Noise hearing loss susceptible population screening method and system

Abstract

The invention relates to the technical field of hearing detection, and discloses a noise-induced hearing loss susceptible population screening method and system, and the method comprises the steps: carrying out the dimension reduction processing of cochlea image data, obtaining a two-dimensional semantic image, carrying out the dimension rising processing of the two-dimensional semantic image, obtaining a three-dimensional semantic image, and carrying out the noise-induced hearing loss susceptible population screening. Determining corresponding attention parameters according to the semantic feature data, fusing the attention parameters and the three-dimensional semantic image to obtain a target semantic image, traversing M groups of noise reaction data to calculate corresponding noise reaction coefficients, determining a noise influence value according to the M noise reaction coefficients, and obtaining a target semantic image; and finally, inputting the target semantic image and the noise influence value into a pre-constructed user classification model to obtain a user category. According to the method, the susceptibility difference of individuals under noise exposure is fully considered, and the susceptibility is accurately evaluated in combination with the three-dimensional semantic image and the noise influence value, so that the user categories of the to-be-screened users can be accurately distinguished.

Description

Technical Field

[0001] The present invention relates to the technical field of hearing detection, and more specifically, to a method and system for screening people susceptible to noise-induced hearing loss. Background Art

[0002] Noise is one of the most common pollutants in modern society. Long-term and / or high-intensity noise exposure can damage the sensory hair cells, spiral neurons and supporting cells of the inner ear cochlea, leading to hearing loss. Since there is no effective medical treatment for noise-induced hearing loss, early monitoring and early prevention are currently the key to reducing the incidence rate. Noise-induced hearing loss is a complex disease caused by the combined effects of genetics and environment. Individual susceptibility varies between individuals exposed to similar noise. With existing technologies, there is little relevant content that can assess individual susceptibility.

[0003] In the prior art, a Chinese patent application with publication number CN116671887A provides a device for screening people at high risk of sudden cardiac death based on photoplethysmographic pulse wave signals. This technology obtains PPG signals through a data acquisition module, and transmits the processed signals to a data processing part through a signal processing and uploading module. The data processing part includes a feature extraction and screening module, and a classification screening module, which classifies and evaluates the risk of sudden cardiac death according to the extracted features; a Chinese patent application with publication number CN117243573A provides a method, system and storage medium for screening sensitive populations and improving ablation sensitivity. This technology classifies the subjects to be tested into sensitive populations or insensitive populations by analyzing medical images and obtaining influencing factors. Sensitive populations have better healing effects after ablation.

[0004] Existing technologies mainly screen for common diseases such as cardiovascular disease, but rarely involve screening for noise-induced hearing loss, and do not fully consider the differences in individual susceptibility to noise exposure. Noise-induced hearing loss is a complex disease caused by the combined effects of genetics and environment, and individual sensitivity varies greatly. Existing screening methods make it difficult to accurately assess susceptibility.

[0005] In view of this, the present invention proposes a method and system for screening people susceptible to noise-induced hearing loss to solve the above-mentioned problem. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method for screening a population susceptible to noise-induced hearing loss.

[0007] To achieve the above object, the present invention provides the following technical solutions: In a first aspect, a method for screening a population susceptible to noise-induced hearing loss is provided, comprising: Obtaining cochlear image data of the user to be screened, performing dimensionality reduction processing on the cochlear image data to obtain a two-dimensional semantic image, and performing dimensionality increase processing on the two-dimensional semantic image to obtain a three-dimensional semantic image; Acquire semantic feature data corresponding to the two-dimensional semantic image, determine corresponding attention parameters according to the semantic feature data, fuse the attention parameters and the three-dimensional semantic image to obtain a target semantic image; Under M groups of noise parameter environments, obtain M groups of noise response data of the users to be screened, traverse the M groups of noise response data to calculate corresponding noise response coefficients, and determine the noise impact value according to the M noise response coefficients; Input the target semantic image and the noise impact value into a pre-built user classification model to obtain user categories, where the user categories at least include susceptible categories and resistant categories; Before the dimension reduction process is performed on the cochlear image data, the following steps are also included: The cochlear image data is sequentially subjected to wavelet decomposition, guided filtering enhancement and fusion processing to generate first image data, the first image data is subjected to basilar membrane curvature correction processing to obtain second image data, and the cochlear image data is replaced by the second image data.

[0008] Furthermore, the method of performing dimensionality reduction processing on the cochlear image data to obtain a two-dimensional semantic image includes: The cochlear image data is projected to obtain W two-dimensional slices, the pixel matrix corresponding to each two-dimensional slice is obtained, the covariance matrix corresponding to the pixel matrix is ​​calculated, the covariance matrix is ​​eigendecomposed to obtain Q initial eigenvectors and corresponding eigenvalues, the Q initial eigenvectors are screened according to the eigenvalues ​​to obtain S target eigenvectors, and the S target eigenvectors are converted into a two-dimensional semantic image.

[0009] Furthermore, the method of calculating the covariance matrix corresponding to the pixel matrix includes: The pixel matrix is ​​expanded into a one-dimensional vector X, the mean of the one-dimensional vector X is calculated, and the mean is subtracted from each pixel value in the one-dimensional vector X, and the covariance corresponding to each pixel value is calculated to generate a covariance matrix.

[0010] Furthermore, the method of performing dimensionality-upgrading processing on the two-dimensional semantic image to obtain a three-dimensional semantic image includes: Based on the preset transposed convolutional layer, S two-dimensional semantic images are mapped into three-dimensional space, and the mapped two-dimensional semantic images are feature spliced ​​to obtain a three-dimensional semantic image.

[0011] Furthermore, the method for determining corresponding attention parameters according to semantic feature data includes: Based on the semantic feature data, standard feature data with the greatest similarity is determined from a pre-constructed database, and standard parameters associated with the standard feature data are used as attention parameters corresponding to the semantic feature data.

[0012] Furthermore, the noise response data includes the number of brain wave cycles, average reaction time, and average blood pressure value. The method of traversing M groups of noise response data to calculate the corresponding noise response coefficient includes: calculating the corresponding noise response coefficient according to the number of brain wave cycles, average reaction time, and average blood pressure value.

[0013] Furthermore, the method for determining the noise impact value according to the M noise response coefficients includes: The occurrence frequency values ​​corresponding to the M groups of noise parameters are obtained, the corresponding weight factors are determined according to the occurrence frequency values ​​of each group of noise parameters, and the noise impact values ​​are calculated according to the weight factors and the M noise response coefficients.

[0014] Furthermore, the method of sequentially performing wavelet decomposition, guided filtering enhancement and fusion processing on the cochlear image data to generate the first image data includes: The cochlear image data is decomposed by using wavelet transform, low-frequency components and high-frequency components are extracted, guided filtering is performed on the high-frequency components to obtain enhanced high-frequency components, and the enhanced high-frequency components are fused with the low-frequency components to obtain the first image data.

[0015] Furthermore, the method of performing basement membrane curvature correction processing on the first image data to obtain the second image data includes: The central spiral line of the first image data is extracted using a skeleton extraction algorithm, and the central spiral line is converted into the frequency domain through Fourier transform to obtain the basilar membrane unfolding line. The first image data with the basilar membrane unfolding line is used as the second image data, wherein the central spiral line represents the axis line of the cochlea.

[0016] In a second aspect, a system for screening a population susceptible to noise-induced hearing loss is provided, which is used to implement the above-mentioned method for screening a population susceptible to noise-induced hearing loss, comprising: Image processing module: used to obtain the cochlear image data of the user to be screened, perform dimensionality reduction processing on the cochlear image data to obtain a two-dimensional semantic image, and perform dimensionality increase processing on the two-dimensional semantic image to obtain a three-dimensional semantic image; Data fusion module: used to obtain the semantic feature data corresponding to the two-dimensional semantic image, determine the corresponding attention parameter according to the semantic feature data, and fuse the attention parameter and the three-dimensional semantic image to obtain the target semantic image; Data generation module: used to obtain M groups of noise response data of the users to be screened under M groups of noise parameter environments, traverse the M groups of noise response data to calculate the corresponding noise response coefficients, and determine the noise impact value according to the M noise response coefficients; User classification module: used to input the target semantic image and noise impact value into the pre-built user classification model to obtain user categories, where the user categories include at least susceptible categories and resistant categories; Before the dimension reduction process is performed on the cochlear image data, the following steps are also included: The cochlear image data is sequentially subjected to wavelet decomposition, guided filtering enhancement and fusion processing to generate first image data, the first image data is subjected to basilar membrane curvature correction processing to obtain second image data, and the cochlear image data is replaced by the second image data.

[0017] Compared with the prior art, the present invention has the following beneficial effects: The present invention first performs dimensionality reduction processing on cochlear image data to obtain a two-dimensional semantic image, performs dimensionality increase processing on the two-dimensional semantic image to obtain a three-dimensional semantic image, then determines corresponding attention parameters according to semantic feature data, fuses the attention parameters and the three-dimensional semantic image to obtain a target semantic image, traverses M groups of noise response data to calculate corresponding noise response coefficients, determines noise impact values ​​according to the M noise response coefficients, and finally inputs the target semantic image and the noise impact values ​​into a pre-built user classification model to obtain user categories. The present invention fully considers the differences in individual susceptibility to noise exposure, combines the three-dimensional semantic image and the noise impact value to accurately evaluate the susceptibility, and can accurately distinguish the user categories of the users to be screened. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of a process of screening a population susceptible to noise-induced hearing loss in the present invention; Figure 2 It is a structural schematic diagram of a system for screening people susceptible to noise-induced hearing loss in the present invention; Figure 3 It is a schematic diagram of the flow of the method for performing dimensionality reduction processing on cochlear image data to obtain a two-dimensional semantic image in the present invention; Figure 4 Schematic diagram of the process of calculating the covariance matrix corresponding to the pixel matrix in the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Example 1

[0021] See also Figure 1 As shown, this embodiment discloses a method for screening a population susceptible to noise-induced hearing loss, comprising: S10: Obtaining cochlear image data of the user to be screened, performing dimensionality reduction processing on the cochlear image data to obtain a two-dimensional semantic image, and performing dimensionality increase processing on the two-dimensional semantic image to obtain a three-dimensional semantic image; In this embodiment, the users to be screened refer to users who need to be monitored for hearing loss. The cochlear image data may be a three-dimensional structural image of the cochlea obtained through MRI technology. MRI technology is a non-invasive medical imaging technology that uses strong magnetic fields and radio waves to generate detailed images of internal structures. Specifically, MRI aligns hydrogen atoms in the human body through a magnetic field, and then uses radio frequency pulses to make these hydrogen atoms emit signals, and finally processes these signals through a computer to generate high-resolution images.

[0022] Before the dimension reduction process is performed on the cochlear image data, the following steps are also included: The cochlear image data is sequentially subjected to wavelet decomposition, guided filtering enhancement and fusion processing to generate first image data, the first image data is subjected to basilar membrane curvature correction processing to obtain second image data, and the cochlear image data is replaced by the second image data.

[0023] The method of sequentially performing wavelet decomposition, guided filtering enhancement and fusion processing on cochlear image data to generate first image data includes: The cochlear image data is decomposed by using wavelet transform, low-frequency components and high-frequency components are extracted, guided filtering is performed on the high-frequency components to obtain enhanced high-frequency components, and the enhanced high-frequency components are fused with the low-frequency components to obtain the first image data.

[0024] It should be noted that in the cochlear images of patients with noise-induced hearing loss, there are usually features such as hair cell damage and spiral neuron degeneration, which may not be obvious in the original MRI / CT images. Directly reducing the dimensionality of the original cochlear image data will lead to the loss of information in these key damaged areas, affecting the accuracy of subsequent classification. Therefore, this embodiment needs to perform special image enhancement and correction according to the characteristics of the cochlear image before dimensionality reduction to ensure that key information is retained to the greatest extent.

[0025] like Figure 3 As shown, the method of performing dimensionality reduction processing on cochlear image data to obtain a two-dimensional semantic image includes: The cochlear image data is projected to obtain W two-dimensional slices, the pixel matrix corresponding to each two-dimensional slice is obtained, the covariance matrix corresponding to the pixel matrix is ​​calculated, the covariance matrix is ​​eigendecomposed to obtain Q initial eigenvectors and corresponding eigenvalues, the Q initial eigenvectors are screened according to the eigenvalues ​​to obtain S target eigenvectors, and the S target eigenvectors are converted into a two-dimensional semantic image.

[0026] It should be noted that the cochlear image data is a three-dimensional structural image of the cochlea, and a two-dimensional slice refers to a two-dimensional plane image cut out from the three-dimensional structural image of the cochlea. These slices can be images cut out along different directions (such as cross-sections, longitudinal sections or coronal planes). Each slice is a two-dimensional projection of the three-dimensional structure of the cochlea on the plane. Since each two-dimensional slice is a two-dimensional plane image, the corresponding pixel matrix can be obtained. The pixel matrix represents the distribution of pixel intensity or brightness values ​​in the two-dimensional slice. Specifically, each element of the pixel matrix represents the grayscale or color intensity of a certain position of the three-dimensional structure of the cochlea on the plane projection, reflecting the tissue density or other characteristics of different regions of the cochlea.

[0027] like Figure 4 As shown, the method for calculating the covariance matrix corresponding to the pixel matrix includes: The pixel matrix is ​​expanded into a one-dimensional vector X, the mean of the one-dimensional vector X is calculated, and the mean is subtracted from each pixel value in the one-dimensional vector X, and the covariance corresponding to each pixel value is calculated to generate a covariance matrix.

[0028] It should be added that expanding the pixel matrix into a one-dimensional vector X refers to converting the pixel matrix into a one-dimensional array, in which each pixel value is arranged in the form of columns or rows. Calculating the covariance is a prior art and will not be described in detail in this embodiment.

[0029] Methods for eigendecomposition of the covariance matrix include: ; In the formula, Represented as the covariance matrix, is the initial eigenvector, is the eigenvalue corresponding to the initial eigenvector, is the identity matrix, is the determinant function of the matrix.

[0030] It should be noted that the matrix Expand it and get a polynomial about λ, called the characteristic polynomial. Solve this polynomial equation to get all the eigenvalues, for example, get Q eigenvalues, and then Solve to obtain the initial eigenvector corresponding to the eigenvalue.

[0031] Methods for screening Q initial eigenvectors according to eigenvalues ​​include: The Q initial feature vectors are traversed, and the initial feature vectors whose feature values ​​are less than the preset feature threshold are eliminated to obtain S target feature vectors.

[0032] It is not difficult to understand that the purpose of eliminating initial feature vectors whose eigenvalues ​​are less than a preset feature threshold in this embodiment is to remove features that have little impact on the data. The size of the eigenvalue usually represents the variance or amount of information occupied by the eigenvalue in the data. The smaller the eigenvalue, the smaller the contribution of the direction represented by the eigenvector to the overall data. Therefore, eliminating these eigenvectors can simplify the model and reduce the computational complexity, while retaining the most important features in the data to make subsequent processing more efficient and accurate. It should be added that converting S target feature vectors into a two-dimensional semantic image refers to reconstructing a two-dimensional image space through the selected target feature vectors. The space is based on the eigenvector and projects the key information in the cochlear image onto a two-dimensional plane to generate a semantic image that can reflect the characteristics of the cochlear image.

[0033] The method of performing dimensionality-upgrading processing on a two-dimensional semantic image to obtain a three-dimensional semantic image includes: Based on the preset transposed convolution layer, S two-dimensional semantic images are mapped into three-dimensional space, and the mapped two-dimensional semantic images are feature spliced ​​to obtain a three-dimensional semantic image.

[0034] Generally, the transposed convolution layer is also called the deconvolution layer, which is widely used in deep learning for tasks such as image generation, upsampling, and feature map expansion. Its core function is to expand a smaller feature map into a larger feature map, and at the same time it can learn the complex features of spatial data. The main function of the transposed convolution layer is to map the feature map from a low-dimensional space to a high-dimensional space by learning back propagation. The above-mentioned feature splicing can be spliced ​​along the spatial dimension, because each two-dimensional semantic image represents a two-dimensional projection of the three-dimensional structure of the cochlea on a certain plane. These two-dimensional semantic images capture the grayscale or color intensity information of different areas of the cochlea, reflecting the tissue density, structural characteristics and other characteristics of the cochlea on this plane. Since each two-dimensional semantic image captures part of the information of the cochlea from different planar perspectives, in the process of dimensionality increase, this information can be fused into the three-dimensional space through feature splicing.

[0035] In this embodiment, the two-dimensional semantic image is first screened through projection dimensionality reduction, and then the three-dimensional semantic image is obtained through dimensionality increase, reconstruction and splicing. The processing of each stage provides more accurate input for the subsequent stage, avoids information loss and redundancy, and enhances the expressive power of the data. This series of combinations ensures that the key information of the cochlear image data can be fully extracted and integrated in different dimensions, providing a reliable foundation for subsequent monitoring and analysis.

[0036] S20: Acquire semantic feature data corresponding to the two-dimensional semantic image, determine corresponding attention parameters according to the semantic feature data, and fuse the attention parameters and the three-dimensional semantic image to obtain a target semantic image; In one embodiment of the present invention, the semantic feature data includes edge strength values, circularity values, and grayscale peak values. It can be understood that the attention parameter reflects the degree of attention of the subsequent model to different areas. For example, the edge areas in the cochlear image, areas with obvious shape features, or areas with large grayscale changes are more important for the characterization of the cochlear structure. Therefore, the attention parameter will pay greater attention to these areas. The two-dimensional semantic image can be pre-divided into several sub-area images, and the semantic feature data corresponding to each sub-area image is obtained. Taking the circularity value as an example, for a sub-area with a higher circularity value, it means that the sub-area has a more regular shape and is more important for the characterization of the cochlear structure. Therefore, higher attention parameters can be assigned to these sub-areas.

[0037] The method of determining the corresponding attention parameters according to the semantic feature data includes: Based on the semantic feature data, standard feature data with the greatest similarity is determined from a pre-constructed database, and standard parameters associated with the standard feature data are used as attention parameters corresponding to the semantic feature data.

[0038] It is not difficult to understand that those skilled in the art can pre-construct several standard feature data based on expert experience. For example, semantic feature data include edge strength values, circularity values ​​and grayscale peak values. Similarly, standard feature data include edge strength values, circularity values ​​and grayscale peak values. By calculating the similarity between semantic feature data and standard feature data, the most similar standard feature data can be determined. The method for calculating the similarity can be achieved by calculating the Euclidean distance or the cosine value, which will not be elaborated in this embodiment.

[0039] In this embodiment, the attention parameter and the three-dimensional semantic image are fused by adjusting the brightness value or transparency value of each sub-region in the three-dimensional semantic image according to the size of the attention parameter. It can be understood that, from the above content, a number of two-dimensional semantic images are mapped and feature-stitched to form a three-dimensional semantic image. Then, each sub-region in the three-dimensional semantic image corresponds to each two-dimensional semantic image. Because the brightness value or transparency value of each sub-region can be adjusted by the size of the attention parameter, the importance of each sub-region in the three-dimensional space can be more clearly highlighted. Regions with larger attention parameters will show higher brightness or lower transparency, making these key regions more prominent in the three-dimensional semantic image, so that the user classification model can more intuitively focus on important structures in the cochlear image data.

[0040] S30: Under M groups of noise parameter environments, obtain M groups of noise response data of the users to be screened, traverse the M groups of noise response data to calculate corresponding noise response coefficients, and determine the noise impact value according to the M noise response coefficients; In one embodiment of the present invention, in an M-group noise parameter environment, it means that noise reaction tests are performed on users to be screened using M groups of different noise parameters. The M-group noise parameters may be 4&12.5kHz, 6&12.5kHz, 4&10kHz, 6&10kHz, 3&12.5kHz, etc. Taking 4&12.5kHz as an example, 4 represents a frequency of 4kHz, and 12.5 represents a frequency of 12.5kHz. 4&12.5kHz is a combination of two frequency parameters. The noise parameters of this frequency combination can be used to observe the noise reaction of users to be screened under a specific frequency combination, thereby evaluating the impact of noise on susceptibility to deafness at different frequencies.

[0041] Among them, the noise response data includes the number of brain wave cycles, average reaction time, and average blood pressure value. The number of brain wave cycles refers to the number of complete oscillations of brain waves during the test period. This value reflects the activity frequency characteristics of the brain in a noisy environment. The larger the number of brain wave cycles, the higher the activity frequency of the brain in a noisy environment, reflecting stronger neural activity. This means that the brain reacts more significantly to noise stimulation and is in a higher state of excitement or stress, indicating that the user is more sensitive to the current noise environment. Therefore, the probability that the user to be screened is a susceptible category user is greater. The more the average reaction time deviates from the standard reaction time, it indicates that the user's cognitive or behavioral response in a noisy environment is affected. This deviation is due to the noise interfering with attention, information processing speed or decision-making ability.

[0042] The method of traversing M groups of noise response data to calculate the corresponding noise response coefficient includes: calculating the corresponding noise response coefficient according to the number of brain wave cycles, the average response time and the average blood pressure value;

[0043] Specifically include: NRC= ; Where NRC is the noise response coefficient, is the average blood pressure value, is the number of brain wave cycles, is the average reaction time, is the standard reaction time, is the hyperbolic cosine function, is a logarithmic function with base e, is the inverse cotangent function, It is an exponential function with e as base, where e is a natural constant.

[0044] In this embodiment, the average blood pressure value is taken as an example. When the average blood pressure value is larger, it indicates that the physiological stress response of the user to be screened in a noisy environment is stronger, which means that the user to be screened is more sensitive to the noise environment, and the noise causes a greater physiological burden on him. Therefore, the probability of him being a susceptible category user may also increase. Therefore, when the noise response coefficient is larger, it indicates that the probability that the user to be screened is a susceptible category user is greater.

[0045] The method for determining the noise impact value based on M noise response coefficients includes:

[0046] The occurrence frequency values ​​corresponding to the M groups of noise parameters are obtained, the corresponding weight factors are determined according to the occurrence frequency values ​​of each group of noise parameters, and the noise impact values ​​are calculated according to the weight factors and the M noise response coefficients.

[0047] It should be noted that the occurrence frequency value refers to the frequency value of each group of noise parameters in the working environment of the user to be screened. For example, M groups of noise parameters can be 4&12.5kHz, 6&12.5kHz, and 4&10kHz. Then the occurrence frequency value corresponding to 4&12.5kHz is 8, the occurrence frequency value corresponding to 6&12.5kHz is 7, and the occurrence frequency value corresponding to 4&10kHz is 5. Then the corresponding weight factors can be 0.4, 0.35, and 0.25. The noise impact value can be obtained by summing the weight factors and the corresponding noise response coefficients.

[0048] In this embodiment, noise response data is first collected under multiple sets of noise parameter environments, and the noise response coefficient is calculated one by one by traversing the data to ensure that the response under each noise condition is accurately quantified. After the noise response coefficient is calculated, a comprehensive evaluation is performed based on the frequency of occurrence of each noise parameter in the user's working environment. The noise impact value finally generated can more accurately reflect the user's susceptibility in a real environment.

[0049] S40: inputting the target semantic image and the noise impact value into a pre-built user classification model to obtain a user category, where the user category at least includes a susceptible category and a resistant category; The construction method of the user classification model includes: Obtain P groups of training data, where P is a positive integer greater than 1, and the training data includes historical target semantic images, historical noise impact values, and historical user categories; The historical target semantic images, historical noise impact values ​​and historical user categories are used as sample sets, and the sample sets are divided into training sets and test sets to build a classifier; The historical target semantic images and historical noise impact values ​​in the training set are used as input data, and the historical user categories in the training set are used as output data to train the classifier and obtain an initial classifier; The initial classifier is tested using the test set, and a classifier that meets the preset accuracy is output as the user classification model. The classifier is preferably a deep neural network model.

[0050] It should be added that the user classification model in this embodiment comprehensively evaluates the target semantic image and the noise impact value to output the user category. It can be seen from the above content that the larger the noise response coefficient is, the greater the probability that the user to be screened is a susceptible category user. Similarly, the larger the noise impact value is, the greater the probability that the user category is a susceptible category. Since the cochlear image data is a three-dimensional structural image of the cochlea, and the target semantic image is generated based on the cochlear image data after dimensionality reduction and dimensionality increase, the target semantic image can reflect the key structural features and tissue density changes of the cochlea. The target semantic image can reflect the tissue density distribution in different areas of the cochlea, which is helpful to analyze the structural integrity of the cochlea or whether there are pathological changes. It is understandable that for susceptible category users, the target semantic image may show abnormal distribution of tissue density or changes in structural characteristics in certain areas of the cochlea, such as reduced density or irregular structure. Therefore, the user classification model in this embodiment can classify the users to be screened according to the target semantic image.

[0051] In this embodiment, the cochlear image data is first subjected to dimensionality reduction processing to obtain a two-dimensional semantic image, and the two-dimensional semantic image is subjected to dimensionality increase processing to obtain a three-dimensional semantic image. Then, corresponding attention parameters are determined according to the semantic feature data, and the attention parameters and the three-dimensional semantic image are fused to obtain a target semantic image. M groups of noise response data are traversed to calculate the corresponding noise response coefficients, and the noise impact value is determined according to the M noise response coefficients. Finally, the target semantic image and the noise impact value are input into a pre-built user classification model to obtain the user category. This embodiment fully considers the differences in individual susceptibility to noise exposure, and combines the three-dimensional semantic image and the noise impact value to accurately evaluate the susceptibility, so as to accurately distinguish the user category of the users to be screened.

[0052] Example 2

[0053] See also Figure 2 As shown, based on the same inventive concept, this embodiment discloses a system for screening people susceptible to noise-induced hearing loss. For details not provided in this embodiment, please refer to the description of the relevant parts in Embodiment 1. The system includes: Image processing module: used to obtain the cochlear image data of the user to be screened, perform dimensionality reduction processing on the cochlear image data to obtain a two-dimensional semantic image, and perform dimensionality increase processing on the two-dimensional semantic image to obtain a three-dimensional semantic image; In this embodiment, the users to be screened refer to users who need to be monitored for hearing loss. The cochlear image data may be a three-dimensional structural image of the cochlea obtained through MRI technology. MRI technology is a non-invasive medical imaging technology that uses strong magnetic fields and radio waves to generate detailed images of internal structures. Specifically, MRI aligns hydrogen atoms in the human body through a magnetic field, and then uses radio frequency pulses to make these hydrogen atoms emit signals, and finally processes these signals through a computer to generate high-resolution images.

[0054] The method of performing dimensionality reduction processing on cochlear image data to obtain a two-dimensional semantic image includes: The cochlear image data is projected to obtain W two-dimensional slices, the pixel matrix corresponding to each two-dimensional slice is obtained, the covariance matrix corresponding to the pixel matrix is ​​calculated, the covariance matrix is ​​eigendecomposed to obtain Q initial eigenvectors and corresponding eigenvalues, the Q initial eigenvectors are screened according to the eigenvalues ​​to obtain S target eigenvectors, and the S target eigenvectors are converted into a two-dimensional semantic image.

[0055] Methods for calculating the covariance matrix corresponding to the pixel matrix include: The pixel matrix is ​​expanded into a one-dimensional vector X, the mean of the one-dimensional vector X is calculated, and the mean is subtracted from each pixel value in the one-dimensional vector X, and the covariance corresponding to each pixel value is calculated to generate a covariance matrix.

[0056] Methods for eigendecomposition of the covariance matrix include: ; In the formula, Represented as the covariance matrix, is the initial eigenvector, is the eigenvalue corresponding to the initial eigenvector, is the identity matrix, is the determinant function of the matrix.

[0057] Methods for screening Q initial eigenvectors according to eigenvalues ​​include: The Q initial feature vectors are traversed, and the initial feature vectors whose feature values ​​are less than the preset feature threshold are eliminated to obtain S target feature vectors.

[0058] The method of performing dimensionality-upgrading processing on a two-dimensional semantic image to obtain a three-dimensional semantic image includes: Based on the preset transposed convolution layer, S two-dimensional semantic images are mapped into three-dimensional space, and the mapped two-dimensional semantic images are feature spliced ​​to obtain a three-dimensional semantic image.

[0059] Data fusion module: used to obtain the semantic feature data corresponding to the two-dimensional semantic image, determine the corresponding attention parameter according to the semantic feature data, and fuse the attention parameter and the three-dimensional semantic image to obtain the target semantic image; The method of determining the corresponding attention parameters according to the semantic feature data includes: Based on the semantic feature data, standard feature data with the greatest similarity is determined from a pre-constructed database, and standard parameters associated with the standard feature data are used as attention parameters corresponding to the semantic feature data.

[0060] Data generation module: used to obtain M groups of noise response data of the users to be screened under M groups of noise parameter environments, traverse the M groups of noise response data to calculate the corresponding noise response coefficients, and determine the noise impact value according to the M noise response coefficients; The method for determining the noise impact value based on M noise response coefficients includes: The occurrence frequency values ​​corresponding to the M groups of noise parameters are obtained, the corresponding weight factors are determined according to the occurrence frequency values ​​of each group of noise parameters, and the noise impact values ​​are calculated according to the weight factors and the M noise response coefficients.

[0061] User classification module: used to input the target semantic image and noise impact value into the pre-built user classification model to obtain user categories, and the user categories include at least susceptible categories and resistant categories.

[0062] Example 3

[0063] Based on Example 1, a method for performing basement membrane curvature correction processing on the first image data to obtain the second image data includes: The central spiral line of the first image data is extracted using a skeleton extraction algorithm, and the central spiral line is converted into the frequency domain through Fourier transform to obtain the basilar membrane unfolding line. The first image data with the basilar membrane unfolding line is used as the second image data, wherein the central spiral line represents the axis line of the cochlea.

[0064] In this embodiment, the skeleton extraction algorithm refers to a morphological image processing algorithm, which is used to extract the central spiral line of the cochlea, that is, the axis line of the basilar membrane, from the cochlear image. The core purpose of the algorithm is to simplify the structure of the cochlea and retain its most basic morphological features, so as to provide an accurate benchmark for subsequent basilar membrane curvature correction and unfolding processing. The purpose of the skeleton extraction algorithm is to simplify the structure of the object in the image into a skeleton with a single pixel width. For the cochlear image, skeletonization will extract the center line of the cochlea, that is, its axis line, which is usually the path where the basilar membrane of the cochlea is located.

[0065] The above-mentioned middle basilar membrane unfolding line is achieved by unfolding the spiral morphology of the cochlea into a standardized straight line through Fourier transform and coordinate mapping. It is used to represent the axis line of the cochlear basilar membrane to ensure uniform distribution of information during subsequent dimensionality reduction and avoid information loss due to the spiral structure. The center line of the cochlea is extracted through Fourier transform and skeleton extraction algorithm, and the spiral structure of the cochlea is unfolded to make information in all regions evenly distributed, ensuring that the data after dimensionality reduction retains complete cochlear characteristics.

[0066] The method of converting the central spiral line into the frequency domain by Fourier transform to obtain the basement membrane expansion line includes: The central spiral line is converted from a Cartesian coordinate system to a polar coordinate system, and the polar coordinates of the central spiral line are subjected to Fourier transform to obtain frequency components, which are adjusted by low-pass filtering until the central spiral line changes from a spiral structure to a straight structure, thereby obtaining the basement membrane unfolding line.

[0067] Methods for adjusting the frequency components by low-pass filtering until the central spiral line changes from a spiral structure to a straight structure include: The total curvature of the central spiral line is calculated and the real-time proportion of the high-frequency component in the frequency component is obtained. When the total curvature is less than the preset curvature threshold and the real-time proportion is less than the preset proportion threshold, it is determined that the central spiral line changes from a spiral structure to a straight structure.

[0068] In this embodiment, Fourier transform is performed on the polar coordinates of the central spiral line to obtain frequency components, which means converting the polar coordinate representation of the central spiral line into the frequency domain to obtain its component information at different frequencies, that is, analyzing the periodic characteristics of the spiral structure. By analyzing these frequency components, the geometric characteristics of the spiral structure can be identified, and the high-frequency and low-frequency components therein can be adjusted so as to unfold the cochlea into a standard straight line shape.

[0069] The total curvature of the above-mentioned central spiral line is used to measure the overall curvature of the central spiral line in the spatial domain, that is, the overall bending degree of the spiral shape. It indicates the degree of deviation of the spiral curve from the straight line. The real-time proportion of the high-frequency component refers to the energy proportion of the high-frequency component in the spectrum after Fourier transform. It can be understood that the smaller the energy proportion of the high-frequency component, the smaller the change in the curvature of the central spiral line, and the spiral structure tends to a straight line shape.

[0070] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center via a wired network or a wireless network. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD) or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

[0071] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in the present invention can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0072] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0073] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of the units is only one, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0074] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0075] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0076] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0077] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for screening people susceptible to noise-induced hearing loss, characterized in that: include: Obtaining cochlear image data of the user to be screened, performing dimensionality reduction processing on the cochlear image data to obtain a two-dimensional semantic image, and performing dimensionality increase processing on the two-dimensional semantic image to obtain a three-dimensional semantic image; Acquire semantic feature data corresponding to the two-dimensional semantic image, determine corresponding attention parameters according to the semantic feature data, fuse the attention parameters and the three-dimensional semantic image to obtain a target semantic image; Under M groups of noise parameter environments, obtain M groups of noise response data of the users to be screened, traverse the M groups of noise response data to calculate corresponding noise response coefficients, and determine the noise impact value according to the M noise response coefficients; Input the target semantic image and the noise impact value into a pre-built user classification model to obtain user categories, where the user categories at least include susceptible categories and resistant categories; Before the dimension reduction process is performed on the cochlear image data, the following steps are also included: The cochlear image data is sequentially subjected to wavelet decomposition, guided filtering enhancement and fusion processing to generate first image data, the first image data is subjected to basilar membrane curvature correction processing to obtain second image data, and the cochlear image data is replaced by the second image data.

2. A method for screening people susceptible to noise-induced hearing loss according to claim 1, characterized in that: The method of performing dimensionality reduction processing on the cochlear image data to obtain a two-dimensional semantic image comprises: The cochlear image data is projected to obtain W two-dimensional slices, the pixel matrix corresponding to each two-dimensional slice is obtained, the covariance matrix corresponding to the pixel matrix is ​​calculated, the covariance matrix is ​​eigendecomposed to obtain Q initial eigenvectors and corresponding eigenvalues, the Q initial eigenvectors are screened according to the eigenvalues ​​to obtain S target eigenvectors, and the S target eigenvectors are converted into a two-dimensional semantic image.

3. A method for screening people susceptible to noise-induced hearing loss according to claim 2, characterized in that: The method for calculating the covariance matrix corresponding to the pixel matrix includes: The pixel matrix is ​​expanded into a one-dimensional vector X, the mean of the one-dimensional vector X is calculated, and the mean is subtracted from each pixel value in the one-dimensional vector X, and the covariance corresponding to each pixel value is calculated to generate a covariance matrix.

4. A method for screening people susceptible to noise-induced hearing loss according to claim 3, characterized in that: The method of performing dimensionality-upgrading processing on a two-dimensional semantic image to obtain a three-dimensional semantic image comprises: Based on the preset transposed convolutional layer, S two-dimensional semantic images are mapped into three-dimensional space, and the mapped two-dimensional semantic images are feature spliced ​​to obtain a three-dimensional semantic image.

5. The method for screening a population susceptible to noise-induced hearing loss according to claim 1, characterized in that: The method for determining corresponding attention parameters according to semantic feature data comprises: Based on the semantic feature data, standard feature data with the greatest similarity is determined from a pre-constructed database, and standard parameters associated with the standard feature data are used as attention parameters corresponding to the semantic feature data.

6. The method for screening a population susceptible to noise-induced hearing loss according to claim 1, characterized in that: The noise response data includes the number of brain wave cycles, average reaction time, and average blood pressure value. The method of traversing M groups of noise response data to calculate the corresponding noise response coefficient includes: calculating the corresponding noise response coefficient according to the number of brain wave cycles, average reaction time, and average blood pressure value.

7. A method for screening people susceptible to noise-induced hearing loss according to claim 6, characterized in that: The method for determining the noise impact value according to M noise response coefficients includes: The occurrence frequency values ​​corresponding to the M groups of noise parameters are obtained, the corresponding weight factors are determined according to the occurrence frequency values ​​of each group of noise parameters, and the noise impact values ​​are calculated according to the weight factors and the M noise response coefficients.

8. The method for screening a population susceptible to noise-induced hearing loss according to claim 1, characterized in that: The method of sequentially performing wavelet decomposition, guided filtering enhancement and fusion processing on the cochlear image data to generate the first image data comprises: The cochlear image data is decomposed by using wavelet transform, low-frequency components and high-frequency components are extracted, guided filtering is performed on the high-frequency components to obtain enhanced high-frequency components, and the enhanced high-frequency components are fused with the low-frequency components to obtain the first image data.

9. A method for screening a population susceptible to noise-induced hearing loss according to claim 8, characterized in that: The method of performing basement membrane curvature correction processing on the first image data to obtain the second image data comprises: The central spiral line of the first image data is extracted using a skeleton extraction algorithm, and the central spiral line is converted into the frequency domain through Fourier transform to obtain the basilar membrane unfolding line. The first image data with the basilar membrane unfolding line is used as the second image data, wherein the central spiral line represents the axis line of the cochlea.

10. A system for screening people susceptible to noise-induced hearing loss, used to implement a method for screening people susceptible to noise-induced hearing loss according to any one of claims 1 to 9, characterized in that: include: Image processing module: used to obtain cochlear image data of the user to be screened, perform dimensionality reduction processing on the cochlear image data to obtain a two-dimensional semantic image, and perform dimensionality increase processing on the two-dimensional semantic image to obtain a three-dimensional semantic image; Data fusion module: used to obtain the semantic feature data corresponding to the two-dimensional semantic image, determine the corresponding attention parameter according to the semantic feature data, and fuse the attention parameter and the three-dimensional semantic image to obtain the target semantic image; Data generation module: used to obtain M groups of noise response data of the users to be screened under M groups of noise parameter environments, traverse the M groups of noise response data to calculate the corresponding noise response coefficients, and determine the noise impact value according to the M noise response coefficients; User classification module: used to input the target semantic image and noise impact value into the pre-built user classification model to obtain user categories, where the user categories include at least susceptible categories and resistant categories; Before the dimension reduction process is performed on the cochlear image data, the following steps are also included: The cochlear image data is sequentially subjected to wavelet decomposition, guided filtering enhancement and fusion processing to generate first image data, the first image data is subjected to basilar membrane curvature correction processing to obtain second image data, and the cochlear image data is replaced by the second image data.

Images