何方法的任何步骤都可以用执行这些步骤的模块、电路或其它构件来执行。
可在不脱离本发明实施方式的精神和范围的情况下，以任何合适方式组合特定实施方式的具体细节。然而，本发明的其它实施方式可以涉及与每个个别方面或这些个别方面的具体组合相关的具体实施方式。
对本发明的实例性实施方式的以上描述已经为了说明和描述的目的而呈现。其并非旨在是穷尽性的或将本发明限制于所描述的精确形式，并且鉴于以上教示许多修改和变化都是可能的。
实施例1
本研究涉及使用神经学测量来构建预测算法，用于在个体水平进行语言和交流能力正常性的发展水平结果的预测。作为示例，本研究使用EEG数据作为神经学测量指标。本研究主要需要：1)EEG测试流程，并且该流程最好是简短的并且包含最有信息的听觉刺激，2)来自一组已经在这种EEG流程上进行过测试的儿童的现有数据，并且这些儿童随后的语言和交流结果是已知的，以及3)用于将EEG和语言和交流的实际评价结果数据相关联的算法，在进行预测中具有较高的特异性和敏感性。
数据源
本研究基于从30个学习汉语的儿童的数据来构建算法。这些儿童在婴儿期进行EEG神经学测量，并使用汉语沟通发展评定量表(Chinese Communicative Development Inventory,CCDI,Tardif et al.,2008)在EEG测量之后几个月评价语言发展结果。
刺激和EEG程序
听觉诱发的EEG程序在多篇文献已有报道(例如，Lau,J.C.Y.,Wong,P.C.M.,&Chandrasekaran,B.(2017).Context-dependent Plasticity in the Subcortical Encoding of Linguistic Pitch Patterns.Journal of Neurophysiology,117(2),594–603；Liu,F.,Maggu,A.R.,Lau,J.C.Y.,&Wong,P.C.M.(2015).Brainstem Encoding of Speech and Musical Stimuli in Congenital Amusia:Evidence from Cantonese Speakers.Frontiers in Human Neuroscience,8；Maggu,A.R.,Liu,F.,Antoniou,M.,&Wong,P.C.M.(2016).Neural Correlates of Indicators of Sound Change in Cantonese:Evidence  from Cortical and Subcortical Processes.Frontiers in Human Neuroscience,10.，通过引用的方式将上述文献的内容整体并入本文，用于全部目的)(基本设置参见图1)。在本研究中，婴儿接受听觉刺激，本研究中的刺激特别设计为反映汉语中词汇音的特性(图2)。在本研究的外部听觉刺激中，将三种声调嵌入音节/ga/：粤语声调/ga2/,普通话声调/ga3/和粤语声调/ga4/(本研究中的婴儿在粤语家庭中成长)。早期延时EEG反应真实地反映了听觉刺激的时程和谱组成。图A显示听觉刺激(左栏)和平均EEG响应(中间栏)的时间进程以及与包围听觉刺激谱的上四分位数的轮廓重叠的EEG响应的平均谱图(右栏)。EEG响应的谱幅度密度集中在听觉刺激谱密度的上四分位数内。图B显示了在1、3、6、9和15月龄的婴儿的听觉刺激的谱图(左栏)和其对应的平均EEG谱图(其余五栏)。婴儿月龄越大，听觉谱结构和EEG谱结构之间的重叠越明显。收集的听觉诱发的神经EEG响应根据Matlab中执行的一系列自动过程进行分析，例如快速傅立叶变换，自相关和均方根(RMS)。按照报道的程序提取诸如基频(F0)、谐波、信噪比和RMS幅度的测量值。
其他变量
除了神经学测量差异之外，其它变量也可能会影响语言发育(例如，性别，出生数据等)。因此，这些变量也可以输入到预测模型中以实现更优的预测性能。
机器学习
本研究的算法使用神经EEG测量，用于在婴儿期间进行EEG记录之后几个月，将个体婴儿分类为较好或较差的语言结果组。本研究使用具有10分交叉验证程序和自举-置换统计的支持向量机(SVM)程序来实现二项分类。
本研究执行了两步法SVM分类：在个体受试者水平和在88个受试者的组水平，其中30名受试者约6月龄时进行了EEG数据收集，并在约1周岁时进行语言和交流能力(CCDI)评估和数据收集。本研究用EEG数据预测6个月后的语言和交流能力发展水平。
在个体受试者的水平上，通过刺激起始标记将经滤波的EEG分割成节段。三个声调中的每一个得到一组1000个EEG节段。在50ms的滑动窗口中利用FFT将每个节段变换到频域，其中窗口之间有50％的重叠。得到每个受试者的T*(E*3)*F张量的矩阵，其中T是时间窗口的数目，E是每个声调的节段的数目，F是来自FFT分析的频率间隔(frequency bin)的数目。每个T的(E*3)*F矩阵首先在行内被归一化，然后在列中被归一化，从而消除谱的绝对幅度的影响，并且仅留下频率相关的模式。然后将T矩阵输入在LIBSVM软件包中执行的三向SVM分类器中作为声调类别的预测器。使用高斯核、C＝100、γ＝1/F。以10倍的方式交叉验证分类，即，数据被随机地分成10份，并且对于每一份，分类器在90％的数据上被训练并且在10％的数据上被测试。交叉验证过程的结果是跨越倍数的平均精确度。结果，对于每个受试者，获得了T分类精度的矢量。图3显示了通过将支持向量机(SVM)分类器应用于对三个汉语词汇音/ga2/、/ga3/和/ga4/的EEG响应的预测分类结果，可以反映婴儿大脑的辨别能力。向SVM分类器输入沿着对三个声调的EEG响应的时间过程的50ms间隔单元的傅立叶变换。对于随机分割数据的10份中的每一份，通过对9/10的数据进行训练和对其余1/10的数据进行测试，对分类进行交叉验证。对于3个声调的分类精度的机会水平是33％。该图说明分类准确度沿着反应的时间进程具有两个峰值，并且准确度随着婴儿的月龄增大而增长。该图说明了跨越个体的分类。在每个单独个体中进 行相同的操作。
在组水平，这些矢量被组合成S*T矩阵，其中S是具有来自EEG记录和CCDI语言和交流结果的受试者的数目。这些矩阵是为EEG记录时间和语言和交流结果产生时间的每个可能的组合创建的，并且在每个时间区间内被归一化。这些组合覆盖了原始83个受试者的部分重叠的子集。这些年龄组合的每一个内的语言和交流结果被转换成具有中值分裂的二进制变量。每个预测器矩阵在每个时间区间内被归一化。然后对每个语言和交流结果以及语言和交流结果时间和EEG记录时间的每个可能的组合执行二进制SVM。SVM参数为高斯核，C＝100,γ＝1/T。用10分法交叉验证分类的准确性。使用自举和置换的组合来计算每个分类的置信区间。对于每个预测器-响应组合，用置换对矩阵(受试者)的行进行重新采样，对实际的和随机置换的受试者标记进行10分交叉验证的SVM分类。重复该过程10000次，获得实际预测精度和置换预测精度的分布(参见图4)。真实标签精确度的平均值高于置换标签精确度的分布的95％，认为精确度显著高于随机水平。
实施例2
在实施例1的基础上，本申请的发明人将研究群体扩大至118名儿童(包括实施例1中的儿童群体，保持实验方案不变)。扩大样本研究的结果表明，本申请构建的预测模型能够将儿童分类为低于或高于平均水平，或者低于或高于25％百分比，并实现高达0.92精准度(AUC)。

Claims (78)

  1. 一种预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的方法，所述方法包括：
    获得所述健康婴儿或幼儿对外部听觉刺激引发的脑电图(EEG)或脑磁图(MEG)波形数据；
    从所述EEG或MEG波形数据中提取定量数据，其中所述定量数据包括表征中枢神经系统对所述外部听觉刺激所引起的测量指标数据；
    使用机器学习分类器分析所述定量数据，所述机器学习分类器被训练为能预测所述健康婴儿或幼儿所属人群的语言和交流能力正常性的发展水平差异的预测评分，其中所述训练是基于从具有已知语言和交流能力正常性的发展水平的多个健康婴儿或幼儿获得的训练数据集的对应定量数据；以及
    基于所述机器学习分类器输出的结果，生成所述健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的预测评分。
  2. 根据权利要求1所述的方法，还包括：
    获得处于静息状态下的所述健康婴儿或幼儿的脑电图(EEG)或脑磁图(MEG)波形数据；和
    从处于静息状态的所述健康婴儿或幼儿的EEG或MEG波形中提取相应的定量数据，其中所述定量数据包括指示中枢神经系统处于静息状态下的测量指标数据。
  3. 根据权利要求2所述的方法，其中所述外部听觉刺激包括多个不同的刺激，并且所述定量数据包括指示中枢神经系统对所述多个外部听觉刺激中一个或多个的反应与处于静息状态下的测量指标数据。
  4. 根据权利要求1至3中任一项所述的方法，其中所述定量数据包括表征听觉中枢相关的处理通路功能活动的数据。
  5. 根据权利要求1至4中任一项所述的方法，其中所述定量数据包括表征下丘及与其连接的中心的功能活动的数据。
  6. 根据权利要求1至5中任一项所述的方法，其中所述定量数据包括表征初级听觉皮层的功能活动的数据，所述初级听觉皮层包括赫氏脑回或与其连接的中心。
  7. 根据权利要求1至6中任一项所述的方法，其中所述外部听觉刺激来自使用音高模式来在词汇或句意水平表达含义的语言和语音信号，优选地，所述外部听觉刺激来自汉语。
  8. 根据权利要求1至6中任一项所述的方法，其中所述外部听觉刺激来自使用音高模式来在短语或句子水平表达含义的语言和语音信号，优选地，所述外部听觉刺激来自汉语。
  9. 根据权利要求1至8中任一项所述的方法，其中所述外部听觉刺激是一个或多个声调的汉语拼音。
  10. 根据权利要求1至9中任一项所述的方法，其中在所述健康婴儿或幼儿处于自然睡眠时或处于清醒状态下执行所述EEG或MEG。
  11. 根据权利要求1至10中任一项所述的方法，其中所述健康婴儿或幼儿的月龄为18个月以下，并且其中当从所述多个健康婴儿或幼儿获得所述训练数据集时，所述多个健康婴儿或幼儿的月龄为18个月以下。
  12. 根据权利要求1至11中任一项所述的方法，其中所述机器学习分类器是支持向量机(SVM)，并且所述预测的语言和交流能力的正常性的发展水平差异是高水平或低水平，或连续性评价方式。
  13. 根据权利要求1至11中任一项所述的方法，其中所述机器学习分类器是支持向量回归算法(SVR)或排序支持向量机(RankSVM)，并且所预测的语言和交流能力正常性的发展水平差异是量化的。
  14. 根据权利要求1至13中任一项所述的方法，其中分析所述定量数据还包括分析与语言和交流能力正常性的发展水平差异相关的一个或多个变量，例如性别和出生数据。
  15. 根据权利要求1至14中任一项所述的方法，其中通过选自以下的方法从所述EEG或MEG波形提取定量数据：自动峰值检测(Automatic peak detection)，快速傅立叶变换(Fast Fourier Transform)，自相关(Autocorrelation)，均方根(Root-Mean-Square，RMS),Morlet小波变换(Morlet Wavelet Transform)，离散小波变换(Discrete Wavelet Transform)，小波散射(Wavelet Scattering)，刺激响应互相关(Stimulus-Response Cross-correlation)，经验模式分解的处理(Empirical Mode Decomposition)，希尔伯特-黄变换(Hilbert-Huang Transform)；和/或
    其中所述定量数据包括以下中的一个或多个：时域峰值幅度(time-domain peak amplitude)，时域峰值时延(time-domain peak latency)，基频(fundamental frequency，F0)，谐波(harmonics)，信噪比(signal-to-noise ratio)，RMS幅度(RMS amplitude)，相关系数(correlation coefficient)，试验间相位相干(inter-trial phase coherence)，锁相系数(phase-locking coefficient)，响应一致性(response consistency)，音高强度(pitch strength)，音高误差(pitch error)，音高跟踪精度(pitch-tracking accuracy)。
  16. 根据权利要求1至15中任一项所述的方法，其中从所述EEG波形提取定量数据包括：通过刺激起始标记将所述EEG波形分割成节段，并且在具有重叠窗口的滑动时间窗口中，利用快速傅立叶变换(FFT)将每个节段变换到频域(frequency domain)。
  17. 如权利要求16所述的方法，其中所述定量数据包括所述健康婴儿或幼儿的T*(E*3)*F张量表示的矩阵，其中T是所述时间窗口的数量，E是每个刺激(例如，一个汉语拼音的一个声调)的节段的数量，F是来自FFT分析的频率间隔(frequency bin)的数量，并且每个T的(E*3)*F矩阵首先在行内被归一化，然后在列中被归一化，从而消除频谱的绝对幅度的影响并且仅保留频率相关随时间变化的模式。
  18. 如权利要求17所述的方法，其中所述机器学习分类器是使用包括高斯核、C和伽马的参数的支持向量机(SVM)。
  19. 根据权利要求1至18中任一项所述的方法，其中由所述机器学习分类器进行的分类是通过交叉验证的，并且交叉验证的结果是在一定倍数的平均准确度、特异性、灵敏度、AUC、奇偶率、相关系数或以上的组合。
  20. 一种计算机可读介质，其存储多个指令，所述多个指令当被计算机系统的处理器执行时控制计算机系统执行包括以下操作：
    获得所述健康婴儿或幼儿对外部听觉刺激引发的脑电图(EEG)或脑磁图(MEG)波形数据；
    从所述EEG或MEG波形数据中提取定量数据，其中所述定量数据包括表征中枢神经系统对所述外部听觉刺激所引起的测量指标数据；
    使用机器学习分类器分析所述定量数据，所述机器学习分类器被训练为能预测所述健康婴儿或幼儿所属人群的语言和交流能力正常性的发展水平差异的预测评分，其中所述训练是基于从具有已知语言和交流能力正常性的发展水平的多个健康婴儿或幼儿获得的训练数据集的对应定量数据；以及
    基于所述机器学习分类器输出的结果，生成所述健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的预测评分。
  21. 根据权利要求20所述的计算机可读介质，其中所述操作还包括：
    获得处于静息状态下的所述健康婴儿或幼儿的脑电图(EEG)或脑磁图(MEG)波形数据；和
    从处于静息状态的所述健康婴儿或幼儿的EEG或MEG波形中提取相应的定量数据，其中所述定量数据包括指示中枢神经系统处于静息状态下的测量指标数据。
  22. 根据权利要求20或21所述的计算机可读介质，其中所述外部听觉刺激包括多个不同的刺激，并且所述定量数据包括指示中枢神经系统对所述多个外部听觉刺激中一个或多个的反应与处于静息状态下的测量指标数据。
  23. 根据权利要求20至22中任一项所述的计算机可读介质，其中所述定量数据包括表征听觉中枢相关的处理通路功能活动的数据。
  24. 根据权利要求20至23中任一项所述的计算机可读介质，其中所述定量数据包括表征下丘及与其连接的中心的功能活动的数据。
  25. 根据权利要求20至24中任一项所述的计算机可读介质，其中所述定量数据包括表征初级听觉皮层的功能活动的数据，所述初级听觉皮层包括赫氏脑回或与其连接的中心。
  26. 根据权利要求20至25中任一项所述的计算机可读介质，其中所述外部听觉刺激来自使用音高模式来在词汇或句意水平表达含义的语言和语音信号，优选地，所述外部听觉刺激来自汉语。
  27. 根据权利要求20至25中任一项所述的计算机可读介质，其中所述外部听觉刺激来自使用音高模式来在短语或句子水平表达含义的语言和语音信号，优选地，所述外部听觉刺激来自汉语。
  28. 根据权利要求20至27中任一项所述的计算机可读介质，其中所述外部听觉刺激是一个或多个声调的汉语拼音。
  29. 根据权利要求20至28中任一项所述的计算机可读介质，其中在所述健康婴儿或幼儿处于自然睡眠时或处于清醒状态下执行所述EEG或MEG。
  30. 根据权利要求20至29中任一项所述的计算机可读介质，其中所述健康婴儿或幼儿的月龄为18个月以下，并且其中当从所述多个健康婴儿或幼儿获得所述训练数据集时，所述多个健康婴儿或幼儿的月龄为18个月以下。
  31. 根据权利要求20至30中任一项所述的计算机可读介质，其中所述机器学习分类器是支持向量机(SVM)，并且所述预测的语言和交流能力的正常性的发展水平差异是高水平或低水平，或连续性评价方式。
  32. 根据权利要求20至30中任一项所述的计算机可读介质，其中所述机器学习分类器是支持向量回归算法(SVR)或排序支持向量机(RankSVM)，并且所预测的语言和交流能力正常性的发展水平差异是量化的。
  33. 根据权利要求20至32中任一项所述的计算机可读介质，其中分析所述定量数据还包括分析与语言和交流能力正常性的发展水平差异相关的一个或多个变量，例如性别和出生数据。
  34. 根据权利要求20至33中任一项所述的计算机可读介质，其中通过选自以下的方法从所述EEG或MEG波形提取定量数据：自动峰值检测(Automatic peak detection)，快速傅立叶变换(Fast Fourier Transform)，自相关(Autocorrelation)，均方根(Root-Mean-Square，RMS),Morlet小波变换(Morlet Wavelet Transform)，离散小波变换(Discrete Wavelet Transform)，小波散射(Wavelet Scattering)，刺激响应互相关(Stimulus-Response Cross-correlation)，经验模式分解的处理(Empirical Mode Decomposition)，希尔伯特-黄变换(Hilbert-Huang Transform)；和/或
    其中所述定量数据包括以下中的一个或多个：时域峰值幅度(time-domain peak amplitude)，时域峰值时延(time-domain peak latency)，基频(fundamental frequency，F0)，谐波(harmonics)，信噪比(signal-to-noise ratio)，RMS幅度(RMS amplitude)，相关系数(correlation coefficient)，试验间相位相干(inter-trial phase coherence)，锁相系数(phase-locking coefficient)，响应一致性(response consistency)，音高强度(pitch strength)，音高误差(pitch error)，音高跟踪精度(pitch-tracking accuracy)。
  35. 根据权利要求20至34中任一项所述的计算机可读介质，其中从所述EEG波形提取定量数据包括：通过刺激起始标记将所述EEG波形分割成节段，并且在具有重叠窗口的滑动时间窗口中，利用快速傅立叶变换(FFT)将每个节段变换到频域(frequency domain)。
  36. 如权利要求35所述的计算机可读介质，其中所述定量数据包括所述健康婴儿或幼儿的T*(E*3)*F张量表示的矩阵，其中T是所述时间窗口的数量，E是每个刺激(例如，一个汉语拼音的一个声调)的节段的数量，F是来自FFT分析的频率间隔(frequency bin)的数量，并且每个T的(E*3)*F矩阵首先在行内被归一化，然后在列中被归一化，从而消除频谱的绝对幅度的影响并且仅保留频率相关随时间变化的模式。
  37. 如权利要求36所述的计算机可读介质，其中所述机器学习分类器是使用包括高斯核、C和伽马的参数的支持向量机(SVM)。
  38. 根据权利要求20至37中任一项所述的计算机可读介质，其中由所述机器学习分类器进行的分类是通过交叉验证的，并且交叉验证的结果是在一定倍数的平均准确度、特异性、灵敏度、AUC、奇偶率、相关系数或以上的组合。
  39. 一种计算机系统，包括：
    如权利要求20-38中任一项所述的计算机可读介质；和
    用于执行存储在所述计算机可读介质上的指令一个或多个处理器。
  40. 一种计算机系统，包括：
    存储器；和
    处理器，所述处理器与所述存储器可通讯地连接并且被配置为执行以下操作：
    获得所述健康婴儿或幼儿对外部听觉刺激引发的脑电图(EEG)或脑磁图(MEG)波形数据；
    从所述EEG或MEG波形数据中提取定量数据，其中所述定量数据包括表征中枢神经系统对所述外部听觉刺激所引起的测量指标数据；
    使用机器学习分类器分析所述定量数据，所述机器学习分类器被训练为能预测所述健康婴儿或幼儿所属人群的语言和交流能力正常性的发展水平差异的预测评分，其中所述训练是基于从具有已知语言和交流能力正常性的发展水平的多个健康婴儿或幼儿获得的训练数据集的对应定量数据；以及
    基于所述机器学习分类器输出的结果，生成所述健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的预测评分。
  41. 根据权利要求40所述的计算机系统，所述处理器还被配置为执行以下操作：
    获得处于静息状态下的所述健康婴儿或幼儿的脑电图(EEG)或脑磁图(MEG)波形数据；和
    从处于静息状态的所述健康婴儿或幼儿的EEG或MEG波形中提取相应的定量数据，其中所述定量数据包括指示中枢神经系统处于静息状态下的测量指标数据。
  42. 根据权利要求40或41所述的计算机系统，其中所述外部听觉刺激包括多个不同的刺激，并且所述定量数据包括指示中枢神经系统对所述多个外部听觉刺激中一个或多个的反应与处于静息状态下的测量指标数据。
  43. 根据权利要求40至42中任一项所述的计算机系统，其中所述定量数据包括表征听觉中枢相关的处理通路功能活动的数据。
  44. 根据权利要求40至43中任一项所述的计算机系统，其中所述定量数据包括表征下丘及与其连接的中心的功能活动的数据。
  45. 根据权利要求40至44中任一项所述的计算机系统，其中所述定量数据包括表征初级听觉皮层的功能活动的数据，所述初级听觉皮层包括赫氏脑回或与其连接的中心。
  46. 根据权利要求40至45中任一项所述的计算机系统，其中所述外部听觉刺激来自使用音高模式来在词汇或句意水平表达含义的语言和语音信号，优选地，所述外部听觉刺激来自汉语。
  47. 根据权利要求40至45中任一项所述的计算机系统，其中所述外部听觉刺激来自使用音高模式来在短语或句子水平表达含义的语言和语音信号，优选地，所述外部听觉刺激来自汉语。
  48. 根据权利要求40至47中任一项所述的计算机系统，其中所述外部听觉刺激是一个或多个声调的汉语拼音。
  49. 根据权利要求40至48中任一项所述的计算机系统，其中在所述健康婴儿或幼儿处于自然睡眠时或处于清醒状态下执行所述EEG或MEG。
  50. 根据权利要求40至49中任一项所述的计算机系统，其中所述健康婴儿或幼儿的月龄为18个月以下，并且其中当从所述多个健康婴儿或幼儿获得所述训练数据集时，所述多个健康婴儿或幼儿的月龄为18个月以下。
  51. 根据权利要求40至50中任一项所述的计算机系统，其中所述机器学习分类器是支持向量机(SVM)，并且所述预测的语言和交流能力的正常性的发展水平差异是高水平或低水平，或连续性评价方式。
  52. 根据权利要求40至51中任一项所述的计算机系统，其中所述机器学习分类器是支持向量回归算法(SVR)或排序支持向量机(RankSVM)，并且所预测的语言和交流能力正常性的发展水平差异是量化的。
  53. 根据权利要求40至52中任一项所述的计算机系统，其中分析所述定量数据还包括分析与语言和交流能力正常性的发展水平差异相关的一个或多个变量，例如性别和出生数据。
  54. 根据权利要求40至53中任一项所述的计算机系统，其中通过选自以下的方法从所述EEG或MEG波形提取定量数据：自动峰值检测(Automatic peak detection)，快速傅立叶变换(Fast Fourier Transform)，自相关(Autocorrelation)，均方根(Root-Mean-Square，RMS),Morlet小波变换(Morlet Wavelet Transform)，离散小波变换(Discrete Wavelet Transform)，小波散射(Wavelet Scattering)，刺激响应互相关(Stimulus-Response Cross-correlation)，经验模式分解的处理(Empirical Mode Decomposition)，希尔伯特-黄变换(Hilbert-Huang Transform)；和/或
    其中所述定量数据包括以下中的一个或多个：时域峰值幅度(time-domain peak amplitude)，时域峰值时延(time-domain peak latency)，基频(fundamental frequency，F0)，谐波(harmonics)，信噪比(signal-to-noise ratio)，RMS幅度(RMS amplitude)，相关系数(correlation coefficient)，试验间相位相干(inter-trial phase coherence)，锁相系数(phase-locking coefficient)，响应一致性(response consistency)，音高强度(pitch strength)，音高误差(pitch error)，音高跟踪精度(pitch-tracking accuracy)。
  55. 根据权利要求40至54中任一项所述的计算机系统，其中从所述EEG波形提取定量数据包括：通过刺激起始标记将所述EEG波形分割成节段，并且在具有重叠窗口的滑动时间窗口中，利用快速傅立叶变换(FFT)将每个节段变换到频域(frequency domain)。
  56. 如权利要求55所述的计算机系统，其中所述定量数据包括所述健康婴儿或幼儿的T*(E*3)*F张量表示的矩阵，其中T是所述时间窗口的数量，E是每个刺激(例如，一个汉语拼音的一个声调)的节段的数量，F是来自FFT分析的频率间隔(frequency bin)的数量，并且每个T的(E*3)*F矩阵首先在行内被归一化，然后在列中被归一化，从而消除频谱的绝对幅度的影响并且仅保留频率相关随时间变化的模式。
  57. 如权利要求56所述的计算机系统，其中所述机器学习分类器是使用包括高斯核、C和伽马的参数的支持向量机(SVM)。
  58. 根据权利要求40至57中任一项所述的计算机系统，其中由所述机器学习分类器进行的分类是通过交叉验证的，并且交叉验证的结果是在一定倍数的平均准确度、特异性、灵敏度、AUC、奇偶率、相关系数或以上的组合。
  59. 一种预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的系统，包括：
    用于获得所述健康婴儿或幼儿对外部听觉刺激引发的脑电图(EEG)或脑磁图(MEG)波形数据的装置；
    用于从所述EEG或MEG波形数据中提取定量数据的装置，其中所述定量数据包括表征中枢神经系统对所述外部听觉刺激所引起的测量指标数据；
    用于使用机器学习分类器分析所述定量数据的装置，所述机器学习分类器被训练为能预测所述健康婴儿或幼儿所属人群的语言和交流能力正常性的发展水平差异的预测评分，其中所述训练是基于从具有已知语言和交流能力正常性的发展水平的多个健康婴儿或幼儿获得的训练数据集的对应定量数据；以及
    用于基于所述机器学习分类器输出的结果，生成所述健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的预测评分的装置。
  60. 根据权利要求59所述的系统，还包括：
    用于获得处于静息状态下的所述健康婴儿或幼儿的脑电图(EEG)或脑磁图(MEG)波形数据的装置；和
    用于从处于静息状态的所述健康婴儿或幼儿的EEG或MEG波形中提取相应的定量数据的装置，其中所述定量数据包括指示中枢神经系统处于静息状态下的测量指标数据。
  61. 根据权利要求59或60所述的系统，其中所述外部听觉刺激包括多个不同的刺激，并且所述定量数据包括指示中枢神经系统对所述多个外部听觉刺激中一个或多个的反应与处于静息状态下的测量指标数据。
  62. 根据权利要求59至61中任一项所述的系统，其中所述定量数据包括表征听觉中枢相关的处理通路功能活动的数据。
  63. 根据权利要求59至62中任一项所述的系统，其中所述定量数据包括表征下丘及与其连接的中心的功能活动的数据。
  64. 根据权利要求59至63中任一项所述的系统，其中所述定量数据包括表征初级听觉皮层的功能活动的数据，所述初级听觉皮层包括赫氏脑回或与其连接的脑区。
  65. 根据权利要求59至64中任一项所述的系统，其中所述外部听觉刺激来自使用音高模式来在词汇或句意水平表达含义的语言和语音信号，优选地，所述外部听觉刺激来自汉语。
  66. 根据权利要求59至64中任一项所述的系统，其中所述外部听觉刺激来自使用音高模式来在短语或句子水平表达含义的语言和语音信号，优选地，所述外部听觉刺激来自汉语。
  67. 根据权利要求59至66中任一项所述的系统，其中所述外部听觉刺激是一个或多个声调的汉语拼音。
  68. 根据权利要求59至67中任一项所述的系统，其中在所述健康婴儿或幼儿处于自然睡眠时或处于清醒状态下执行所述EEG或MEG。
  69. 根据权利要求59至68中任一项所述的系统，其中所述健康婴儿或幼儿的月龄为18个月以下，并且其中当从所述多个健康婴儿或幼儿获得所述训练数据集时，所述多个健康婴儿或幼儿的月龄为18个月以下。
  70. 根据权利要求59至69中任一项所述的系统，其中所述机器学习分类器是支持向量机(SVM)，并且所述预测的语言和交流能力的正常性的发展水平差异是高水平或低水平，或连续性评价方式。
  71. 根据权利要求59至70中任一项所述的系统，其中所述机器学习分类器是支持向量回归算法(SVR)或排序支持向量机(RankSVM)，并且所预测的语言和交流能力正常性的发展水平差异是量化的。
  72. 根据权利要求59至71中任一项所述的系统，其中分析所述定量数据还包括分析与语言和交流能力正常性的发展水平差异相关的一个或多个变量，例如性别和出生数据。
  73. 根据权利要求59至72中任一项所述的系统，其中通过选自以下的方法从所述EEG或MEG波形提取定量数据：自动峰值检测(Automatic peak detection)，快速傅立叶变换(Fast Fourier Transform)，自相关(Autocorrelation)，均方根(Root-Mean-Square，RMS),Morlet小波变换(Morlet Wavelet Transform)，离散 小波变换(Discrete Wavelet Transform)，小波散射(Wavelet Scattering)，刺激响应互相关(Stimulus-Response Cross-correlation)，经验模式分解的处理(Empirical Mode Decomposition)，希尔伯特-黄变换(Hilbert-Huang Transform)；和/或
    其中所述定量数据包括以下中的一个或多个：时域峰值幅度(time-domain peak amplitude)，时域峰值时延(time-domain peak latency)，基频(fundamental frequency，F0)，谐波(harmonics)，信噪比(signal-to-noise ratio)，RMS幅度(RMS amplitude)，相关系数(correlation coefficient)，试验间相位相干(inter-trial phase coherence)，锁相系数(phase-locking coefficient)，响应一致性(response consistency)，音高强度(pitch strength)，音高误差(pitch error)，音高跟踪精度(pitch-tracking accuracy)。
  74. 根据权利要求59至73中任一项所述的系统，其中从所述EEG波形提取定量数据包括：通过刺激起始标记将所述EEG波形分割成节段，并且在具有重叠窗口的滑动时间窗口中，利用快速傅立叶变换(FFT)将每个节段变换到频域(frequency domain)。
  75. 如权利要求74所述的系统，其中所述定量数据包括所述健康婴儿或幼儿的T*(E*3)*F张量表示的矩阵，其中T是所述时间窗口的数量，E是每个刺激(例如，一个汉语拼音的一个声调)的节段的数量，F是来自FFT分析的频率间隔(frequency bin)的数量，并且每个T的(E*3)*F矩阵首先在行内被归一化，然后在列中被归一化，从而消除频谱的绝对幅度的影响并且仅保留频率相关随时间变化的模式。
  76. 如权利要求75所述的系统，其中所述机器学习分类器是使用包括高斯核、C和伽马的参数的支持向量机(SVM)。
  77. 根据权利要求59至76中任一项所述的系统，其中由所述机器学习分类器进行的分类是通过交叉验证的，并且交叉验证的结果是在一定倍数的平均准确度、特异性、灵敏度、AUC、奇偶率、相关系数或以上的组合。
  78. 一种生成用于预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的机器学习分类器的方法，所述方法包括：
    获得一名或多名健康婴儿或幼儿对外部听觉刺激引发的脑电图(EEG)或脑磁图(MEG)波形数据，并且从所述EEG或MEG波形数据中提取定量数据，其中所述定量数据包括表征中枢神经系统对所述外部听觉刺激所引起的测量指标数据；
    在一定时间(例如6-12个月)后，对所述一名或多名健康婴儿或幼儿进行语言和交流能力测试，并获得表征语言和交流能力的定性或定量数据；
    将从所述EEG或MEG波形数据中提取的定量数据和所述表征语言和交流能力的定性或定量数据输入机器学习分类器进行训练，以生成用于预测健康婴儿或幼儿的语言和交流能力正常性的发展水平差异的机器学习分类器。
PCT/CN2021/106684 2020-07-17 2021-07-16 预测婴幼儿的语言发展和交流能力的方法和工具 Ceased WO2022012653A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
US18/097,578 US20250268510A1 (en) 2020-07-17 2021-07-16 Method and tool for predicting language development and communication capabilities of infant and toddler
AU2021308673A AU2021308673A1 (en) 2020-07-17 2021-07-16 Method and tool for predicting language development and communication capabilities of infant and toddler
GB2302076.1A GB2612256B (en) 2020-07-17 2021-07-16 Method and tool for forecasting language development and communication capabilities of infant and toddler

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010692253.5A CN113951895B (zh) 2020-07-17 2020-07-17 预测婴幼儿的发育语言和交流能力的方法和工具
CN202010692253.5 2020-07-17

Publications (1)

Publication Number Publication Date
WO2022012653A1 true WO2022012653A1 (zh) 2022-01-20

Family

ID=79459348

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/106684 Ceased WO2022012653A1 (zh) 2020-07-17 2021-07-16 预测婴幼儿的语言发展和交流能力的方法和工具

Country Status (5)

Country Link
US (1) US20250268510A1 (zh)
CN (2) CN116211321A (zh)
AU (1) AU2021308673A1 (zh)
GB (1) GB2612256B (zh)
WO (1) WO2022012653A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116269447A (zh) * 2023-05-17 2023-06-23 之江实验室 一种基于语音调制和脑电信号的言语认知评估系统
CN118216934A (zh) * 2024-05-22 2024-06-21 中国石油大学(华东) 小波散射变换与微分熵特征融合的脑电信号分类方法

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2025092724A1 (en) * 2023-10-30 2025-05-08 The Chinese University Of Hong Kong Forecasting of future autism diagnosis and response to intervention using neural data and machine learning
EP4702925A1 (en) * 2024-09-02 2026-03-04 Universitat de Barcelona A computer implemented method and system for predicting a risk that an infant presents a neural speech encoding disorder
CN120837100B (zh) * 2025-09-23 2025-12-02 安徽大学 一种静息态常模脑电质量评估方法及其系统

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130178731A1 (en) * 2010-08-13 2013-07-11 Children's Medical Center Corporation Methods and apparatus for risk assessment of developmental disorders during early cognitive development
CN104757968A (zh) * 2014-08-08 2015-07-08 沈阳工业大学 儿童失神性癫痫发作情况的中间数据统计评价方法
US20190192285A1 (en) * 2017-12-21 2019-06-27 The Chinese University Of Hong Kong Neural predictors of language-skill outcomes in cochlear implantation patients
CN110582228A (zh) * 2017-03-22 2019-12-17 皇家飞利浦有限公司 用于确定婴儿的健康状态的方法和装置
CN110604565A (zh) * 2019-08-02 2019-12-24 北京脑陆科技有限公司 一种基于便携式eeg设备的脑部健康筛查方法
CN110650681A (zh) * 2017-05-15 2020-01-03 布鲁姆技术公司 用于监测胎儿健康的系统和方法

Family Cites Families (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6993381B2 (en) * 2002-10-25 2006-01-31 Connolly John F Linking neurophysiological and neuropsychological measures for cognitive function assessment in a patient
KR101484930B1 (ko) * 2006-10-13 2015-01-21 러트거스, 더 스테이트 유니버시티 청각 처리에서 결함들을 나타내는 유아들의 검사 및 치료 방법들 및 시스템들
AU2011269693A1 (en) * 2010-06-22 2013-01-24 National Research Council Of Canada Cognitive function assessment in a patient
KR101186298B1 (ko) * 2011-04-26 2012-09-26 광주대학교산학협력단 영유아 대상의 음성자극 분할제시 및 청취능력 평가시스템
SG11201406192VA (en) * 2012-03-30 2014-10-30 Agency Science Tech & Res Method for assessing the treatment of attention-deficit/hyperactivity disorder
CN103054587B (zh) * 2012-12-20 2014-12-31 清华大学 基于听觉稳态响应的言语测听方法
UA83076U (ru) * 2013-03-04 2013-08-27 Таврийский Национальный Университет Имени В.И. Вернадського Способ определения уровня развития речи у детей в возрасте от трех с половиной лет
RU2646745C2 (ru) * 2015-11-26 2018-03-06 Ксения Викторовна Носова Способ оценки уровня развития детей с интеллектуальными и сенсорными нарушениями
EP3267698B1 (en) * 2016-07-08 2024-10-09 Oticon A/s A hearing assistance system comprising an eeg-recording and analysis system
KR102028797B1 (ko) * 2017-03-02 2019-10-04 이화여자대학교 산학협력단 퇴행성 질환의 조기 발견을 위한 언어능력진단 시스템, 방법 및 이를 제공하기 위한 컴퓨터 프로그램이 기록된 컴퓨터로 판독가능한 기록매체
CN107307865A (zh) * 2017-08-04 2017-11-03 湖州健凯康复产品有限公司 一种自闭症儿童辅助交流装置
CN107280666A (zh) * 2017-08-08 2017-10-24 中山大学孙逸仙纪念医院 一种基于机器学习的耳聋患者ci术后康复预测方法及系统
WO2020049430A1 (en) * 2018-09-04 2020-03-12 Johnson & Johnson Consumer Inc. Apparatus and method for assessing emotion of infants and young children
WO2020058761A1 (en) * 2018-09-18 2020-03-26 NeuroSteer Ltd. Systems and methods for cooperative invasive and noninvasive brain stimulation

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130178731A1 (en) * 2010-08-13 2013-07-11 Children's Medical Center Corporation Methods and apparatus for risk assessment of developmental disorders during early cognitive development
CN104757968A (zh) * 2014-08-08 2015-07-08 沈阳工业大学 儿童失神性癫痫发作情况的中间数据统计评价方法
CN110582228A (zh) * 2017-03-22 2019-12-17 皇家飞利浦有限公司 用于确定婴儿的健康状态的方法和装置
CN110650681A (zh) * 2017-05-15 2020-01-03 布鲁姆技术公司 用于监测胎儿健康的系统和方法
US20190192285A1 (en) * 2017-12-21 2019-06-27 The Chinese University Of Hong Kong Neural predictors of language-skill outcomes in cochlear implantation patients
CN110604565A (zh) * 2019-08-02 2019-12-24 北京脑陆科技有限公司 一种基于便携式eeg设备的脑部健康筛查方法

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116269447A (zh) * 2023-05-17 2023-06-23 之江实验室 一种基于语音调制和脑电信号的言语认知评估系统
CN116269447B (zh) * 2023-05-17 2023-08-29 之江实验室 一种基于语音调制和脑电信号的言语认知评估系统
CN118216934A (zh) * 2024-05-22 2024-06-21 中国石油大学(华东) 小波散射变换与微分熵特征融合的脑电信号分类方法

Also Published As

Publication number Publication date
US20250268510A1 (en) 2025-08-28
GB2612256B (en) 2023-12-20
GB202302076D0 (en) 2023-03-29
CN113951895B (zh) 2023-04-25
GB2612256A (en) 2023-04-26
CN113951895A (zh) 2022-01-21
AU2021308673A1 (en) 2023-03-02
CN116211321A (zh) 2023-06-06

Similar Documents

Publication Publication Date Title
CN113951895B (zh) 预测婴幼儿的发育语言和交流能力的方法和工具
Amorim et al. Changes in vocal emotion recognition across the life span.
Winkler Interpreting the mismatch negativity
Volkmer et al. Cortical responses to tone and phoneme mismatch as a predictor of dyslexia? A systematic review
Zhang et al. Neural coding of formant‐exaggerated speech in the infant brain
Correia et al. EEG decoding of spoken words in bilingual listeners: from words to language invariant semantic-conceptual representations
François et al. Enhanced neonatal brain responses to sung streams predict vocabulary outcomes by age 18 months
Arenillas‐Alcón et al. Prenatal daily musical exposure is associated with enhanced neural representation of speech fundamental frequency: Evidence from neonatal frequency‐following responses
Hu et al. Dissociation of tone and vowel processing in M andarin idioms
Žarić et al. Crossmodal deficit in dyslexic children: practice affects the neural timing of letter-speech sound integration
Choi et al. On the early neural perceptual integrality of tones and vowels
Gibbon et al. Machine learning accurately classifies neural responses to rhythmic speech vs. non-speech from 8-week-old infant EEG
Parmar et al. A novel and efficient Wavelet Scattering Transform approach for primitive-stage dyslexia-detection using electroencephalogram signals
Kailaheimo-Lönnqvist et al. Infant event-related potentials to speech are associated with prelinguistic development
François et al. Attenuated brain responses to speech sounds in moderate preterm infants at term age
Schaadt et al. Maturation of consonant perception, but not vowel perception, predicts lexical skills at 12 months
Blanco et al. Functional Brain Adaptations During Speech Processing in 4‐Month‐Old Bilingual Infants
Virtala et al. Neural phoneme discrimination in variable speech in newborns–Associations with dyslexia risk and later language skills
Navarrete-Arroyo et al. Infant neural speech encoding is associated with pre-reading skill development
Beckers et al. Neural processing of short-term recurrence in songbird vocal communication
Bakhtiar et al. Categorical perception of speech sounds in adults who stutter
Ou et al. Induced gamma oscillations index individual differences in speech sound perception and production
Pettigrew et al. Review The relationship between the mismatch negativity (MMN) and psycholinguistic models of spoken word processing
Jeng Infant and childhood development: intersections between development and language experience
HK40096456A (zh) 预测婴幼儿的发育语言和交流能力的方法和工具

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21841784

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 18097578

Country of ref document: US

ENP Entry into the national phase

Ref document number: 202302076

Country of ref document: GB

Kind code of ref document: A

Free format text: PCT FILING DATE = 20210716

WWE Wipo information: entry into national phase

Ref document number: 2302076.1

Country of ref document: GB

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2021308673

Country of ref document: AU

Date of ref document: 20210716

Kind code of ref document: A

WWP Wipo information: published in national office

Ref document number: 2302076.1

Country of ref document: GB

122 Ep: pct application non-entry in european phase

Ref document number: 21841784

Country of ref document: EP

Kind code of ref document: A1

WWG Wipo information: grant in national office

Ref document number: 2302076.1

Country of ref document: GB

WWP Wipo information: published in national office

Ref document number: 18097578

Country of ref document: US