CBCT image-based periodontal problem data positioning method

Abstract

The invention provides a periodontal problem data positioning method based on a CBCT image. The method comprises the following steps: acquiring a CBCT original image of an oral cavity area of a target object, then filtering an original frequency domain image corresponding to the CBCT original image by using a high-pass filter to generate a filtered frequency domain image, generating a CBCT processed image according to the filtered frequency domain image, and then enhancing the CBCT processed image to generate a CBCT enhanced image. The CBCT processing image and the CBCT enhanced image are subjected to fusion processing to generate the to-be-recognized image, and then the periodontal problem position information is determined according to the to-be-recognized image, so that the influence of artifacts caused by the metal implant on the image quality is effectively avoided, and the image quality is improved. And accurate positioning of the periodontal problem position of the target object of which the oral cavity area is provided with the metal implant is realized.

Description

Technical Field

[0001] The present application relates to data processing technology, and in particular to a method for locating periodontal problem data based on CBCT images. Background Art

[0002] The diagnosis and treatment of periodontal disease has always been a crucial topic in the field of dentistry. Periodontal disease not only affects a patient's oral health but can also negatively impact overall health. Therefore, accurate and timely localization of periodontal problems is crucial for both prevention and treatment.

[0003] Cone Beam Computed Tomography (CBCT), an advanced medical imaging technology, has been widely used in dentistry due to its high resolution, three-dimensional imaging capabilities, and low radiation dose. CBCT images provide detailed structural information of the oral cavity, including teeth, gums, alveolar bone, and periodontal ligament, providing a powerful basis for the diagnosis of periodontal disease.

[0004] However, in practice, the quality of CBCT images is often affected by a variety of factors. In particular, when metal implants are present in the oral cavity, X-ray absorption and scattering by the metal can produce artifacts in CBCT images. These artifacts can interfere with image display and make it difficult to locate periodontal problems. Traditional image processing methods often struggle to effectively remove these artifacts, hindering accurate location of periodontal problems. Summary of the Invention

[0005] The present application provides a periodontal problem data positioning method based on CBCT images, which is used to effectively avoid the influence of artifacts caused by metal implants on image quality and realize accurate positioning of periodontal problem positions for target objects equipped with metal implants in the oral area.

[0006] In a first aspect, the present application provides a method for locating periodontal problem data based on CBCT images, comprising:

[0007] Acquiring a CBCT original image of an oral region of a target subject, wherein the oral region is equipped with a metal implant, and wherein the CBCT original image contains artifact components associated with the metal implant;

[0008] performing filtering processing on an original frequency domain image corresponding to the original CBCT image through a high-pass filter to generate a filtered frequency domain image, and generating a CBCT processed image based on the filtered frequency domain image, wherein a cutoff frequency of the high-pass filter is associated with the artifact component;

[0009] performing enhancement processing on the CBCT processed image to generate a CBCT enhanced image, and performing fusion processing on the CBCT processed image and the CBCT enhanced image to generate an image to be recognized;

[0010] The periodontal problem location information is determined according to the image to be identified.

[0011] In the above scheme, high-pass filtering is performed by setting a cutoff frequency corresponding to the metal implant artifact component, effectively reducing the low-frequency artifact components (such as scattering artifacts and beam hardening artifacts) in the original CBCT image. The cutoff frequency here is directly related to the artifact distribution characteristics, so that the filtering process significantly reduces the impact of metal artifacts on image quality while retaining high-frequency anatomical structure information. The frequency domain processing method based on two-dimensional Fourier transform and its inverse transform can accurately separate and filter out artifact-related frequency domain components, generating a CBCT processed image with artifact weakened in the spatial domain. This process retains the spatial frequency characteristics of the gingival tissue edge and fine structure, providing a basis for subsequent enhancement processing. Next, the CBCT processed image is enhanced, and the complementary information of the processed image and the enhanced image is fused. Finally, the periodontal problem location information is determined based on the fused image to be identified, so as to accurately locate the periodontal problem location for the target object with metal implants in the oral area.

[0012] Optionally, filtering the original frequency domain image corresponding to the original CBCT image by a high-pass filter to generate a filtered frequency domain image, and generating a CBCT processed image based on the filtered frequency domain image includes:

[0013] Converting the CBCT original image from the spatial domain to the frequency domain using a two-dimensional Fourier transform to generate the original frequency domain image;

[0014] filtering the frequency domain elements in the original frequency domain image that are lower than the cutoff frequency according to the high-pass filter to generate the filtered frequency domain image;

[0015] The filtered frequency domain image is converted from the frequency domain to the spatial domain using a two-dimensional inverse Fourier transform to generate the CBCT processed image.

[0016] In the above scheme, the raw CBCT image is converted to the frequency domain via a two-dimensional Fourier transform, achieving decoupling of the frequency components of the spatial domain information. This operation causes large-area low-frequency artifacts generated by metal implants (such as low-frequency banding and global beam hardening artifacts) to be concentrated in the frequency domain, providing a spectral resolution basis for frequency-domain selective filtering. A high-pass filter with an adjustable cutoff frequency is applied in the frequency domain to specifically filter out frequency domain components below the cutoff frequency. These low-frequency components correspond to the primary energy distribution area of metal artifacts. The filtering operation effectively suppresses artifacts while retaining effective high-frequency information representing periodontal biological tissue (including high-frequency details of tooth margins, alveolar bone microstructure, and gingival texture). The filtered frequency domain image is reconstructed into a processed CBCT image in the spatial domain via a two-dimensional inverse Fourier transform. This process maintains the geometric consistency of important anatomical structures while removing low-frequency artifacts, ensuring that the processed image reduces artifact interference while maintaining the spatial correlation of the original grayscale distribution.

[0017] Optionally, the enhancing the CBCT processed image to generate a CBCT enhanced image, and fusing the CBCT processed image and the CBCT enhanced image to generate an image to be identified, includes:

[0018] performing enhancement processing on the CBCT processed image according to a multi-layer complex wavelet transform and a noise suppression threshold to generate the CBCT enhanced image;

[0019] The image to be identified is generated according to the CBCT processed image and the CBCT enhanced image.

[0020] In this solution, a multi-layer decomposition framework based on complex wavelet transforms separates high-frequency image details from low-frequency background at multiple scales. By adjusting the gain coefficient layer by layer and combining it with a dynamic noise suppression threshold, the following technical effects are achieved:

[0021] Direction-sensitive feature enhancement: The multi-directional decomposition characteristics of the complex wavelet transform in the complex domain accurately capture the multi-directional gradient information of periodontal features such as the gingival margin and alveolar bone microfractures;

[0022] Inter-layer adaptive enhancement: Dynamically attenuates the gain coefficient according to the number of decomposition layers to avoid excessive noise in high-frequency layers. It also uses nonlinear gain compensation based on the amplitude matrix to distinguish the energy difference between valid signals and noise.

[0023] Noise suppression threshold control: Dynamically calculate the layer-by-layer threshold through the statistical characteristics of CBCT image processing, while enhancing weak tissue textures and suppressing the amplification effect of high-frequency noise in the wavelet domain.

[0024] Optionally, before performing enhancement processing on the CBCT processed image according to the multi-layer complex wavelet transform and the noise suppression threshold to generate the CBCT enhanced image, the method further includes:

[0025] The noise suppression threshold is determined according to the CBCT processed image and a preset layer attenuation factor.

[0026] In the above scheme, a preset layer attenuation factor causes the noise suppression threshold to decay layer by layer as the number of complex wavelet decomposition layers increases. This design is more adaptable to the multi-scale characteristics of the wavelet transform. Low-decomposition layers primarily contain high-frequency noise and details, requiring a higher threshold to suppress noise. High-decomposition layers retain low-frequency, weak signals, requiring a lower threshold to avoid excessive truncation of valid signals, thereby achieving gradient adjustment of noise suppression intensity across layers. The high threshold of low-decomposition layers focuses on suppressing high-frequency noise, while the low threshold of high-decomposition layers preserves potential weak tissue texture. Compared with fixed thresholds or single statistic methods, this strategy ensures that the noise suppression intensity of each decomposition layer in the wavelet domain matches the signal energy distribution characteristics, suppressing speckle noise while reducing the loss of valid high-frequency components.

[0027] Optionally, before performing enhancement processing on the CBCT processed image according to the multi-layer complex wavelet transform and the noise suppression threshold to generate the CBCT enhanced image, the method further includes:

[0028] The gain coefficient of the corresponding layer of the complex wavelet transform is determined according to the preset gain intensity coefficient, the preset layer attenuation coefficient and the amplitude matrix of the complex wavelet transform coefficient of each layer. The gain coefficient is used to configure the enhancement degree during the enhancement processing of the CBCT enhanced image and is positively correlated with the enhancement degree.

[0029] In the above scheme, the gain coefficient is attenuated using a preset layer attenuation coefficient, and a control logic is established in which the gain intensity exponentially decays with the increasing number of wavelet decomposition layers, wherein a larger gain coefficient enhances high-frequency details, and the attenuated gain coefficient avoids excessive enhancement of low-frequency artifact residues and background noise, thereby maintaining the overall grayscale fidelity of the image. This mechanism always ensures that the enhancement intensity of different decomposition layers is adapted to the frequency band characteristics represented by the wavelet coefficients. Among them, corresponding to significant tissue structures (such as the edge of the alveolar ridge), the enhancement amplitude is automatically suppressed by the denominator to avoid oversaturation artifacts; corresponding to weak signal or noise-dominated areas, the expression of potential pathological characteristics can be enhanced by increasing the proportion of effective signals.

[0030] Optionally, determining the periodontal problem location information according to the image to be identified includes:

[0031] Performing grayscale processing on the image to be identified, determining a gingival boundary point set according to the grayscale gradient, and performing curve fitting on the gingival boundary point set to determine a gingival line curve;

[0032] A characteristic segment is determined according to the gum line curve, and a corresponding periodontal problem area is determined in the image to be identified according to the characteristic segment, so as to mark and display the periodontal problem area in the image to be identified.

[0033] In the above scheme, the grayscale gradient operator is used to extract edge features of the image to be identified. The grayscale distribution and gradient intensity of the image are used to determine the transition zone characteristics between the gingival tissue and the tooth body, and generate a set of candidate gingival boundary points. Compared with the traditional threshold segmentation method, the gradient response operator has higher orientation sensitivity to weak contrast areas such as the periodontal pocket edge and the gingival recession line, which effectively reduces the missed detection rate of soft tissue boundary detection. Parametric curve fitting is performed on the discrete boundary point set to construct a continuous and smooth gingival line curve model. This process suppresses the interference of isolated noise points and local abnormal points on the integrity of the boundary through mathematical optimization, compensates for the boundary breakage problem caused by local image blur or residual metal artifacts, and improves the accuracy of the description of the gingival line morphological characteristics.

[0034] Optionally, determining a characteristic segment according to the gum line curve, and determining a corresponding periodontal problem area in the image to be identified according to the characteristic segment, includes:

[0035] forming a reference curve according to at least two lowest boundary points or two highest boundary points in the gingival boundary point set;

[0036] Determine each reference feature area according to the reference curve and the gum line curve, wherein the reference feature area is a closed area formed by the reference curve and the gum line curve;

[0037] If the maximum width of the closed area is greater than a preset width threshold and / or the area of the closed area is greater than a preset area threshold, the closed area is determined to be the periodontal problem area.

[0038] In the above scheme, the lowest or highest boundary point in the gingival boundary point set is selected to generate a reference curve, and the extreme point of the point set is used as a reference anchor mark. This method uses the inherent geometric topological features around the teeth to establish a stable reference benchmark, avoids the subjective deviation of manual annotation, and ensures the consistency of the measurement benchmark. Then, a closed area is generated by spatially enclosing the reference curve and the gingival line curve, and the maximum horizontal width of the closed area is determined to identify areas beyond normal gingival leakage or ulcer areas (such as horizontal expansion of periodontal pockets). In addition, the pixel area of the closed area can be counted to detect abnormal volume defects caused by bone absorption or gingival atrophy, effectively improving the objectivity and repeatability of lesion detection. Furthermore, through extreme benchmark positioning, multi-parameter quantification of geometric features and combined criteria decision-making, the accuracy of periodontal problem area detection is improved, and objective identification and spatial positioning of pathological changes are achieved in the interference environment of metal implants.

[0039] In a second aspect, the present application provides a periodontal problem data positioning device based on CBCT images, comprising:

[0040] an acquisition module, configured to acquire a CBCT original image of an oral region of a target subject, wherein the oral region is equipped with a metal implant, and wherein the CBCT original image contains an artifact component associated with the metal implant;

[0041] a processing module, configured to filter an original frequency domain image corresponding to the original CBCT image using a high-pass filter to generate a filtered frequency domain image, and generate a processed CBCT image based on the filtered frequency domain image, wherein a cutoff frequency of the high-pass filter is associated with the artifact component;

[0042] The processing module is further configured to perform enhancement processing on the CBCT processed image to generate a CBCT enhanced image, and perform fusion processing on the CBCT processed image and the CBCT enhanced image to generate an image to be recognized;

[0043] The processing module is further configured to determine periodontal problem location information based on the image to be identified.

[0044] Optionally, the processing module is specifically configured to:

[0045] Converting the CBCT original image from the spatial domain to the frequency domain using a two-dimensional Fourier transform to generate the original frequency domain image;

[0046] filtering the frequency domain elements in the original frequency domain image that are lower than the cutoff frequency according to the high-pass filter to generate the filtered frequency domain image;

[0047] The filtered frequency domain image is converted from the frequency domain to the spatial domain using a two-dimensional inverse Fourier transform to generate the CBCT processed image.

[0048] Optionally, the processing module is specifically configured to:

[0049] performing enhancement processing on the CBCT processed image according to a multi-layer complex wavelet transform and a noise suppression threshold to generate the CBCT enhanced image;

[0050] The image to be identified is generated according to the CBCT processed image and the CBCT enhanced image.

[0051] Optionally, the processing module is specifically configured to:

[0052] The noise suppression threshold is determined according to the CBCT processed image and a preset layer attenuation factor.

[0053] Optionally, the processing module is specifically configured to:

[0054] The gain coefficient of the corresponding layer of the complex wavelet transform is determined according to the preset gain intensity coefficient, the preset layer attenuation coefficient and the amplitude matrix of the complex wavelet transform coefficient of each layer. The gain coefficient is used to configure the enhancement degree during the enhancement processing of the CBCT enhanced image and is positively correlated with the enhancement degree.

[0055] Optionally, the processing module is specifically configured to:

[0056] Performing grayscale processing on the image to be identified, determining a gingival boundary point set according to the grayscale gradient, and performing curve fitting on the gingival boundary point set to determine a gingival line curve;

[0057] A characteristic segment is determined according to the gum line curve, and a corresponding periodontal problem area is determined in the image to be identified according to the characteristic segment, so as to mark and display the periodontal problem area in the image to be identified.

[0058] Optionally, the processing module is specifically configured to:

[0059] forming a reference curve according to at least two lowest boundary points or two highest boundary points in the gingival boundary point set;

[0060] Determine each reference feature area according to the reference curve and the gum line curve, wherein the reference feature area is a closed area formed by the reference curve and the gum line curve;

[0061] If the maximum width of the closed area is greater than a preset width threshold and / or the area of the closed area is greater than a preset area threshold, the closed area is determined to be the periodontal problem area.

[0062] In a third aspect, the present application provides an electronic device, comprising:

[0063] processor; and,

[0064] a memory for storing executable instructions of the processor;

[0065] The processor is configured to perform any possible method described in the first aspect by executing the executable instructions.

[0066] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement any possible method described in the first aspect.

[0067] The present application provides a periodontal problem data positioning method based on CBCT images, which obtains a CBCT original image of the oral area of a target object, then uses a high-pass filter to filter the original frequency domain image corresponding to the CBCT original image to generate a filtered frequency domain image, and generates a CBCT processed image based on the filtered frequency domain image. Subsequently, the CBCT processed image is enhanced to generate a CBCT enhanced image, and the CBCT processed image and the CBCT enhanced image are fused to generate an image to be identified, and then the periodontal problem location information is determined based on the image to be identified, so as to effectively avoid the influence of artifacts caused by metal implants on the image quality, and realize accurate positioning of the periodontal problem position of the target object with metal implants in the oral area. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0069] Figure 1 1 is a flow chart of a method for locating periodontal problem data based on CBCT images according to an exemplary embodiment of the present application;

[0070] Figure 2 1 is a flow chart of a method for locating periodontal problem data based on CBCT images according to another exemplary embodiment of the present application;

[0071] Figure 3 1 is a schematic structural diagram of a periodontal problem data positioning device based on CBCT images according to an exemplary embodiment of the present application;

[0072] Figure 4 It is a structural diagram of an electronic device according to an exemplary embodiment of the present application.

[0073] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0074] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0075] Figure 1 FIG. 1 is a flow chart of a method for locating periodontal problem data based on CBCT images according to an exemplary embodiment of the present application. Figure 1 As shown, the method provided in this embodiment includes:

[0076] S101: Acquire a CBCT original image of the oral cavity area of a target object.

[0077] In this step, a CBCT original image of the oral region of the target object is acquired. The oral region is equipped with a metal implant, and artifact components associated with the metal implant exist in the CBCT original image.

[0078] Specifically, a CBCT scanner was used to scan the patient's oral area. Scanning parameters were set to: voltage 100-120 kVp, current 4-8 mA, voxel size 0.2 mm × 0.2 mm × 0.2 mm, and scan time 20-30 seconds, ensuring coverage of the jaw area containing the metal implant (e.g., within 2 cm around the implant).

[0079] In the DICOM format CBCT original images generated by scanning, beam hardening artifacts (manifested as radial stripes) and scattering artifacts (local grayscale unevenness) exist in the area around the metal implant.

[0080] S102 , filtering the original frequency domain image corresponding to the CBCT original image through a high-pass filter to generate a filtered frequency domain image, and generating a CBCT processed image based on the filtered frequency domain image.

[0081] In this step, the original frequency domain image corresponding to the CBCT original image is filtered by a high-pass filter to generate a filtered frequency domain image, and a CBCT processed image is generated based on the filtered frequency domain image, wherein the cutoff frequency of the high-pass filter is associated with the artifact component.

[0082] The CBCT raw image can be converted from the spatial domain to the frequency domain using a two-dimensional Fourier transform to generate a raw frequency domain image. Frequency domain elements below a cutoff frequency in the raw frequency domain image are then filtered using a high-pass filter to generate a filtered frequency domain image. The filtered frequency domain image is then converted from the frequency domain to the spatial domain using an inverse two-dimensional Fourier transform to generate a processed CBCT image.

[0083] Specifically, a high-pass filter is a filter that allows high-frequency components to pass through while suppressing low-frequency components. In periodontal problem data positioning, the focus is on the edges and details of teeth and periodontal tissues, which usually correspond to high-frequency components. Therefore, we need to use a high-pass filter to remove the low-frequency components in the original frequency domain image. The design of the high-pass filter can be based on a variety of methods, such as an ideal high-pass filter, a Butterworth high-pass filter, a Gaussian high-pass filter, etc. In this embodiment, a Butterworth high-pass filter can be selected because it has a smooth transition between the passband and the stopband, which helps to reduce the ringing effect in the filtering process.

[0084] In this approach, the original CBCT image is converted to the frequency domain via a two-dimensional Fourier transform, decoupling the frequency components of the spatial domain information. This operation allows large-area low-frequency artifacts generated by metal implants (such as low-frequency banding and global beam hardening artifacts) to be concentrated in the frequency domain, providing a spectral resolution basis for frequency-domain selective filtering.

[0085] A high-pass filter with an adjustable cutoff frequency is applied in the frequency domain to specifically remove frequency components below the cutoff frequency. These low-frequency components correspond to the primary energy distribution of metal artifacts. This filtering operation effectively suppresses artifacts while preserving the high-frequency information that characterizes periodontal tissue (including high-frequency details of tooth margins, alveolar bone microstructure, and gingival texture).

[0086] Next, the filtered frequency-domain image is reconstructed into a spatial-domain CBCT processed image using a two-dimensional inverse Fourier transform. This process maintains the geometric consistency of important anatomical structures while removing low-frequency artifacts, ensuring that the processed image reduces artifact interference while maintaining the spatial correlation of the original grayscale distribution.

[0087] The above steps form a closed-loop processing link through a bidirectional conversion framework of spatial domain-frequency domain-spatial domain, achieving adaptive suppression of the physical properties of metal artifact components, and providing high-quality input data with artifact attenuation for subsequent image enhancement and feature extraction.

[0088] S103 , performing enhancement processing on the CBCT processed image to generate a CBCT enhanced image, and performing fusion processing on the CBCT processed image and the CBCT enhanced image to generate an image to be recognized.

[0089] In a first possible implementation, global histogram equalization can be performed on the CBCT image. This redistributes the image's grayscale levels, making the grayscale distribution more uniform and thus enhancing image contrast. Specifically, the number of pixels at each grayscale level in the CBCT image can be counted to obtain a grayscale histogram. Then, based on the grayscale histogram, the cumulative distribution function of each grayscale level is calculated. The cumulative distribution function is then used to map the original grayscale levels to obtain new grayscale levels. Global histogram equalization achieves a preliminary improvement in the contrast of the CBCT image.

[0090] Although global histogram equalization can improve the overall contrast, it may introduce noise or over-enhance certain areas when processing images with large local contrast differences. Therefore, this embodiment further uses adaptive histogram equalization to optimize the image enhancement effect. The image can be divided into multiple small blocks (sub-regions), and histogram equalization is performed on each sub-region separately. This can better handle local contrast differences while maintaining global contrast improvement. Specifically, the CBCT-processed image can be divided into multiple overlapping or non-overlapping sub-regions. Histogram equalization is performed on each sub-region separately. At the sub-region boundary, a bilinear interpolation method is used to smooth the grayscale differences between adjacent sub-regions to avoid block effects. Through adaptive histogram equalization, the local contrast of the CBCT-processed image is further improved, while reducing noise and over-enhancement. After the above two steps, we obtain an enhanced CBCT image, that is, a CBCT enhanced image.

[0091] Next, the CBCT processed image and the CBCT enhanced image are fused to generate an image to be identified. To fully leverage the advantages of the CBCT processed and enhanced images, this embodiment may employ a fusion method based on a Laplacian pyramid. First, a Laplacian pyramid is constructed for each of the CBCT processed and enhanced images. Specifically, the two images may be Gaussian filtered and downsampled to construct a Gaussian pyramid. Then, a Laplacian pyramid is constructed by performing a difference calculation between adjacent layers of the Gaussian pyramid. After obtaining the Laplacian pyramids of the two images, a fusion strategy based on weighted averaging is employed to generate a fused Laplacian pyramid. The coefficients of each layer in the Laplacian pyramid may be weighted averaged according to certain weights. The weights may be determined based on indicators such as local contrast and clarity of the image. The fused Laplacian pyramid is then used to reconstruct the fused image through upsampling and Gaussian filtering.

[0092] A second possible implementation method can also utilize bilateral filtering and contrast-limited adaptive histogram equalization. Specifically, weights can be calculated based on the spatial distance between pixels, with closer distances giving greater weights. Weights can also be calculated based on the grayscale value differences between pixels, with smaller differences giving greater weights. A combined weight is obtained by multiplying the spatial proximity weight and the pixel similarity weight. Using this combined weight, a weighted average of neighboring pixels is performed to obtain the filtered pixel value. Bilateral filtering effectively suppresses noise in CBCT-processed images while preserving edge information.

[0093] The CBCT image is then divided into multiple small blocks (subregions). Histogram equalization is performed on each subregion, but the degree of contrast enhancement is limited. At subregion boundaries, bilinear interpolation is used to smooth grayscale differences between adjacent subregions. This significantly improves the local contrast of the CBCT image while avoiding noise amplification and detail loss. After these two steps, we obtain the enhanced CBCT image, known as the CBCT-enhanced image.

[0094] Next, to fully leverage the advantages of CBCT-processed and CBCT-enhanced images, this embodiment employs a fusion method based on principal component analysis. Specifically, the two images can be concatenated row-by-row or column-by-column to form a two-dimensional matrix. The covariance matrix of the image matrix is then calculated. Eigenvalue decomposition is performed on the covariance matrix to obtain eigenvalues and corresponding eigenvectors. The original image is projected into the space formed by the eigenvectors to obtain the transformed image.

[0095] After obtaining the transformed images, a weighted fusion strategy is employed to generate the image to be identified. Specifically, weights can be calculated based on the eigenvalues or variances of the principal component transformed images. Larger eigenvalues or variances indicate greater information contained in the image, and therefore, greater weights. The calculated weights are then used to perform a weighted average of the two transformed images to generate a fused image. This fused image is then projected back into the original image space to obtain the final image to be identified.

[0096] In a third possible implementation, the CBCT processed image can be enhanced using multi-layer complex wavelet transform and noise suppression threshold. Specifically, the CBCT enhanced image can be determined, which involves parameters such as the layer gain coefficient, the maximum number of decomposition layers of the complex wavelet transform, the complex wavelet transform coefficients (including amplitude and phase information), the gradient operator, the spatial gradient of the complex wavelet coefficients, and the layer noise suppression threshold. Before the enhancement process, the noise suppression threshold must be determined based on the CBCT processed image and the preset layer attenuation factor. At the same time, the gain coefficient of the corresponding layer is determined based on the preset gain intensity coefficient, the preset layer attenuation coefficient, and the amplitude matrix of the complex wavelet transform coefficients of each layer. The image to be identified is generated based on the CBCT processed image and the CBCT enhanced image.

[0097] In the first possible implementation, global histogram equalization can initially improve the overall contrast of the image, making details clearer. Adaptive histogram equalization further addresses areas with large local contrast differences, avoiding the noise amplification and over-enhancement issues that global equalization can cause. The Laplacian pyramid fusion method effectively captures image details at different scales. Using a weighted average fusion strategy, the fused image retains overall structural information while enhancing detail.

[0098] In the second possible implementation, bilateral filtering can remove noise while preserving image edge information, making it suitable for processing noisy CBCT images. While limiting the extent of contrast enhancement, it effectively improves local image contrast, avoiding noise amplification and detail loss. Data dimensionality reduction and feature extraction are then used to effectively capture the image's key features, generating a high-quality image for recognition using a weighted fusion strategy.

[0099] In the third possible implementation, a multi-layer complex wavelet transform analyzes images at multiple scales, extracting detailed information across different frequency bands. By setting a noise suppression threshold, image noise can be effectively removed, improving image quality. The gain coefficients and fusion weights in this method can be adjusted based on image characteristics, offering significant flexibility and robustness.

[0100] Compared to the first possible implementation, the third possible implementation can analyze images at multiple scales and extract detailed information in different frequency bands. This multi-scale analysis capability enables more refined image enhancement processing and can capture more subtle image features. Although the first implementation combines global histogram equalization with adaptive histogram equalization, it lacks multi-scale analysis capabilities and cannot process the different frequency components in the image as finely as the multi-layer complex wavelet transform. The third implementation, by setting a noise suppression threshold, can effectively remove noise from the image and improve image quality, which is particularly important when processing noisy CBCT images.

[0101] Furthermore, while the first implementation optimizes local contrast through adaptive histogram equalization, it may not be as direct and effective as the third implementation in terms of noise suppression. The gain coefficients and fusion weights in the third implementation can be adjusted based on image characteristics, providing greater flexibility and robustness. This makes this method adaptable to different types of CBCT images, achieving better enhancement results.

[0102] Compared with the second possible implementation method mentioned above, the third possible implementation method can analyze images at multiple scales, and through multi-layer complex wavelet transform and weight fusion, it can enhance image contrast while retaining more detail information.

[0103] Moreover, the third implementation is computationally more efficient than the second. Although multi-layer complex wavelet transform and subsequent fusion processing require certain computing resources, the bilateral filtering and PCA in the second implementation are computationally expensive, especially when processing large-scale CBCT images.

[0104] Furthermore, the third implementation is highly adaptable and robust, capable of processing complex CBCT images, including those containing multiple frequency components, noise, and fine details. While the second implementation can also process complex images to a certain extent, it is not as stable and effective as the third implementation when dealing with extremely complex or noisy images.

[0105] S104: Determine periodontal problem location information based on the image to be recognized.

[0106] In the first implementation, the image to be identified may be grayscale processed, and the gingival boundary point set may be determined based on the grayscale gradient. Curve fitting is performed on the gingival boundary point set to determine the gingival line curve. Feature segments are determined based on the gingival line curve, and the corresponding periodontal problem areas are determined in the image to be identified. Specifically, a reference curve may be formed based on at least two lowest boundary points or highest boundary points in the gingival boundary point set, and then each reference feature area (i.e., closed area) may be determined based on the reference curve and the gingival line curve. If the maximum width value of the closed area is greater than a preset width threshold and / or the area of the closed area is greater than a preset area threshold, the closed area is determined to be a periodontal problem area. The periodontal problem area is marked and displayed in the image to be identified so that the doctor can diagnose and treat it.

[0107] In this embodiment, a CBCT original image of the oral region of the target object is obtained, and then the original frequency domain image corresponding to the CBCT original image is filtered using a high-pass filter to generate a filtered frequency domain image, and a CBCT processed image is generated based on the filtered frequency domain image. Subsequently, the CBCT processed image is enhanced to generate a CBCT enhanced image, and the CBCT processed image and the CBCT enhanced image are fused to generate an image to be identified, and then the position information of the periodontal problem is determined based on the image to be identified, so as to effectively avoid the influence of artifacts caused by metal implants on the image quality, and achieve accurate positioning of the periodontal problem position for the target object with metal implants in the oral region.

[0108] Figure 2 FIG. 1 is a flow chart of a method for locating periodontal problem data based on CBCT images according to another exemplary embodiment of the present application. Figure 2 As shown, the periodontal problem data positioning method based on CBCT images provided in this embodiment includes:

[0109] S201: Acquire a CBCT original image of the oral cavity area of a target object.

[0110] In this step, a CBCT original image of the oral region of the target object is acquired. The oral region is equipped with a metal implant, and artifact components associated with the metal implant exist in the CBCT original image.

[0111] S202 , performing enhancement processing on the CBCT processed image according to multi-layer complex wavelet transform and noise suppression threshold to generate a CBCT enhanced image.

[0112] Specifically, the CBCT processed image I(x,y) may be enhanced using Formula 1 and multi-layer complex wavelet transform and noise suppression threshold to generate a CBCT enhanced image E(x,y), where Formula 1 is:

[0113]

[0114] Where I(x,y) is the CBCT processed image, E(x,y) is the CBCT enhanced image, and λ k is the k-th layer gain coefficient, N is the maximum decomposition layer number of complex wavelet transform, W k (I)(x,y) is the complex wavelet transform coefficient of the kth layer, which includes amplitude and phase information. Re(W k (I)(x,y)) is the real part of the complex wavelet transform coefficient, is the gradient operator, is the spatial gradient of the complex wavelet coefficients, σ k is the noise suppression threshold of the kth layer.

[0115] It is worth noting that Formula 1 is based on a multi-directional decomposition architecture in the complex domain, which improves the gradient feature representation capability of multi-directional structures such as the gingival margin and alveolar bone microcracks. In Formula 1, By stacking the real part coefficients of the gains of different decomposition layers layer by layer, the edge information of the tissue structure is reconstructed, wherein the gain coefficient of the kth layer can be dynamically attenuated with the number of decomposition layers, thereby suppressing the excessive amplification of high-frequency noise. The spatial gradient constraint of the complex wavelet coefficient amplitude is used to enhance the consistency of direction and reduce the edge artifacts caused by local gain mutation.

[0116] S203 : Generate an image to be recognized based on the CBCT processed image and the CBCT enhanced image.

[0117] Specifically, Formula 2 can be used to generate the image to be identified F(x, y) based on the CBCT processed image I(x, y) and the CBCT enhanced image E(x, y), where Formula 2 is:

[0118]

[0119] Among them, F(x,y) is the image to be recognized, H I (x,y) and H E (x, y) is the local window entropy of the CBCT processed image I(x, y) and the CBCT enhanced image E(x, y) centered at the pixel position (x, y), Z(x, y) is the normalization factor, α and β are the adjustment hyperparameters, ω I With ω E are the first weight coefficient and the second weight coefficient respectively, and ω I +ω E =1.

[0120] It's worth noting that in Equation 2, the local window entropy centered at pixel position (x,y) of the processed and enhanced CBCT images, I(x,y), and E(x,y), respectively, describes the information complexity within a local window of the processed and enhanced CBCT images. Weighted fusion is then performed, achieving a weighted fusion of the two. The normalization factor ensures that the grayscale values of the fused image remain within a reasonable range. This fusion strategy preserves the spatial structure of the original image while highlighting high-frequency details and edge features.

[0121] As can be seen from S202-S203, the multi-layer complex wavelet transform significantly enhances high-frequency details in the image (such as tooth edges and gingival margin structures), effectively suppresses noise, and improves image clarity and contrast. Furthermore, a noise suppression threshold control strategy dynamically adjusts the noise characteristics of different decomposition layers, effectively avoiding the amplification effect of high-frequency noise and preserving the effective signal in the image. Therefore, by fusing the CBCT-processed and CBCT-enhanced images, the complementary information between the two is fully utilized, effectively improving the overall image quality and providing a reliable basis for the accurate identification of periodontal problems.

[0122] In other words, the aforementioned enhancement and fusion method enhances the weak signals of key periodontal anatomical structures through the inter-layer gain control and gradient constraint mechanism of the complex wavelet transform. Combined with a local entropy-driven adaptive fusion strategy, it optimizes the balance between artifact suppression and detail preservation. The complex wavelet-domain nonlinear enhancement of Equation 1 and the entropy-gradient hybrid weighted fusion of Equation 2 improve the edge sharpness and texture contrast of periodontal problem areas under the interference of metal implants, providing high-quality input images for subsequent gingival boundary detection and lesion localization.

[0123] Furthermore, before enhancing the CBCT processed image according to the multi-layer complex wavelet transform and the noise suppression threshold to generate the CBCT enhanced image, the noise suppression threshold may be determined according to the CBCT processed image and a preset layer attenuation factor.

[0124] Specifically, the noise suppression threshold may be determined using Formula 3 according to the CBCT processed image I(x,y) and a preset layer attenuation factor δ, where Formula 3 is:

[0125]

[0126] Where S is the total number of pixels in the CBCT processed image, δ is the preset layer attenuation factor, 0<δ<1, median is the median operation, δ k-1 is the attenuation factor of the k-1th layer.

[0127] The S in Formula 3 represents the total number of pixels in the CBCT processed image, which reflects the overall scale of the image. The preset layer attenuation factor δ is used to control the attenuation rate of the noise suppression threshold with the number of decomposition layers. By combining the two, Formula 3 can adaptively adjust the noise suppression thresholds of different decomposition layers to meet the needs of different images and different decomposition layers. In addition, Formula 3 performs a median operation on the real part of the complex wavelet transform coefficients of the kth layer, and calculates the median of the absolute value of the difference between it and the median. This step aims to capture the fluctuation of the complex wavelet transform coefficients of this layer to reflect the level of noise in this layer. By combining it with ln(S) and δ k-1 Multiplying and dividing by 0.6745 (this constant comes from the relationship between the standard deviation of Gaussian noise and the median absolute deviation), Formula 3 can accurately calculate the noise suppression threshold for this layer.

[0128] Because noise characteristics vary at different decomposition layers, different noise suppression thresholds need to be set for each layer. Formula 3 allows for dynamic adjustment of the noise suppression threshold, aligning the noise suppression strength with the signal energy distribution. At lower decomposition layers, where noise levels are higher, a higher noise suppression threshold is set to effectively suppress noise. At higher decomposition layers, where noise levels are lower and signal energy is stronger, a lower noise suppression threshold is set to preserve the valid signal.

[0129] Furthermore, the noise suppression threshold determined by Equation 3 can effectively suppress noise during the multi-layer complex wavelet transform process while preserving the effective signal in the image. This step is crucial for improving the quality of CBCT enhanced images because it directly determines the effectiveness and accuracy of subsequent enhancement processing.

[0130] Before enhancing CBCT images using Equation 1, the noise suppression threshold determined using Equation 3 provides an accurate noise suppression foundation for subsequent processing. This helps effectively avoid the amplification of high-frequency noise during the enhancement process, improving image clarity and contrast. Furthermore, because Equation 3 accurately calculates the noise suppression thresholds at different decomposition layers, it also helps improve the effectiveness of subsequent enhancement. By combining multi-layer complex wavelet transforms with a noise suppression threshold control strategy, effective enhancement of CBCT images can be achieved, generating high-quality enhanced CBCT images.

[0131] Furthermore, before enhancing the CBCT processed image according to the multi-layer complex wavelet transform and the noise suppression threshold to generate a CBCT enhanced image, the gain coefficient of the corresponding layer of the complex wavelet transform can be determined according to the preset gain intensity coefficient, the preset layer attenuation coefficient and the amplitude matrix of the complex wavelet transform coefficients of each layer. The gain coefficient is used to configure the enhancement degree during the enhancement process of the CBCT enhanced image and is positively correlated with the enhancement degree.

[0132] Specifically, the gain coefficient of the corresponding layer can be determined using Formula 4 according to the preset gain intensity coefficient, the preset layer attenuation coefficient, and the amplitude matrix of the complex wavelet transform coefficients of each layer. Formula 4 is:

[0133]

[0134] Wherein, μ is the preset gain intensity coefficient, θ is the preset layer attenuation coefficient, 0.7<θ<0.95, D k is the amplitude matrix of the k-th layer complex wavelet transform coefficients, η is the preset gain lower limit protection constant, SNR base is the preset reference signal-to-noise ratio, θ k-1 is the attenuation coefficient of the k-1th layer.

[0135] In the above scheme, μ in Formula 4 represents the preset gain intensity coefficient, which determines the strength of the overall gain. The preset layer attenuation coefficient θ is used to control the attenuation rate of the gain coefficient with the number of decomposition layers. By combining μ and θ, Formula 4 can adaptively adjust the gain coefficients of different decomposition layers to meet the needs of different images and different decomposition layers. D in Formula 4 k Represents the amplitude matrix of the complex wavelet transform coefficients of the kth layer, which reflects the energy distribution of the high-frequency details of the layer. By calculating D k The median and noise standard deviation σ k The ratio of can evaluate the intensity of the high-frequency details of this layer relative to the noise.

[0136] Furthermore, since the energy distribution of high-frequency details varies across decomposition layers, different gain coefficients need to be set for each decomposition layer. Formula 4 allows for dynamic adjustment of the gain coefficient. At low decomposition layers, high-frequency detail energy is weaker and noise levels are higher, so a lower gain coefficient is set to avoid introducing excessive noise. At high decomposition layers, high-frequency detail energy is stronger and noise levels are relatively lower, so a higher gain coefficient is set to effectively enhance high-frequency detail.

[0137] Furthermore, the gain coefficient determined by Equation 4 effectively enhances high-frequency details during the multi-layer complex wavelet transform process while avoiding the introduction of excessive noise. This step is crucial for improving the quality of CBCT enhanced images because it directly determines the accuracy and reliability of subsequent image fusion and periodontal problem identification.

[0138] By combining multi-layer complex wavelet transforms, noise suppression threshold control, and adaptive gain coefficient determination strategies, the quality of CBCT images can be significantly improved, providing strong support for the accurate diagnosis of periodontal problems. Furthermore, this technical solution exhibits excellent robustness and generalization capabilities, adapting to the needs of diverse oral environments and periodontal disease conditions.

[0139] Furthermore, before generating the image to be identified F(x,y) based on the CBCT processed image I(x,y) and the CBCT enhanced image E(x,y), Formula 5 can also be used to determine the local window entropy H of each pixel position of the CBCT processed image I(x,y) based on the CBCT processed image I(x,y). I (x, y), and determine the local window entropy H of each pixel position of the CBCT enhanced image E(x, y) according to the CBCT enhanced image E(x, y) E (x,y), where Formula 5 is:

[0140]

[0141] Among them, M is the quantization level, hist i is the frequency of grayscale value quantized to the i-th interval in the local window, Ω is the local window centered at the pixel position (x, y), R Ω is the pixel gray value extractor of the local window of the CBCT processed image, and |Ω| is the total pixels of the local window.

[0142] Based on the quantitative statistical results of grayscale values, Equation 5 uses the entropy calculation formula to calculate the local window entropy at each pixel location for both the processed and enhanced CBCT images. A higher entropy value indicates greater texture complexity and information content in that local region. As an information metric, local window entropy effectively reflects the information richness of a local region. By calculating the local window entropy of both processed and enhanced CBCT images, we can understand the texture complexity and information distribution across different regions of the image. During image fusion, the results of local window entropy calculations can provide an important basis for developing fusion strategies. For example, in periodontal data localization, information from the enhanced image can be prioritized for fusion in regions with high information richness (i.e., regions with high local window entropy) to improve the accuracy and reliability of the image to be identified. For regions with lower information richness, more information from the processed image can be retained to avoid introducing unnecessary noise. By calculating the local window entropy of both processed and enhanced CBCT images and fully considering this entropy value during the image fusion process, adaptive fusion of information from different regions can be achieved. This adaptive fusion strategy helps optimize image fusion and improve the quality of the image to be identified. In addition, since the calculation of local window entropy can reflect the information richness of the local area of the image, this information can be used to more accurately locate the position and scope of periodontal problems during periodontal problem identification.

[0143] Furthermore, before generating the image to be identified F(x,y) based on the CBCT processed image I(x,y) and the CBCT enhanced image E(x,y), Formula 6 can also be used, and the local window entropy H of each pixel position of the CBCT processed image I(x,y) is calculated. I (x, y) and the local window entropy H at each pixel position of the CBCT enhanced image E(x, y) E (x, y) determines the normalization factor Z(x, y) at the corresponding position, where Formula 6 is:

[0144]

[0145] in, It is the average of the average entropy of the entire CBCT processed image I(x,y) and the average entropy of the entire CBCT enhanced image E(x,y).

[0146] Specifically, by calculating the normalization factor, Equation 6 balances the information weighting of the processed and enhanced CBCT images during the fusion process. In regions with high information richness, the normalization factor decreases accordingly, reducing the proportion of that region's information in the fusion result. Conversely, in regions with low information richness, the normalization factor increases, increasing the proportion of that region's information in the fusion result. This balancing mechanism helps ensure that the fused image fully and accurately reflects the characteristics of periodontal problems.

[0147] Furthermore, the application of normalization factors can optimize the image fusion process and reduce information bias between different image regions. By balancing the information weights of different image regions, normalization factors help improve the quality of fused images, making them clearer and more accurate, providing strong support for subsequent identification and analysis of periodontal problems. By considering the information richness and weight distribution of different image regions, more rational fusion strategies can be designed to achieve effective information integration and optimization.

[0148] Furthermore, before generating the to-be-identified image F(x,y) based on the CBCT processed image I(x,y) and the CBCT enhanced image E(x,y), Formula 7 can also be used to determine and adjust hyperparameters based on the CBCT processed image I(x,y) and the CBCT enhanced image E(x,y). Formula 7 is:

[0149]

[0150] Among them, G x is the horizontal gradient output by the Sobel operator, G y It is the vertical gradient output by the Sobel operator.

[0151] In the above scheme, by calculating and adjusting hyperparameters, Equation 7 can adaptively adjust the weights of the CBCT-processed and CBCT-enhanced images during the fusion process based on their gradient information. In regions with high edge intensity (i.e., regions with large gradient amplitudes), the weight of the corresponding image is increased, thereby preserving more of that region's information in the fusion result. Conversely, in regions with low edge intensity, the weight of the corresponding image is reduced to avoid introducing unnecessary noise. Adaptive adjustment of fusion weights helps optimize the image fusion process and improve the quality of the fused image. By rationally assigning weights based on the actual information characteristics of the images, the fused image can be made clearer and more accurate, better reflecting the characteristics of periodontal problems.

[0152] S204 , performing grayscale processing on the image to be identified, determining a gingival boundary point set according to the grayscale gradient, and performing curve fitting on the gingival boundary point set to determine a gingival line curve.

[0153] Based on the grayscale image, a grayscale gradient operator is used to perform gradient analysis on the image. Grayscale gradient reflects the degree of brightness change in the image and is an important basis for edge detection. Common grayscale gradient operators include the Sobel operator and the Prewitt operator. In this method, the Sobel operator can be used to calculate the horizontal and vertical gradients of the image, thereby obtaining the gradient magnitude and gradient direction for each pixel.

[0154] Grayscale gradient analysis can identify areas of image brightness with significant variations, often corresponding to object edges. In periodontal data localization, the boundary between the gums and the tooth itself is of interest. Therefore, based on the magnitude and direction of the grayscale gradient, possible gum boundary points can be screened to form a gum boundary point set. This point set contains the approximate location of the gum boundary but may contain noise and errors.

[0155] To obtain an accurate gum line curve, curve fitting is required for the gum boundary point set, estimating a continuous curve from a series of discrete points. In this embodiment, parametric curve fitting methods such as polynomial fitting and spline fitting can be used. Through curve fitting, a smooth gum line curve can be obtained that accurately reflects the boundary between the gums and the tooth body.

[0156] S205 : Determine a characteristic segment according to the gum line curve, and determine a corresponding periodontal problem area in the image to be identified according to the characteristic segment.

[0157] In this step, after obtaining the gum line curve, characteristic segments can be identified based on its shape and characteristics. Characteristic segments are areas of the gum line curve that have distinctive shapes or significant variations, and these areas are often closely associated with periodontal problems. For example, a concave portion of the gum line curve may correspond to periodontal pocket formation, while a convex portion of the curve may correspond to gingival hyperplasia or swelling. By identifying these characteristic segments, the area of periodontal problem can be further narrowed down.

[0158] Finally, the corresponding periodontal problem areas are identified in the image to be identified based on the characteristic segments and marked. These markings can use different colors, shapes, or symbols to indicate the location and severity of the periodontal problem. For example, a red circle or rectangle can be used to mark the location and size of periodontal pockets, while a green arrow or line can be used to indicate the direction of gum hyperplasia or swelling. This marking display allows for intuitive visualization of the distribution of periodontal problems, providing strong support for subsequent diagnosis and treatment.

[0159] In one possible implementation, a reference curve can be formed based on at least two lowest or highest boundary points in a set of gingival boundary points. Then, reference feature regions are determined based on the reference curve and the gingival line curve. A reference feature region is a closed region formed by the reference curve and the gingival line curve. If the maximum width of the closed region is greater than a preset width threshold and / or the area of the closed region is greater than a preset area threshold, the closed region is determined to be a periodontal problem area.

[0160] Specifically, at least two boundary points of special significance are selected from the set of gingival boundary points, such as the lowest boundary point or the highest boundary point. These points are typically located at the extremes of the gingival line curve and can serve as anchor points for the reference curve. In practice, the set of gingival boundary points can be sorted and the first and last points (or points at other specific locations as needed) can be selected as reference points.

[0161] Connect the selected reference points to form a reference curve. This curve will serve as the baseline for subsequently defining the reference feature area. The shape and position of the reference curve depend on the selected reference points. It should roughly reflect the overall trend of the gum line curve while not completely overlapping it, so that a closed area can be formed later.

[0162] Based on the reference curve, it intersects with the gingival line curve to form multiple closed areas. These closed areas are enclosed by the reference curve and the gingival line curve, representing specific segments on the gingival line curve.

[0163] In practice, closed regions are formed by calculating the intersection points between the reference curve and the gum line curve and then connecting adjacent intersection points. For each closed region, its maximum width and area are calculated. These properties are used to subsequently determine whether the region is a periodontal problem area. The maximum width can be calculated by finding the two points in the closed region that are farthest apart, while the area can be estimated using integration or other numerical methods.

[0164] Preset width thresholds and area thresholds. These thresholds are set based on clinical experience, statistical data, or expert opinion to distinguish between normal periodontal tissue and potential problem areas. The setting of thresholds should take into account the differences between individuals and the diversity of periodontal problems. The maximum width value and area of each enclosed area are compared with the preset thresholds. If the maximum width value of the enclosed area is greater than the preset width threshold, or the area of the enclosed area is greater than the preset area threshold (or both conditions are met), the enclosed area is determined to be a periodontal problem area.

[0165] Identified periodontal problem areas are marked and displayed in the image to be identified. Markers can use different colors, shapes, or symbols to indicate the location and severity of the problem. This marking display allows for intuitive visualization of the distribution of periodontal problems, providing strong support for subsequent diagnosis and treatment.

[0166] Figure 3 FIG. 1 is a schematic diagram of a periodontal problem data location device based on CBCT images according to an exemplary embodiment of the present application. Figure 3 As shown, the periodontal problem data positioning device 300 based on CBCT images provided in this embodiment includes:

[0167] An acquisition module 310 is configured to acquire a CBCT original image of an oral region of a target subject, wherein the oral region is equipped with a metal implant, and wherein the CBCT original image contains artifact components associated with the metal implant;

[0168] a processing module 320 configured to filter the original frequency domain image corresponding to the original CBCT image using a high-pass filter to generate a filtered frequency domain image, and generate a processed CBCT image based on the filtered frequency domain image, wherein a cutoff frequency of the high-pass filter is associated with the artifact component;

[0169] The processing module 320 is further configured to perform enhancement processing on the CBCT processed image to generate a CBCT enhanced image, and perform fusion processing on the CBCT processed image and the CBCT enhanced image to generate an image to be recognized;

[0170] The processing module 320 is further configured to determine periodontal problem location information based on the image to be identified.

[0171] Optionally, the processing module 320 is specifically configured to:

[0172] Converting the CBCT original image from the spatial domain to the frequency domain using a two-dimensional Fourier transform to generate the original frequency domain image;

[0173] filtering the frequency domain elements in the original frequency domain image that are lower than the cutoff frequency according to the high-pass filter to generate the filtered frequency domain image;

[0174] The filtered frequency domain image is converted from the frequency domain to the spatial domain using a two-dimensional inverse Fourier transform to generate the CBCT processed image.

[0175] Optionally, the processing module 320 is specifically configured to:

[0176] performing enhancement processing on the CBCT processed image according to a multi-layer complex wavelet transform and a noise suppression threshold to generate the CBCT enhanced image;

[0177] The image to be identified is generated according to the CBCT processed image and the CBCT enhanced image.

[0178] Optionally, the processing module 320 is specifically configured to:

[0179] The noise suppression threshold is determined according to the CBCT processed image and a preset layer attenuation factor.

[0180] Optionally, the processing module 320 is specifically configured to:

[0181] The gain coefficient of the corresponding layer of the complex wavelet transform is determined according to the preset gain intensity coefficient, the preset layer attenuation coefficient and the amplitude matrix of the complex wavelet transform coefficient of each layer. The gain coefficient is used to configure the enhancement degree during the enhancement processing of the CBCT enhanced image and is positively correlated with the enhancement degree.

[0182] Optionally, the processing module 320 is specifically configured to:

[0183] Performing grayscale processing on the image to be identified, determining a gingival boundary point set according to the grayscale gradient, and performing curve fitting on the gingival boundary point set to determine a gingival line curve;

[0184] A characteristic segment is determined according to the gum line curve, and a corresponding periodontal problem area is determined in the image to be identified according to the characteristic segment, so as to mark and display the periodontal problem area in the image to be identified.

[0185] Optionally, the processing module 320 is specifically configured to:

[0186] forming a reference curve according to at least two lowest boundary points or two highest boundary points in the gingival boundary point set;

[0187] Determine each reference feature area according to the reference curve and the gum line curve, wherein the reference feature area is a closed area formed by the reference curve and the gum line curve;

[0188] If the maximum width of the closed area is greater than a preset width threshold and / or the area of the closed area is greater than a preset area threshold, the closed area is determined to be the periodontal problem area.

[0189] Figure 4 FIG. 1 is a schematic diagram of the structure of an electronic device according to an exemplary embodiment of the present application. Figure 4 As shown, this embodiment provides an electronic device 400 including: a processor 401 and a memory 402; wherein:

[0190] The memory 402 is used to store computer programs, and the memory may also be a flash memory.

[0191] The processor 401 is configured to execute the execution instructions stored in the memory to implement each step in the above method. For details, please refer to the relevant description in the above method embodiment.

[0192] Optionally, the memory 402 may be independent or integrated with the processor 401 .

[0193] When the memory 402 is a device independent of the processor 401, the electronic device 400 may further include:

[0194] The bus 403 is used to connect the memory 402 and the processor 401 .

[0195] This embodiment further provides a readable storage medium, in which a computer program is stored. When at least one processor of an electronic device executes the computer program, the electronic device executes the methods provided in the various aforementioned embodiments.

[0196] This embodiment further provides a program product, which includes a computer program stored in a readable storage medium. At least one processor of an electronic device can read the computer program from the readable storage medium, and at least one processor can execute the computer program to cause the electronic device to implement the methods provided in the various embodiments described above.

[0197] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present application that follow the general principles of the present application and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered merely as exemplary, and the true scope and spirit of the present application are indicated by the claims.

[0198] It should be understood that the present application is not limited to the exact structure described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for locating periodontal problem data based on CBCT images, characterized in that: include: Acquiring a CBCT original image of an oral region of a target subject, wherein the oral region is equipped with a metal implant, and wherein the CBCT original image contains artifact components associated with the metal implant; performing filtering processing on an original frequency domain image corresponding to the original CBCT image through a high-pass filter to generate a filtered frequency domain image, and generating a CBCT processed image based on the filtered frequency domain image, wherein a cutoff frequency of the high-pass filter is associated with the artifact component; performing enhancement processing on the CBCT processed image to generate a CBCT enhanced image, and performing fusion processing on the CBCT processed image and the CBCT enhanced image to generate an image to be recognized; The periodontal problem location information is determined according to the image to be identified.

2. The method for periodontal problem data positioning based on CBCT images according to claim 1, characterized in that: The filtering process of the original frequency domain image corresponding to the original CBCT image by a high-pass filter to generate a filtered frequency domain image, and generating a CBCT processed image according to the filtered frequency domain image includes: Converting the CBCT original image from the spatial domain to the frequency domain using a two-dimensional Fourier transform to generate the original frequency domain image; filtering the frequency domain elements in the original frequency domain image that are lower than the cutoff frequency according to the high-pass filter to generate the filtered frequency domain image; The filtered frequency domain image is converted from the frequency domain to the spatial domain using a two-dimensional inverse Fourier transform to generate the CBCT processed image.

3. The method for periodontal problem data positioning based on CBCT images according to claim 1, characterized in that: The step of enhancing the CBCT processed image to generate a CBCT enhanced image, and fusing the CBCT processed image and the CBCT enhanced image to generate an image to be identified includes: performing enhancement processing on the CBCT processed image according to a multi-layer complex wavelet transform and a noise suppression threshold to generate the CBCT enhanced image; The image to be identified is generated according to the CBCT processed image and the CBCT enhanced image.

4. The method for periodontal problem data positioning based on CBCT images according to claim 3, characterized in that: Before performing enhancement processing on the CBCT processed image according to the multi-layer complex wavelet transform and the noise suppression threshold to generate the CBCT enhanced image, the method further includes: The noise suppression threshold is determined according to the CBCT processed image and a preset layer attenuation factor.

5. The method for periodontal problem data positioning based on CBCT images according to claim 3, characterized in that: Before performing enhancement processing on the CBCT processed image according to the multi-layer complex wavelet transform and the noise suppression threshold to generate the CBCT enhanced image, the method further includes: The gain coefficient of the corresponding layer of the complex wavelet transform is determined according to the preset gain intensity coefficient, the preset layer attenuation coefficient and the amplitude matrix of the complex wavelet transform coefficient of each layer. The gain coefficient is used to configure the enhancement degree during the enhancement processing of the CBCT enhanced image and is positively correlated with the enhancement degree.

6. The method for periodontal problem data positioning based on CBCT images according to claim 1, characterized in that: The determining of periodontal problem location information according to the image to be identified includes: Performing grayscale processing on the image to be identified, determining a gingival boundary point set according to the grayscale gradient, and performing curve fitting on the gingival boundary point set to determine a gingival line curve; A characteristic segment is determined according to the gum line curve, and a corresponding periodontal problem area is determined in the image to be identified according to the characteristic segment, so as to mark and display the periodontal problem area in the image to be identified.

7. The method for periodontal problem data positioning based on CBCT images according to claim 6, characterized in that: The determining of a characteristic segment according to the gum line curve, and determining a corresponding periodontal problem area in the image to be identified according to the characteristic segment, includes: forming a reference curve according to at least two lowest boundary points or two highest boundary points in the gingival boundary point set; Determine each reference feature area according to the reference curve and the gum line curve, wherein the reference feature area is a closed area formed by the reference curve and the gum line curve; If the maximum width of the closed area is greater than a preset width threshold and / or the area of the closed area is greater than a preset area threshold, the closed area is determined to be the periodontal problem area.

8. A periodontal problem data positioning device based on CBCT images, characterized in that: include: an acquisition module, configured to acquire a CBCT original image of an oral region of a target subject, wherein the oral region is equipped with a metal implant, and wherein the CBCT original image contains an artifact component associated with the metal implant; a processing module, configured to filter an original frequency domain image corresponding to the original CBCT image using a high-pass filter to generate a filtered frequency domain image, and generate a processed CBCT image based on the filtered frequency domain image, wherein a cutoff frequency of the high-pass filter is associated with the artifact component; The processing module is further configured to perform enhancement processing on the CBCT processed image to generate a CBCT enhanced image, and perform fusion processing on the CBCT processed image and the CBCT enhanced image to generate an image to be recognized; The processing module is further configured to determine periodontal problem location information based on the image to be identified.

9. An electronic device, characterized in that: include: processor; as well as, a memory for storing executable instructions of the processor; The processor is configured to perform the method according to any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 7 when executed by a processor.

Images