Hearing screening and tinnitus assessment method based on multi-frequency cross-analysis

Abstract

This invention discloses a hearing screening and tinnitus assessment method based on multi-frequency cross-analysis, specifically relating to the fields of hearing testing and neural information processing technology. It constructs an initial hearing vector in the individual frequency domain, generates a deviation feature vector based on a standard population database, extracts a set of abnormal frequency bands, and constructs a frequency domain focusing model. Multiple-intensity sound stimulation is applied to the abnormal frequency bands, and neural response parameter sequences are collected. Further cross-spectral analysis is performed to construct an individualized tinnitus resonance map, and the dominant frequency features and response intensity are extracted. Combined with subjective tinnitus scores, a tinnitus risk index is calculated, and a screening conclusion is output. This invention can achieve objective, quantitative, and spectral-level tinnitus identification, improving screening accuracy and clinical applicability.

Description

Technical Field

[0001] This invention relates to the field of hearing detection and neural information processing technology, specifically to a hearing screening and tinnitus assessment method based on multi-frequency cross-analysis. Background Technology

[0002] Tinnitus is a common but complex subjective symptom in clinical otolaryngology, and it is often difficult to objectively assess using conventional methods. Traditional hearing screening methods mainly use pure-tone audiometry, which can measure changes in a patient's hearing threshold, but it is difficult to accurately assess the frequency characteristics, masking properties, and potential associations between tinnitus and multi-band hearing abnormalities.

[0003] Especially in children, the elderly, and patients with "hidden tinnitus" who complain of tinnitus but have normal hearing thresholds, conventional single-frequency pure-tone testing is easily affected by subjective cooperation, making it difficult to comprehensively characterize their auditory frequency domain response characteristics. Studies have shown that tinnitus may involve abnormal neural activity in specific frequency bands, which often deviate from the patient's subjective description, leading to low screening efficiency and poor diagnostic accuracy. On the other hand, existing tinnitus assessment methods mostly rely on questionnaires or masking threshold tests, lacking a systematic and parameterized frequency domain analysis mechanism, and thus failing to form accurate assessment models that cater to individual differences. Summary of the Invention

[0004] The purpose of this invention is to provide a hearing screening and tinnitus assessment method based on multi-frequency cross-analysis to address the shortcomings of the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a hearing screening and tinnitus assessment method based on multi-frequency cross-analysis, comprising: Obtain the individuals to be screened in multiple frequency bands Initial auditory response value Construct the initial listening vector in the frequency domain where n is the total number of frequency bands; Based on a standard population hearing database, the relative deviation coefficients of the individuals in each frequency band were calculated. And construct the multi-frequency deviation feature vector D; Based on the feature vector D, the set of frequency bands F′ with abnormal deviations is extracted, and a frequency domain focusing model P is constructed to indicate sensitive frequency bands where tinnitus or hearing loss may exist. For each frequency band in the frequency band set F′ Multiple-intensity incremental sound stimulations were performed, and the corresponding neural response parameter sequences were recorded. , where m is the neural response parameter; For the parameter sequence Cross-spectral analysis was performed to construct an individualized tinnitus resonance feature map E(f, t), and the tinnitus dominant frequency features were extracted from the map. With response strength ; According to the above and Combined with individual subjective tinnitus level scores The tinnitus risk index TRI is calculated, and the final screening conclusion is output based on the threshold.

[0006] Preferably, the construction of the multi-frequency deviation feature vector D includes: A pre-defined standard population hearing database is used, which contains average hearing threshold data for multiple age groups, genders, and population categories, with each frequency band... Corresponding to a standard hearing threshold This constitutes a standard hearing vector. ; Based on the initial frequency domain hearing vector of the individual to be screened Each frequency band Individual threshold With the corresponding standard hearing threshold Compare and calculate the relative deviation coefficient. ; Arrange the relative deviation coefficients of all frequency bands in frequency order to construct a deviation feature vector. It is used to characterize the degree of frequency domain deviation of an individual's auditory response relative to a standard level.

[0007] Preferably, the step of extracting the set of frequency bands F′ with abnormal deviations based on the feature vector D includes: For eigenvectors Deviation coefficients for each frequency band Perform statistical analysis and set an abnormal deviation threshold. ,like The absolute value is greater than Then the corresponding frequency band Marked as an abnormal frequency range; All of them must be satisfied | | > frequency band Collect a set of abnormal frequency bands, F′; Based on the frequency distribution characteristics of the abnormal frequency band set F′, a piecewise response weight function W(f) is constructed in the frequency domain. This function is assigned high weight in the F′ region and low weight in the non-abnormal frequency band region, forming a frequency domain focusing model P(f) = W(f) × D(f). The frequency domain focusing model P is applied to the tinnitus assessment process to indicate the sensitive frequency range in which an individual may have abnormal tinnitus perception or hidden hearing loss.

[0008] Preferably, each frequency band in the frequency band set F′ Perform multi-intensity incremental sound stimulation, including: In each frequency band Internally set sound intensity sequence ,in The intensity is the i-th increasing tone. Apply frequency to individuals sequentially The intensity is Pure tone stimulation; Collect individual neurophysiological responses at each stimulus intensity and extract neural response parameters. This includes amplitude, latency, signal-to-noise ratio, and waveform duration; In frequency All neural response parameters corresponding to the intensity sequence L Composition of response parameter sequence This is used for subsequent construction of spectral resonance maps and analysis of tinnitus dominant frequency inference.

[0009] Preferably, the parameter sequence Cross-spectral analysis was performed to construct an individualized tinnitus resonance feature map E(f, t), including: For each frequency band in the frequency band set F′ The neural response parameter sequence Time axis interpolation and alignment are performed to construct a unified time series matrix R( , t); The time series matrix R is analyzed based on short-time Fourier transform or continuous wavelet transform. Perform a frequency-time domain joint transformation on (f, t) to obtain the neural response energy spectrum in the frequency f and time t dimensions; For all frequency bands The transformation results are fused to construct a two-dimensional resonance feature map E(f, t), where the horizontal axis is frequency f and the vertical axis is time t. The value corresponding to each point in the map is the neural response intensity per unit time. Extract the dominant frequency point with local maximum energy density from the spectrum E(f, t). and its corresponding peak response intensity These are respectively used as the dominant frequency and response amplitude characteristics of individual tinnitus.

[0010] Preferably, the calculation of the tinnitus risk index TRI includes: Obtain individual tinnitus dominant frequency characteristics Its corresponding maximum response strength , respectively, represent the frequency position and resonance amplitude identified in the tinnitus resonance feature spectrum; Receive individual subjective tinnitus rating The scoring is based on the subjects' self-report questionnaire converted into standardized rating values, which are used to reflect the perceived severity of tinnitus. according to , and A multi-factor weighted model was constructed to calculate the tinnitus risk index TRI.

[0011] Preferably, the step of determining and outputting the final screening conclusion based on the threshold includes: The TRI value is compared with the preset risk threshold δTRI. If TRI ≥ δTRI, the screening conclusion of "high risk of tinnitus" is output; otherwise, the result of "low risk of tinnitus" or "controllable risk" is output.

[0012] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. This invention provides a hearing screening and tinnitus assessment method based on multi-frequency cross-analysis. By introducing frequency domain deviation feature vectors, abnormal frequency band sets, neural response parameter sequences, and tinnitus resonance maps, it achieves full-process information acquisition and modeling from acoustic stimulation to neural response. Compared with traditional assessment methods that rely on pure-tone hearing thresholds or subjective questionnaires, this invention can identify abnormal neural responses and resonance characteristics in specific frequency bands, significantly improving the accuracy of tinnitus dominant frequency identification and the objectivity of assessment. It is particularly suitable for screening occult tinnitus in individuals who complain of tinnitus but have normal hearing.

[0013] 2. The tinnitus risk index (TRI) model constructed in this invention comprehensively considers subjective perception, physiological response intensity, and frequency risk distribution, achieving multi-source information fusion and automated risk assessment through a unified numerical model. This index can be used not only for high-throughput initial screening but also for generating personalized treatment recommendations and evaluating treatment efficacy. It possesses significant advantages such as clear structure, high sensitivity, and strong adaptability, providing a novel solution for intelligent screening and assisted diagnosis of tinnitus. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0015] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] For examples, please refer to Figure 1 As shown in this embodiment, the hearing screening and tinnitus assessment method based on multi-frequency cross-analysis includes: Obtain the individuals to be screened in multiple frequency bands Initial auditory response value Construct the initial listening vector in the frequency domain where n is the total number of frequency bands; Based on a standard population hearing database, the relative deviation coefficients of the individuals in each frequency band were calculated. And construct the multi-frequency deviation feature vector D; Based on the feature vector D, the set of frequency bands F′ with abnormal deviations is extracted, and a frequency domain focusing model P is constructed to indicate sensitive frequency bands where tinnitus or hearing loss may exist. For each frequency band in the frequency band set F′ Multiple-intensity incremental sound stimulations were performed, and the corresponding neural response parameter sequences were recorded. , where m is the neural response parameter; For the parameter sequence Cross-spectral analysis was performed to construct an individualized tinnitus resonance feature map E(f, t), and the tinnitus dominant frequency features were extracted from the map. With response strength ; According to the above and Combined with individual subjective tinnitus level scores The tinnitus risk index TRI is calculated, and the final screening conclusion is output based on the threshold.

[0018] To comprehensively assess an individual's sensitivity across different auditory frequency bands, this invention sets the test frequency range to 100 Hz to 12000 Hz (i.e., 0.1 kHz to 12 kHz), covering the range of common human auditory frequencies. This frequency range is evenly divided into n frequency segments, where n is a positive integer greater than or equal to 8, with a recommended value of 16 or 24, to increase the frequency domain sampling density without compromising test time.

[0019] For each frequency band (where i is an integer between 1 and n), auditory stimulation is provided using an increased-order sine wave audio signal. An increased-order sine wave refers to an audio signal with a given frequency... The sound pressure level (i.e., sound intensity) gradually increases according to a set step value, constructing a stimulus sequence from low to high. For example, the sound intensity sequence is set as L={20 dB, 30 dB, 40 dB, 50 dB, 60 dB, 70 dB, 80 dB}, and the system sequentially plays multiple sound intensity signals at the corresponding frequencies until the individual experiences a subjective auditory response.

[0020] Individual auditory response detection is based on behavioral feedback patterns. Specifically, feedback can be collected using one of the following methods: Signal presentation is performed using a standard hearing assessment system (such as a clinical pure-tone audiometer) in conjunction with an in-ear monitor. Whether the subject perceives the auditory stimulus is confirmed through interactive methods such as button response, voice recognition, or eye tracking.

[0021] Once an individual is at a certain frequency The first sound detected, and its corresponding sound pressure level This is recorded as the subjective hearing threshold at that frequency. and One-to-one correspondence yields frequency-threshold pairs ( , ).

[0022] Combining the response results across all frequency bands yields the initial hearing vector in the frequency domain. Vector ;in, This represents the i-th test frequency band. This represents the individual's subjective auditory perception threshold in this frequency band. It is a two-dimensional ordered set of pairs of length n, which can be used to characterize the frequency domain hearing sensitivity profile of an individual.

[0023] To eliminate deviations caused by individual reaction delays, equipment errors, or environmental interference, the initial vector... Outlier removal is performed using the following process: Response time window setting: Feedback must be given within 2 seconds after each stimulus sound is played. Data points that do not respond within this time will be marked as "hysteresis response" and discarded. Threshold range constraint: The acceptable range for hearing threshold is set between 20 dB and 80 dB. Data points exceeding this range are marked as "invalid response". Statistical residual detection: Calculate the mean and standard deviation of all thresholds, and remove outliers that deviate from the mean by more than twice the standard deviation (i.e., use the 3σ rule for removal).

[0024] The processed vector is denoted as ′ is the effective hearing vector in the frequency domain after filtering, noise reduction, and stability enhancement.

[0025] This step can be implemented in the following two ways: Software-driven pure-tone audiometry systems include: The hardware components include: plug-in earphones, audio output module, subject feedback button, and main control terminal; The software component is based on MATLAB, LabVIEW, or Python GUI frameworks to control stimulus generation, audio output, and feedback acquisition. The system automatically controls the sound intensity presentation and threshold judgment of each frequency band, realizing the acquisition of high-throughput auditory response data.

[0026] Mobile hearing screening system: Suitable for remote screening scenarios; The system outputs a pure tone signal in a specified frequency band via Bluetooth headset; Users can confirm whether they have heard the sound by touching the screen or using voice. The application uses a built-in data calibration module to dynamically adjust for differences in audio output across different devices, ensuring consistent testing standards.

[0027] This invention pre-defines a standard population hearing database as a reference template for calculating individual hearing threshold deviations. This database is constructed based on clinical audiometry experiments, literature review data, and existing national standards (such as GB / T 7582-2004), and covers the following dimensions: Frequency band division: The test frequency bands should be consistent with the initial vector, covering the complete human hearing frequency range from 100 Hz to 12000 Hz. It is recommended to use a geometric division or a 1 / 3 octave band division method, with the number of frequency bands n being a positive integer greater than or equal to 8.

[0028] Population classification factors: Age stratification (e.g., 0–18 years, 19–40 years, 41–60 years, 60 years and older). Gender classification (male, female); Occupational exposure (e.g., history of long-term noise exposure); Health status label (e.g., history of ear disease).

[0029] Hearing threshold setting: for each frequency band Corresponding to the average hearing threshold of a standard population (Unit: sound pressure level in decibels, dB SPL), and stored in vector form to form a standard hearing vector: Standard hearing vector ;in, Indicates frequency Below, the average hearing threshold level of the reference population is used. This standard data is automatically selected according to the user's screening input information to determine the corresponding category.

[0030] Obtain the initial frequency domain hearing vector of the individual to be screened. Then, extract its frequency bands respectively. The lower hearing threshold , and the standard hearing vector The corresponding frequency band Standard threshold below Compare them.

[0031] To avoid distortion caused by differences in hearing threshold units or amplitudes, this invention employs a "relative deviation coefficient" algorithm to standardize and compare data across frequency bands. The specific calculation formula is as follows: In the i-th frequency band Above, the relative deviation coefficient of an individual ;in, For individuals at frequency The measured hearing threshold; The standard hearing threshold at the same frequency is defined in the standard database. This is the relative deviation coefficient for that frequency band, representing the degree of deviation in an individual's auditory perception within that frequency band compared to a standard hearing level. If... A positive value indicates that the individual's hearing is weaker than that of the general population (the threshold is higher); a negative value indicates that the individual may have sensitivity or over-hearing in that frequency band.

[0032] All the above frequency bands to The calculated deviation coefficient to Arranged in ascending order of frequency, they form the deviation eigenvector D, defined as follows: Multi-frequency deviation eigenvector The vector D is a one-dimensional real number sequence with a length equal to the total number of frequency bands n, reflecting the distribution characteristics of an individual's hearing deviation across the entire auditory frequency domain. By analyzing the vector D, the frequency bands of hearing loss, the degree of deviation, and the trend of spectral morphology can be identified, thus providing a key parameter basis for subsequent steps such as focusing on sensitive frequency bands (e.g., constructing a frequency domain focusing model) and generating tinnitus spectrograms.

[0033] To ensure good comparability and input stability of D vectors among different samples, this invention performs normalization processing after vector construction, including: Mean normalization (min-max scaling): For each element in vector D... Mapped to the range of 0 to 1; Z-score normalization: Normalizes the D vector to a mean of 0 and a standard deviation of 1, which is suitable for subsequent cluster analysis or neural network input scenarios; Outlier handling: Identify and remove outlier frequency bands whose absolute value of the deviation coefficient exceeds 2.5 times the standard deviation to improve the robustness of subsequent modeling.

[0034] The degree of deviation of an individual in different frequency bands is determined by the deviation eigenvector. It means that, among them For individuals in frequency bands The relative deviation coefficient is calculated as follows: ,in It is the individual at frequency The hearing threshold below This represents the average hearing threshold corresponding to the standard population.

[0035] To identify regions in the frequency domain where significant hearing shifts may exist, this invention employs a threshold method for determining the absolute value of the deviation, i.e., if a certain... The absolute value is greater than the preset deviation threshold. If so, then it is considered that there is an anomaly in that frequency band.

[0036] To improve the adaptability and robustness of the algorithm, Instead of using fixed constants, the calculation is dynamically based on the statistical properties of the D vector. The specific method is as follows: Calculate the mean μ of the deviation vector D; Calculate the standard deviation σ of the deviation vector D; Set the deviation adjustment coefficient k (the empirical value ranges from 1.5 to 2.5). Then the deviation threshold = μ + k × σ; This method allows the system to automatically adapt to the deviation distribution of different hearing types, age groups, and sample groups, thereby improving the accuracy of identifying abnormal frequency bands in individuals.

[0037] Iterate through all elements in vector D, for each If its absolute value | |greater than Then its corresponding frequency Add the abnormal frequency range set F′. The final definition of F′ is: F′ is a high-bias subset of the D vector in the frequency domain, reflecting the sensitive frequency range in which an individual may have auditory perception abnormalities or potential tinnitus sources.

[0038] To ensure that subsequent tinnitus analysis focuses on the individual's sensitive frequency range, this invention further proposes constructing a frequency-domain focusing model P, which serves as a weighted modulation function in the frequency domain. This model is used to highlight the weight of abnormal frequency ranges and weaken the influence of non-abnormal regions. The construction of this model includes the following steps: Define a weighting function W(f) in the frequency domain, with a range of [0,1]. Assign higher weight values ​​to the abnormal frequency band set F′ and lower weight values ​​to other frequency bands.

[0039] The specific construction method is as follows: If f ∈ F′, then W(f) = 1; If f ∉ F′, then W(f) = α, 0 < α < 1, where α is the background weight attenuation coefficient (a value of 0.2–0.3 is recommended).

[0040] Using frequency as the independent variable, the characteristic deviation vector D is mapped to the function D(f), and then multiplied by the weighting function W(f) to obtain the frequency domain focusing model P(f), that is: P(f) = D(f) × W(f); where: P(f) is the focusing value at frequency f; D(f) is the deviation function after linear interpolation; W(f) is the frequency weighting function.

[0041] The final model P has the strongest focusing response in the abnormal frequency band and maintains background suppression in the non-abnormal region, thus achieving targeted extraction and focused amplification of frequency domain information.

[0042] In each frequency band A set of intensity level steps is set in the middle. ,in: to It represents the increasing sound intensity from the 1st to the mth order, and the unit is sound pressure level (decibel SPL). The recommended setting is 20 decibels. The intensity should not exceed 80 dB to prevent cochlear overload. The increment for each level can be set to 5 or 10 dB, dynamically adjusted according to the subject's age and tolerance. A recommended intensity sequence example is: L = {20, 30, 40, 50, 60, 70} (6 intensity levels in total). This setting helps observe the threshold sensitivity, saturation, and nonlinear trends of individual neural responses.

[0043] For frequencies of The intensity is The stimulus is presented using pure tones (i = 1 to m), as follows: The duration of each pure-tone stimulus is set to 300 milliseconds to 1 second, with 500 milliseconds recommended. The interval between two consecutive stimuli should be no less than 2 seconds to avoid neural adaptive inhibition and auditory fatigue. The stimulus waveform uses a continuous sine wave signal, output from a standard signal generator or a dedicated pure-tone generation module. The output signal is transmitted to the external auditory canal through an in-ear headphone, maintaining consistent sound source localization (e.g., unilateral or bilateral). If using electrophysiological equipment (such as an event-related potential (ERP) system), each intensity is repeated 5 times for average waveform calculation. After each stimulus, the system enters the acquisition phase, and a state buffer is required before proceeding to the next stimulus level.

[0044] At each intensity level The following methods are used: acquiring individual electrophysiological response signals using neural response detection equipment. The following equipment and techniques can be employed: Brainstem evoked potential (ABR) system: used to record auditory nerve responses at the brainstem level, with the commonly used waveform being IV; Event-related potential (ERP) system: used to detect time-locked neural responses at the cortical level; Otoacoustic emission (OAE) system: used to detect the function of outer hair cells in the cochlea; Multichannel EEG system: used for monitoring more complex auditory evoked activity.

[0045] The system extracts the following neural response parameters (m dimensions) for each stimulus response waveform: Amplitude: The difference in amplitude between the main peak and the baseline in a neural response waveform, measured in microvolts; Latency: The time delay from the onset of the stimulus sound to the peak of the main response, measured in milliseconds; Signal-to-noise ratio (SNR): The ratio between the neural response signal and the background noise, characterizing the validity of the data; Waveform duration: The length of time from the first significant rise to the recovery to the baseline.

[0046] The above parameters can be extracted using automatic algorithms (such as peak detection and curve integration) or manual labeling.

[0047] frequency All intensity The corresponding neural response parameter sets are combined to construct a complete response parameter sequence. The definition is as follows: = { , , ..., },in ;in: For frequency The complete neural response trajectory is shown below; For sound intensity The neural response parameter set below; Indicates amplitude, Indicates the incubation period. Indicates the signal-to-noise ratio. Indicates the duration of the waveform. It can be viewed as a four-dimensional time series, where sound intensity is the input variable, and each order corresponds to a complete multidimensional neurophysiological response.

[0048] Neural response parameter sequence As defined in the preceding steps: Each of them ; respectively represent amplitude (Amplitude), incubation period (Latency), Signal-to-noise ratio (SNR), Waveform Duration (Duration).

[0049] Due to different frequency bands The timing and duration of the stimulation may not be consistent; therefore, it is necessary to first analyze all... Perform uniform timeline interpolation and alignment operations on the sequences: Linear interpolation or spline interpolation is used to uniformly map each parameter sequence onto a standardized time axis t ∈ [0,T], where T is the total acquisition window duration (e.g., 5 seconds); the parameter function is obtained after interpolation. .

[0050] To enhance the comprehensiveness of spectral analysis, a normalized neural response intensity function R( , t), constructed as follows: ;in, , , This is an adjustable weighting coefficient used to balance the effects of amplitude, signal-to-noise ratio, and latency. The recommended default value is [value missing]. = 0.5, = 0.3, = 0.2.

[0051] To extract the coupling relationship between the frequency domain and the time domain, this invention employs Short-Time Fourier Transform (STFT) or Continuous Wavelet Transform (CWT) to analyze the neural response function R( Process , t) For example, the STFT implementation is as follows: Use a sliding time window (e.g., 200 milliseconds) to segment the t-axis; For each segment, the frequency response spectrum within that time period is calculated using Fast Fourier Transform (FFT); The output is a joint power spectrum of frequency f and time t. (f, t).

[0052] For all frequency bands The frequency-time spectrum obtained after processing (e.g.) or The two-dimensional resonance spectrum E(f, t) is constructed by fusing the two signals and is defined as follows: ;in: (f, t) represents the frequency band. The response spectrum below; is a Gaussian smoothing weight function used for continuous fusion of nearest-neighbor frequency bands in the frequency domain; f is the frequency on the horizontal axis of the spectrum, and t is the time on the vertical axis of the spectrum. Each pixel in the spectrum E(f, t) represents the instantaneous resonance intensity of the nervous system at a specific time t and frequency f.

[0053] The tinnitus dominant frequency is extracted from the constructed resonance spectrum E(f, t). and its corresponding maximum response strength The method is as follows: Calculate the total energy density corresponding to each frequency f in the spectrum: E_total(f) = The total energy spectrum E_total(f) on the frequency axis is obtained, which is used to reflect the cumulative neural response intensity of each frequency band.

[0054] clock speed Defined as the frequency point at which E_total(f) reaches its maximum value: = argmax [E_total(f)]; that is, the frequency with the highest energy response in the frequency domain is the individual's possible subjective tinnitus dominant frequency.

[0055] The response intensity corresponding to the main frequency is defined as: = max [E( [, t)]; that is, at the main frequency The maximum resonance amplitude at a certain point is used to quantify the neural activity level of tinnitus perception.

[0056] The tinnitus resonance feature map E(f, t) has been constructed in the aforementioned steps. It is a two-dimensional function map that reflects the distribution of neural response energy at different frequencies f and times t.

[0057] and These two indicators, which respectively reflect the subjective perceived frequency of tinnitus and the intensity level of corresponding neural activity, are the core objective indicators for constructing a tinnitus risk index.

[0058] Tinnitus is highly subjective, and objective neurological parameters alone may not fully reflect its clinical severity. Therefore, this invention introduces an individual subjective tinnitus rating system. This serves as a supplementary factor. The score is based on a standardized questionnaire (such as the Tinnitus Handicap Inventory, THI, or the Visual Analogue Scale, VAS), which is then algorithmically converted into uniform level values. The standard is a subjective tinnitus rating, with a range of 0 to 10. The higher the value, the more severe the individual's perceived tinnitus.

[0059] The following quantification method is recommended: THI ≤ 16: = 1 (slight); THI 18–36: = 3 (Light to Medium); THI 38–56: = 5 (Medium); THI 58–76: = 7 (medium heavy); THI ≥ 78: = 9 or 10 (severe); Combining this subjective rating with resonance parameters helps to achieve integrated modeling of physiological data and self-perception.

[0060] To achieve a quantitative expression of tinnitus risk level, this invention proposes the following multi-factor weighted model for calculating the tinnitus risk index TRI, expressed as follows: Where: TRI: Tinnitus Risk Index, the final output value, with no upper limit on the value range; The maximum neural response intensity corresponding to the dominant frequency is normalized to 0 to 1. Subjective tinnitus score, ranging from 0 to 10; g( ): Tinnitus dominant frequency risk function, used to measure The probability of physiological abnormalities in the frequency band; α, β, γ: weighting coefficients used to adjust the relative influence of the three factors. Recommended values ​​are: α = 0.4, β = 0.4, γ = 0.2.

[0061] Tinnitus occurs more frequently in certain frequency ranges (such as high frequencies ≥ 8000 Hz or low frequencies ≤ 250 Hz), g( This is used to model the pattern. It is defined as follows: like If ∈ high-risk frequency bands (e.g., 8000–12000 Hz or 125–250 Hz), then g( If ) = 1; If ∈ the medium-risk frequency band (e.g., 4000–8000 Hz), then g( If ) = 0.5; If ∈ the center of the normal hearing frequency range (e.g., 500–3000Hz), then g( = 0.2; This can be achieved using piecewise linear functions or Gaussian distribution modeling. This allows the risk of the frequency band containing the dominant frequency to be mapped as a quantified value for use in TRI calculations.

[0062] After calculating the TRI value, the results need to be classified and judged based on the preset risk threshold δTRI, and the screening conclusion should be output. The classification rules are as follows: If TRI ≥ δTRI, the output conclusion is "high risk of tinnitus", and it is recommended to proceed to further clinical intervention or re-examination. If TRI < δTRI, then output "Low risk of tinnitus" or "Risk is controllable"; The recommended setting for the threshold δTRI is: Initial screening scenario: δTRI = 5.5 (moderate sensitivity); High-sensitivity early warning scenarios: δTRI = 4.0 (improves recall rate); δTRI can be optimized based on population data using methods such as ROC curves and Youden Index to achieve the best screening accuracy.

[0063] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A hearing screening and tinnitus assessment method based on multi-frequency cross-analysis, characterized in that: include: Obtain the individuals to be screened in multiple frequency bands Initial auditory response value Construct the initial listening vector in the frequency domain where n is the total number of frequency bands; Based on a standard population hearing database, the relative deviation coefficients of the individuals in each frequency band were calculated. And construct the multi-frequency deviation feature vector D; Based on the feature vector D, the set of frequency bands F′ with abnormal deviations is extracted, and a frequency domain focusing model P is constructed to indicate sensitive frequency bands where tinnitus or hearing loss may exist. For each frequency band in the frequency band set F′ Multiple-intensity incremental sound stimulations were performed, and the corresponding neural response parameter sequences were recorded. , where m is the neural response parameter; For the parameter sequence Cross-spectral analysis was performed to construct an individualized tinnitus resonance feature map E(f, t), and the tinnitus dominant frequency features were extracted from the map. With response strength ; According to the above and Combined with individual subjective tinnitus level scores The tinnitus risk index TRI is calculated, and the final screening conclusion is output based on the threshold.

2. The hearing screening and tinnitus assessment method based on multi-frequency cross-analysis according to claim 1, characterized in that: The construction of the multi-frequency deviation feature vector D includes: A pre-defined standard population hearing database is used, which contains average hearing threshold data for multiple age groups, genders, and population categories, with each frequency band... Corresponding to a standard hearing threshold This constitutes a standard hearing vector. ; Based on the initial frequency domain hearing vector of the individual to be screened Each frequency band Individual threshold With the corresponding standard hearing threshold Compare and calculate the relative deviation coefficient. ; Arrange the relative deviation coefficients of all frequency bands in frequency order to construct a deviation feature vector. It is used to characterize the degree of frequency domain deviation of an individual's auditory response relative to a standard level.

3. The hearing screening and tinnitus assessment method based on multi-frequency cross-analysis according to claim 2, characterized in that: The step of extracting the set of frequency bands F′ with abnormal deviations based on the feature vector D includes: For eigenvectors Deviation coefficients for each frequency band Perform statistical analysis and set an abnormal deviation threshold. ,like The absolute value is greater than Then the corresponding frequency band Marked as an abnormal frequency range; All of them must be satisfied | | > frequency band Collect a set of abnormal frequency bands, F′; Based on the frequency distribution characteristics of the abnormal frequency band set F′, a piecewise response weight function W(f) is constructed in the frequency domain. This function is assigned high weight in the F′ region and low weight in the non-abnormal frequency band region, forming a frequency domain focusing model P(f) = W(f) × D(f); The frequency domain focusing model P is applied to the tinnitus assessment process to indicate the sensitive frequency range in which an individual may have abnormal tinnitus perception or hidden hearing loss.

4. The hearing screening and tinnitus assessment method based on multi-frequency cross-analysis according to claim 3, characterized in that: Each frequency band in the frequency band set F′ Perform multi-intensity incremental sound stimulation, including: In each frequency band Internally set sound intensity sequence ,in The intensity is the i-th increasing tone. Apply frequency to individuals sequentially The intensity is Pure tone stimulation; Collect individual neurophysiological responses at each stimulus intensity and extract neural response parameters. This includes amplitude, latency, signal-to-noise ratio, and waveform duration; In frequency All neural response parameters corresponding to the intensity sequence L Composition of response parameter sequence This is used for subsequent construction of spectral resonance maps and analysis of tinnitus dominant frequency inference.

5. The hearing screening and tinnitus assessment method based on multi-frequency cross-analysis according to claim 4, characterized in that: The parameter sequence Cross-spectral analysis was performed to construct an individualized tinnitus resonance feature map E(f, t), including: For each frequency band in the frequency band set F′ The neural response parameter sequence Time axis interpolation and alignment are performed to construct a unified time series matrix R( , t); The time series matrix R is analyzed based on short-time Fourier transform or continuous wavelet transform. Perform a frequency-time domain joint transformation on (f, t) to obtain the neural response energy spectrum in the frequency f and time t dimensions; For all frequency bands The transformation results are fused to construct a two-dimensional resonance feature map E(f, t), where the horizontal axis is frequency f and the vertical axis is time t. The value corresponding to each point in the map is the neural response intensity per unit time. Extract the dominant frequency point with local maximum energy density from the spectrum E(f, t). and its corresponding peak response intensity These are respectively used as the dominant frequency and response amplitude characteristics of individual tinnitus.

6. The hearing screening and tinnitus assessment method based on multi-frequency cross-analysis according to claim 5, characterized in that: The calculation of the tinnitus risk index TRI includes: Obtain individual tinnitus dominant frequency characteristics Its corresponding maximum response strength , respectively, represent the frequency position and resonance amplitude identified in the tinnitus resonance feature spectrum; Receive individual subjective tinnitus rating The scoring is based on the subjects' self-report questionnaire converted into standardized rating values, which are used to reflect the perceived severity of tinnitus. according to , and A multi-factor weighted model was constructed to calculate the tinnitus risk index TRI.

7. The hearing screening and tinnitus assessment method based on multi-frequency cross-analysis according to claim 6, characterized in that: The final screening conclusion is output based on the threshold determination, including: The TRI value is compared with the preset risk threshold δTRI. If TRI ≥ δTRI, the screening conclusion of "high risk of tinnitus" is output; otherwise, the result of "low risk of tinnitus" or "controllable risk" is output.

Images