Oral risk assessment method and system based on image recognition and medium

Abstract

The invention discloses an oral risk assessment method and system based on image recognition and a medium, and relates to the technical field of image recognition, and the method comprises the steps: carrying out the block segmentation of a user oral image, activating a lesion recognizer to recognize the features of an image block, carrying out the matching of an oral disease, and obtaining a corresponding flora type set; imaging the target flora by using a fluorescence microscope, and extracting fluorescence characteristics; and generating an oral risk decision of the target user in combination with the identified disease information and the fluorescence characteristics. Therefore, the technical effects of multi-dimensional image comprehensive analysis and evaluation accuracy and efficiency improvement are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and particularly relates to an oral risk assessment method, system and medium based on image recognition. Background Art

[0002] Due to the particularity of the long voyage operation environment, the microbial flora in the oral cavity, respiratory tract and gastrointestinal tract of long voyage personnel will change, leading to a series of pathological changes, various diseases, and affecting the health of crew members.

[0003] Traditional methods for diagnosing oral diseases mainly rely on the clinical examinations and empirical judgments of dentists, which are subjective and limited to a certain extent and affect efficiency at the same time. Meanwhile, the utilization of oral images is not sufficient enough, and the analysis and processing of images mostly stay on the surface, making it difficult to deeply excavate the relevant information contained in the images. Summary of the Invention

[0004] The present invention provides an oral risk assessment method, system and medium based on image recognition to solve the technical problems of insufficient recognition accuracy and insufficient comprehensiveness of risk decision-making in the prior art, and achieve the technical effects of comprehensive multi-dimensional image analysis, improving assessment accuracy and efficiency.

[0005] In a first aspect, the present invention provides an oral risk assessment method based on image recognition, wherein the oral risk assessment method based on image recognition includes: Performing block segmentation on a target oral image of a target user collected by a medical imaging device to obtain a set of target image blocks.

[0006] Activating a lesion recognizer to analyze and recognize first image features of a first image block to obtain a first lesion recognition result, wherein the first image block refers to any one image block in the set of target image blocks.

[0007] If the first lesion recognition result meets a predetermined constraint, matching a first oral disease, and traversing in an oral disease flora database to obtain a set of first flora types corresponding to the first oral disease.

[0008] Performing fluorescence imaging acquisition on a first flora in the set of first flora types through a fluorescence microscope to obtain a first fluorescence image.

[0009] Collecting first fluorescence features of the first fluorescence image and combining with the first oral disease to form a target oral risk decision of the target user.

[0010] In a feasible implementation manner, before activating the lesion recognizer to analyze and recognize the first image features of the first image block to obtain the first lesion recognition result, it includes: Determine the first oral region corresponding to the first image block.

[0011] Construct a first image sample set of the first oral region, where the first image sample set includes a first healthy image and a first diseased image.

[0012] Construct a training data set based on the first healthy image and the first diseased image, and perform supervised machine learning on the training data set to obtain the lesion recognizer.

[0013] Among them, constructing a training data set based on the first healthy image and the first diseased image includes: Obtain the first feature information of the first healthy image and form a first data set.

[0014] Obtain the second feature information of the first diseased image and form a second data set.

[0015] The first data set and the second data set form the training data set.

[0016] In a feasible implementation manner, if the first lesion recognition result meets a predetermined constraint, match the first oral disease, and traverse in the oral disease flora database to obtain a first flora type set corresponding to the first oral disease, including: Obtain a healthy oral sample, and the healthy oral sample has a healthy flora type set.

[0017] Obtain a first oral sample of the first oral disease, and the first oral sample has a disease flora type set.

[0018] Compare the healthy flora type set with the disease flora type set to obtain the first flora type set of the first oral disease.

[0019] Construct the oral disease flora database according to the corresponding relationship between the first oral disease and the first flora type set.

[0020] In a feasible implementation manner, perform fluorescence imaging acquisition on the first flora in the first flora type set through a fluorescence microscope to obtain a first fluorescence image, including: Obtain a first antibody corresponding to the first flora.

[0021] Perform fluorescence labeling on the first antibody to obtain a first labeling result.

[0022] Obtain the first fluorescence image through the fluorescence microscope, where the first fluorescence image refers to the combined imaging of the first labeling result and the first flora.

[0023] In a feasible implementation manner, collecting the first fluorescence features of the first fluorescence image and combining with the first oral disease to form the target oral risk decision of the target user includes: Performing normalization weighted calculation on the first fluorescence features to obtain a first fluorescence index.

[0024] Taking the first fluorescence index as the first risk index of the first oral disease.

[0025] Forming the target oral risk decision according to the first corresponding relationship between the first oral disease and the first risk index.

[0026] In a feasible implementation manner, the first fluorescence features at least include a first fluorescence coverage rate and a first fluorescence dispersion degree.

[0027] In a feasible implementation manner, after collecting the first fluorescence features of the first fluorescence image and combining with the first oral disease to form the target oral risk decision of the target user, it further includes: Obtaining a first predetermined amplification curve of the first bacterial flora.

[0028] Taking the first predetermined amplification curve as a constraint and combining with the first fluorescence features to obtain the predicted fluorescence features of the first bacterial flora at a target time.

[0029] Analyzing the predicted fluorescence features to obtain a predicted risk index and forming the target oral risk prediction of the target user.

[0030] In a feasible implementation manner, taking the first predetermined amplification curve as a constraint and combining with the first fluorescence features to obtain the predicted fluorescence features of the first bacterial flora at a target time includes: Performing polynomial fitting on the first predetermined amplification curve to obtain a first fitting formula.

[0031] Inputting the target time into the first fitting formula to obtain a first fitting value.

[0032] Taking the first fitting value as a first predicted amplification coefficient and performing weighted adjustment on the first fluorescence features to obtain the predicted fluorescence features.

[0033] In a second aspect, the present invention further provides an oral risk assessment system based on image recognition. Among them, the oral risk assessment system based on image recognition includes: An image block segmentation module, configured to perform block segmentation on the target oral image of the target user collected by a medical imaging device to obtain a set of target image blocks.

[0034] A lesion recognition module, configured to activate a lesion recognizer to analyze and recognize the first image features of the first image patch, and obtain a first lesion recognition result, where the first image patch refers to any one image patch in the target image patch set.

[0035] An oral disease matching module, configured to match a first oral disease if the first lesion recognition result meets a predetermined constraint, and traverse in an oral disease flora database to obtain a first flora type set corresponding to the first oral disease.

[0036] A fluorescence imaging acquisition module, configured to perform fluorescence imaging acquisition on the first flora in the first flora type set through a fluorescence microscope to obtain a first fluorescence image.

[0037] An oral risk decision-making module, configured to collect the first fluorescence features of the first fluorescence image, and combine with the first oral disease to form a target oral risk decision for the target user.

[0038] In a third aspect, the present invention further provides a computer-readable storage medium storing a computer program, which when executed by a processor, implements the oral risk assessment method based on image recognition provided by the present invention.

[0039] The present invention discloses an oral risk assessment method, system and medium based on image recognition, including: performing image patch segmentation on a target oral image of a target user collected by a medical imaging device to construct a target image patch set; activating a lesion recognizer to perform feature extraction and recognition on any first image patch in the target image patch set to obtain a first lesion recognition result; if the first lesion recognition result meets a predetermined recognition constraint, matching to obtain a corresponding first oral disease, and searching in a preset oral disease flora database for a first flora type set corresponding to the first oral disease; performing fluorescence imaging on a target first flora in the first flora type set based on a fluorescence microscope to obtain a first fluorescence image; extracting the first fluorescence features of the first fluorescence image, and performing risk reasoning in combination with the first oral disease to form a target oral risk decision result for the target user. The oral risk assessment method, system and medium based on image recognition disclosed by the present invention solve the technical problems of insufficient recognition accuracy and insufficient comprehensiveness of risk decision-making, and achieve the technical effects of multi-dimensional image comprehensive analysis, improving assessment accuracy and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a schematic flowchart of the oral risk assessment method based on image recognition according to the present invention.

[0041] Figure 2 It is a schematic structural diagram of the oral risk assessment system based on image recognition according to the present invention.

[0042] Description of the reference numerals in the drawings: Image block segmentation module 11, lesion recognition module 12, oral disease matching module 13, fluorescence imaging acquisition module 14, oral risk decision-making module 15. Detailed implementation manners

[0043] The above technical solutions will be described in detail below in combination with the drawings in the specification and specific implementation manners to better understand the above technical solutions. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments only used to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention. In addition, it should be noted that, for the sake of convenience of description, only parts related to the present invention are shown in the drawings instead of all.

[0044] Embodiment 1, as Figure 1 is a schematic flow chart of the oral risk assessment method based on image recognition of the present invention. Among them, the oral risk assessment method based on image recognition includes: S100: Perform block segmentation on the target oral image of the target user collected by a medical imaging device to obtain a set of target image blocks.

[0045] Specifically, the medical imaging device refers to a medical device capable of collecting oral images, such as oral CBCT, digital panoramic dental X-ray machine, etc., which can provide detailed image information of the oral cavity of the target user for subsequent analysis.

[0046] Specifically, block segmentation is an image processing technology used to perform regional segmentation operations according to a preset spatial partitioning strategy, so as to obtain a set of image blocks with local features, thereby facilitating separate analysis of different regions of the image and improving the accuracy and efficiency of analysis.

[0047] Specifically, first, collect an image of the oral cavity area of the target user through a medical imaging device to obtain the original image data containing the complete oral cavity structure; subsequently, based on the image size, resolution and the distribution characteristics of the oral cavity structure, set the image block segmentation parameters, including the image block size, overlap rate and segmentation method, etc.

[0048] Furthermore, in the segmentation process, the target oral image can be spatially segmented by means of a sliding window or grid division to generate multiple image blocks, and each image block corresponds to a local area in the oral image; at the same time, mark the position index of different image blocks in the original image to form a set of target image blocks for subsequent processing modules to call.

[0049] Through the above splitting operation, local regions with different features can be extracted from the overall image, that is, the obtained set of target image patches, providing a basis for subsequent in-depth analysis of each image patch.

[0050] S200: Activate the lesion recognizer to analyze and recognize the first image features of the first image patch, and obtain a first lesion recognition result, where the first image patch refers to any one of the set of target image patches.

[0051] Specifically, the lesion recognizer is a preset image analysis module for automatically recognizing and classifying possible lesion regions in oral image patches. An image feature extraction sub-module and a lesion type discrimination sub-module based on a convolutional neural network are integrated inside the lesion recognizer, which can model and recognize the pathological features in the input image patch. Among them, the first image patch refers to any one selected from the set of target image patches, containing local structural information in the oral image, and can be input into the lesion recognizer as an independent analysis unit for processing; the first image features are multi-dimensional image feature data extracted from the first image patch, for example, including texture features, edge features, color distribution, spatial morphology, etc.

[0052] Through the above process, lesion recognition analysis can be performed on any image patch in the set of target image patches, which helps to improve the automation and intelligence level of lesion detection, reduce the repetitive labor intensity of relevant personnel, and provide reliable data support for the formulation of subsequent personalized treatment plans.

[0053] In some embodiments, before activating the lesion recognizer to analyze and recognize the first image features of the first image patch and obtain a first lesion recognition result, it includes: Determine the first oral site corresponding to the first image patch; form a first image sample set of the first oral site, where the first image sample set includes a first healthy image and a first lesion image; form a training data set based on the first healthy image and the first lesion image, and perform supervised machine learning on the training data set to obtain the lesion recognizer; where forming a training data set based on the first healthy image and the first lesion image includes: Obtain the first feature information of the first healthy image and form a first data set; obtain the second feature information of the first lesion image and form a second data set; the first data set and the second data set form the training data set.

[0054] Specifically, the first oral region refers to the specific area corresponding to the first image patch in the target oral image, including but not limited to the incisor region, molar region, gingival margin, lingual region, etc., which is used to guide the construction of the subsequent sample set, so as to achieve targeted training enhancement of the model. The first image sample set is a representative image set collected and sorted for the first oral region, including the first healthy images and the first lesion images collected at the same or similar sites. Among them, the first healthy image is a standard structural image without obvious lesion characteristics, and the first lesion image is a labeled image with a confirmed lesion area, which is used to construct a training data set to improve the recognition accuracy of the lesion recognizer in this region.

[0055] Specifically, after the image patch segmentation is completed, first, the first oral region corresponding to the first image patch is determined by the position information of the image patch in the original image in combination with the oral structure mapping relationship; then, the image samples corresponding to this oral region are called from the historical database or expert-annotated samples to form the first image sample set, which contains multiple first healthy images and first lesion images; next, the image features of the first healthy images are extracted, including first feature information such as texture, edge, color, shape, etc. to form the first data set; at the same time, the image features of the first lesion images are extracted to obtain the second feature information and form the second data set; furthermore, the first data set and the second data set are merged to construct a training data set, and a supervised machine learning algorithm is used to construct and train a lesion recognizer based on methods such as convolutional neural network, support vector machine or ensemble learning.

[0056] Through the above process, a training data set with regional pertinence is constructed according to the oral anatomical region corresponding to the first image patch, and a dedicated recognition model is trained through a supervised machine learning method, so that the lesion recognizer has stronger regional adaptability and classification accuracy, which helps to improve the early recognition accuracy of lesions in specific oral regions.

[0057] S300: If the first lesion recognition result meets the predetermined constraints, match the first oral disease, and traverse the oral disease flora database to obtain the first flora type set corresponding to the first oral disease.

[0058] Specifically, the predetermined constraints are a set of conditions (parameters or indicators) preset for judging whether the lesion recognition result is reliable, such as a specific eigenvalue range, image pattern matching degree, confidence level of the first lesion recognition result, etc.

[0059] Specifically, the oral disease flora database is a data set that stores information on different types of oral diseases and their corresponding flora. In other words, the oral disease flora database contains detailed information such as the types, quantities, and distributions of the flora in the oral cavity under various oral disease states. By traversing and searching in this database, a first flora type set corresponding to the first oral disease can be obtained, that is, a relevant flora combination specific to this oral disease, which is used to provide a reference for subsequent further collection and analysis.

[0060] In the above processing flow, on the one hand, by making a judgment on the preset constraints of the lesion recognition result, unreliable recognition results are filtered, thereby improving the accuracy of subsequent risk assessment; on the other hand, the process of matching oral diseases and obtaining the corresponding flora type set broadens the dimension of risk assessment, and the subsequent assessment combining oral flora indicators helps to improve the comprehensiveness of the assessment results.

[0061] In some embodiments, if the first lesion recognition result meets the predetermined constraints, then match the first oral disease, and traverse in the oral disease flora database to obtain the first flora type set corresponding to the first oral disease, including: Obtain a healthy oral sample, and the healthy oral sample has a healthy flora type set; obtain the first oral sample of the first oral disease, and the first oral sample has a disease flora type set; compare the healthy flora type set with the disease flora type set to obtain the first flora type set of the first oral disease; construct the oral disease flora database according to the corresponding relationship between the first oral disease and the first flora type set.

[0062] Specifically, a healthy oral sample is a sample taken from the oral cavity of a healthy individual, which can represent the oral microbial composition in a normal and healthy state. Among them, the healthy flora type set is the combined information of various healthy microorganisms contained in this sample.

[0063] Specifically, the first oral sample of the first oral disease refers to the oral sample of a patient with a specific oral disease (i.e., the first oral disease), and the microbial combination present in it is the disease flora type set.

[0064] Specifically, first, collect an oral sample from an individual without clinical symptoms, that is, a healthy oral sample; after microbiome sequencing or flora analysis, determine the healthy flora type set, including the flora composition and its abundance distribution under the normal microecological state. At the same time, collect a sample of a patient diagnosed with the first oral disease, that is, the first oral sample; similarly, through the same type of microbiome analysis and processing, obtain its corresponding disease flora type set, which is used to reflect the flora imbalance characteristics under this disease state.

[0065] Furthermore, perform differential analysis on the healthy flora type set and the diseased flora type set, including but not limited to differences in species diversity, changes in dominant species, significance tests for the abundance of characteristic flora, etc., and extract the significantly different flora types based on statistical methods (such as LEfSe analysis, Wilcoxon rank-sum test, principal component analysis PCA, etc.), and output them as the first flora type set of the first oral disease. This flora type set may include indicators such as a list of characteristic species, abundance change trends, and abnormal related metabolic pathways.

[0066] Furthermore, based on the mapping relationship between the first oral disease and the first flora type set, construct an oral disease flora database for recording different oral diseases (such as dental caries, gingivitis, periodontitis, oral candidiasis infection, etc.) and their corresponding characteristic flora type sets.

[0067] Exemplarily, the structure of the oral disease flora database is shown in Table 1 below: Table 1 Exemplary Oral Disease Flora Database Preferably, information such as flora regulation strategies, probiotic intervention programs, and individualized treatment suggestions can be further associated in the oral disease flora database to facilitate improving the corresponding response efficiency after risk assessment.

[0068] Through the above process, the lesion results of the image recognition layer can be effectively linked with the flora characteristics of the microecological layer, realizing cross-modal information fusion from lesion detection to flora association, and providing data support for subsequent analysis.

[0069] S400: Collect fluorescence imaging of the first flora in the first flora type set through a fluorescence microscope to obtain a first fluorescence image.

[0070] Specifically, a fluorescence microscope is a precision instrument that can observe and image samples using fluorescence characteristics, that is, by exciting fluorescent substances in the sample to emit light with a specific wavelength of light, and then showing the distribution and morphology of specific components in the sample. By observing the first flora through a fluorescence microscope, the observed fluorescence signal can be converted into image data to obtain a first fluorescence image. This image is used to intuitively display the morphological, distribution and other characteristic information of the first flora after fluorescence labeling, providing data support for subsequent quantitative analysis and risk assessment of flora characteristics.

[0071] Exemplarily, first, fluorescence imaging acquisition of the first microbial community in the first microbial community type concentration is performed through a fluorescence microscope to obtain a first fluorescence image. Among them, the first microbial community is a target strain isolated from a first oral sample, and the microbial community is specifically stained in advance through fluorescence labeling means (such as fluorescence in situ hybridization FISH, fluorescent protein labeling, fluorescent dye staining, etc.) to enhance its detectability during the fluorescence imaging process. The obtained first fluorescence image records the distribution characteristics, aggregation morphology, and relative abundance information of the target microbial community in the oral sample.

[0072] In some embodiments, fluorescence imaging acquisition of the first microbial community in the first microbial community type concentration is performed through a fluorescence microscope to obtain a first fluorescence image, including: Obtain a first antibody corresponding to the first microbial community; perform fluorescence labeling on the first antibody to obtain a first labeling result; obtain the first fluorescence image through the fluorescence microscope, where the first fluorescence image refers to the combined imaging of the first labeling result and the first microbial community.

[0073] Specifically, first, according to the specific antigen of the target bacterial species in the first microbial community, a first antibody with a highly specific binding ability to it is obtained. Among them, the first antibody can be a monoclonal antibody or a polyclonal antibody, preferably designed for the surface antigen of the target strain to ensure the specificity of binding and the accuracy of imaging. Then, the first antibody is subjected to fluorescence labeling treatment to obtain a first labeling result. Exemplarily, the fluorescence labeling can use commonly used fluorescent molecules such as fluorescein isothiocyanate (FITC), rhodamine, Alexa Fluor series dyes, etc., so that the antibody generates a detectable fluorescence signal under specific wavelength excitation.

[0074] Furthermore, incubate the first labeling result with the first microbial community to make the fluorescently labeled antibody bind to the surface antigen of the target microbial community to form a stable complex; subsequently, perform imaging acquisition on the combined sample area through a fluorescence microscope to obtain a first fluorescence image.

[0075] The above process plays a role in converting microbiological information into analyzable image data, realizing an important transformation from the determination of microbial community type to the visualization of specific microbial community characteristics, and providing key image materials for subsequent quantitative analysis and risk decision-making based on fluorescence characteristics.

[0076] S500: Collect the first fluorescence characteristics of the first fluorescence image and combine them with the first oral disease to form the target oral risk decision of the target user.

[0077] Specifically, the first fluorescence feature refers to the image feature information extracted from the first fluorescence image, including but not limited to parameters such as brightness distribution, color channel intensity, fluorescence intensity gradient, edge morphology, etc., which are used to characterize the optical properties of the lesion area. There is a specific pattern matching relationship between the first fluorescence feature and the first oral disease. According to this relationship, the oral risks existing in the target user can be identified correspondingly, and a target oral risk decision can be generated.

[0078] In some implementation manners, the first fluorescence feature at least includes a first fluorescence coverage rate and a first fluorescence dispersion degree.

[0079] Specifically, the first fluorescence feature of the first fluorescence image includes indicators that can reflect the growth state of the flora, such as the fluorescence coverage rate and the fluorescence dispersion degree. Among them, the first fluorescence coverage rate refers to the area ratio of the regions detected as abnormal fluorescence regions (such as weak fluorescence or strong fluorescence regions) in the first fluorescence image, which is used to measure the distribution breadth of the lesions or dental plaque in the entire oral image; the first fluorescence dispersion degree refers to the degree of dispersion of the fluorescence abnormal regions in the image, reflecting whether the lesion regions are concentrated in a certain local area or widely distributed in multiple positions. Exemplarily, it can be quantified by means such as centroid distribution, standard deviation or spatial clustering coefficient, etc., which is used to assist in judging the severity and development trend of the lesions.

[0080] Specifically, first, the fluorescence abnormal regions are identified through a fluorescence intensity threshold segmentation algorithm, and the area ratio of them in the whole image is calculated to obtain the first fluorescence coverage rate. For example, in an image with a resolution of 1024×1024 pixels, if the total area of the abnormal fluorescence regions is 50000 pixels, then the first fluorescence coverage rate is approximately 4.77%. Subsequently, the spatial distribution of these abnormal regions is analyzed, the centroid coordinates of them are calculated, and the standard deviation or average distance of their distribution is evaluated, so as to obtain the first fluorescence dispersion degree.

[0081] In some embodiments, collecting the first fluorescence feature of the first fluorescence image and combining it with the first oral disease to form the target oral risk decision of the target user includes: Performing a normalized weighted calculation on the first fluorescence feature to obtain a first fluorescence index; using the first fluorescence index as the first risk index of the first oral disease; forming the target oral risk decision according to the first corresponding relationship between the first oral disease and the first risk index.

[0082] Specifically, after extracting features such as the first fluorescence coverage rate and the first fluorescence dispersion degree, normalization is first performed. For example, the coverage rate is mapped from 0% - 100% to the interval of 0 - 1, and the dispersion degree is normalized to the same range according to the maximum and minimum values. Subsequently, weights are set in an empirical or data-driven manner. For example, the weight of the coverage rate is 0.6, and the weight of the dispersion degree is 0.4. Calculate the first fluorescence index = 0.6 × normalized coverage rate + 0.4 × normalized dispersion degree. For example, if the normalized coverage rate of a certain user is 0.7 and the dispersion degree is 0.5, then the first fluorescence index is 0.62. Then, this fluorescence index is used as the first risk index for the first oral disease (such as dental caries), and according to the preset corresponding relationship (such as 0 - 0.3 is low risk, 0.3 - 0.6 is medium risk, above 0.6 is high risk), it is determined that the user is at a high risk level, and the target oral risk decision is output in combination with this risk level, such as suggesting that the user follow up recently or strengthen daily oral cleaning. Among them, the corresponding relationship between the first oral disease and the first risk index can be a risk threshold or a risk model obtained through clinical data statistical analysis.

[0083] Through the above process, the complex multi-dimensional image feature information can be transformed into a unified and interpretable risk quantification index, simplifying the subsequent risk assessment and clinical decision-making processes, and at the same time improving the objectivity and consistency of the assessment results.

[0084] In some embodiments, after collecting the first fluorescence characteristics of the first fluorescence image and combining with the first oral disease to form the target oral risk decision of the target user, it further includes: Obtain the first predetermined amplification curve of the first bacterial community; with the first predetermined amplification curve as a constraint, combine the first fluorescence characteristics to obtain the predicted fluorescence characteristics of the first bacterial community at the target time; analyze the predicted fluorescence characteristics to obtain a predicted risk index, and form the target oral risk prediction of the target user.

[0085] Specifically, the first predetermined amplification curve refers to the growth trend or fluorescence signal change model of specific oral bacteria (such as Streptococcus mutans, Porphyromonas gingivalis, etc.) under specific environmental conditions. Preferably, this first predetermined amplification curve is obtained by fitting experimental data and can reflect the reproduction rate or metabolic activity change of the bacterial community in the time dimension. The predicted fluorescence characteristics are the fluorescence image characteristics (such as coverage rate, dispersion degree, etc.) at a future time point deduced by combining the bacterial community amplification model on the basis of the current fluorescence image characteristics, and are used to predict the development trend of the lesion.

[0086] Specifically, the predicted risk index is the future disease risk level calculated based on the above-mentioned predicted fluorescence characteristics, which is a time extension of the current risk index (i.e., a prospective risk index). By obtaining the predicted risk index, it is possible to further provide the trend of oral health changes in the future based on the current risk assessment, which has forward-looking and intervention guiding significance.

[0087] Specifically, after completing the current fluorescence image risk assessment, the microbial amplification model related to the current user is further called. For example, based on the saliva sample or historical data, it is identified that the main pathogenic bacterium of this user is Porphyromonas gingivalis, and its fluorescence signal amplification curve under standard conditions is called (such as the fluorescence coverage rate exponentially increases with time, the initial value is the current coverage rate of 0.62, and the growth rate is +0.05 per day). Then, taking this amplification curve as a constraint condition, combined with the first fluorescence characteristics extracted from the current fluorescence image, the predicted fluorescence coverage rate and dispersion at several future time points (such as 3 days and 7 days later) are calculated, forming the predicted fluorescence characteristics. Assuming that the predicted coverage rate after 3 days is 0.77 and the dispersion is 0.58, the predicted fluorescence index is 0.70 through normalized weighted calculation.

[0088] Furthermore, according to the mapping relationship between the fluorescence index and oral diseases, it is determined that this user will be in a high-risk state after 3 days, and the corresponding target oral risk prediction result is output, such as: it is expected that the risk of gingivitis will increase after 3 days, and it is recommended to strengthen the use of dental floss and have a reexamination.

[0089] Through the above process, predicting the future lesion risk based on the microbial amplification trend can identify potential health threats in advance, and thus help improve the timeliness and scientificity of individualized prevention and intervention.

[0090] In some implementation manners, taking the first predetermined amplification curve as a constraint, and combining the first fluorescence characteristics to obtain the predicted fluorescence characteristics of the first microbial community at the target time includes: Performing polynomial fitting on the first predetermined amplification curve to obtain a first fitting formula; inputting the target time into the first fitting formula to obtain a first fitting value; using the first fitting value as the first predicted amplification coefficient to perform weighted adjustment on the first fluorescence characteristics to obtain the predicted fluorescence characteristics.

[0091] Specifically, first, extract the fluorescence response values of the first microbial community (such as Streptococcus mutans) at multiple time points from experiments or historical data, and perform polynomial fitting on these data to obtain a fitting function; then, substitute the target time (such as 3 days in the future) into the fitting formula to obtain the predicted fluorescence response value, and perform a ratio operation on this fitting value and the initial value to obtain the first predicted amplification coefficient.

[0092] Subsequently, the current fluorescence feature is weighted and adjusted according to the first predicted amplification coefficient (normalized if it exceeds the maximum value), and the adjusted feature will be used as the predicted fluorescence feature at the target time point for subsequent risk index calculation.

[0093] Through the above process, it is possible to deduce the change of the image feature at a future time point based on the current fluorescence image feature while considering the growth kinetics of the flora, thereby enhancing the biological rationality and temporal continuity of the prediction model and providing a clearer time window and risk basis for subsequent intervention strategies.

[0094] In summary, the oral risk assessment method based on image recognition provided by the present invention has the following technical effects: By performing image block segmentation on the target oral image of the target user collected by a medical imaging device to construct a set of target image blocks; activating a lesion recognizer to extract and recognize features of any first image block in the set of target image blocks to obtain a first lesion recognition result; if the first lesion recognition result meets a predetermined recognition constraint, matching to obtain a corresponding first oral disease, and searching for a first flora type set corresponding to the first oral disease in a preset oral disease flora database; performing fluorescence imaging on the target first flora in the first flora type set based on a fluorescence microscope to obtain a first fluorescence image; extracting the first fluorescence feature of the first fluorescence image and performing risk inference in combination with the first oral disease to form a target oral risk decision result of the target user, thereby achieving the technical effects of multi-dimensional image comprehensive analysis, improving the accuracy and efficiency of the assessment.

[0095] Embodiment 2, as Figure 2 is a schematic structural diagram of the oral risk assessment system based on image recognition of the present invention. For example, Figure 1 The flowchart of the oral risk assessment method based on image recognition in the present invention can be implemented by a structure such as Figure 2 shown.

[0096] Based on the same concept as the oral risk assessment method based on image recognition in the above embodiment, the oral risk assessment system based on image recognition provided by the present invention further includes: An image block segmentation module 11, configured to perform block segmentation on the target oral image of the target user collected by a medical imaging device to obtain a set of target image blocks.

[0097] A lesion recognition module 12, configured to activate a lesion recognizer to analyze and recognize the first image feature of the first image block to obtain a first lesion recognition result, where the first image block refers to any one image block in the set of target image blocks.

[0098] An oral disease matching module 13, configured to match a first oral disease if the first lesion recognition result meets a predetermined constraint, and traverse in an oral disease flora database to obtain a first flora type set corresponding to the first oral disease.

[0099] A fluorescence imaging acquisition module 14, configured to perform fluorescence imaging acquisition on the first flora in the first flora type set through a fluorescence microscope to obtain a first fluorescence image.

[0100] An oral risk decision-making module 15, configured to collect first fluorescence features of the first fluorescence image, and combine with the first oral disease to form a target oral risk decision for the target user.

[0101] In some embodiments, the lesion recognition module 12 includes: A first oral site determination unit, configured to determine a first oral site corresponding to the first image block.

[0102] A first image sample set construction unit, configured to construct a first image sample set of the first oral site, where the first image sample set includes a first healthy image and a first lesion image.

[0103] A training data set construction and lesion recognizer training unit, configured to construct a training data set based on the first healthy image and the first lesion image, and perform supervised machine learning on the training data set to obtain the lesion recognizer.

[0104] A first data set acquisition unit, configured to acquire first feature information of the first healthy image and form a first data set.

[0105] A second data set acquisition unit, configured to acquire second feature information of the first lesion image and form a second data set.

[0106] A training data set construction unit, configured to construct the training data set from the first data set and the second data set.

[0107] In some embodiments, the oral disease matching module 13 includes: A healthy oral sample acquisition unit, configured to acquire a healthy oral sample, and the healthy oral sample has a healthy flora type set.

[0108] A first oral sample acquisition unit, configured to acquire a first oral sample of the first oral disease, and the first oral sample has a disease flora type set.

[0109] A flora type set comparison unit, configured to compare the healthy flora type set with the disease flora type set to obtain the first flora type set of the first oral disease.

[0110] An oral disease flora database construction unit for constructing the oral disease flora database according to the correspondence between the first oral disease and the first flora type set.

[0111] In some embodiments, the fluorescence imaging acquisition module 14 includes: A first antibody acquisition unit for acquiring a first antibody corresponding to the first flora.

[0112] A fluorescence labeling unit for performing fluorescence labeling on the first antibody to obtain a first labeling result.

[0113] A first fluorescence image acquisition unit for acquiring the first fluorescence image through the fluorescence microscope, where the first fluorescence image refers to the combined imaging of the first labeling result and the first flora.

[0114] In some embodiments, the oral risk decision-making module 15 includes: A first antibody acquisition unit for acquiring a first antibody corresponding to the first flora.

[0115] A fluorescence labeling unit for performing fluorescence labeling on the first antibody to obtain a first labeling result.

[0116] A first fluorescence image acquisition unit for acquiring the first fluorescence image through the fluorescence microscope, where the first fluorescence image refers to the combined imaging of the first labeling result and the first flora.

[0117] In some embodiments, the oral risk decision-making module 15 further includes: A first predetermined amplification curve acquisition unit for acquiring a first predetermined amplification curve of the first flora.

[0118] A predicted fluorescence feature calculation unit for obtaining the predicted fluorescence feature of the first flora at the target time by combining the first fluorescence feature with the constraint of the first predetermined amplification curve.

[0119] A target oral risk prediction formation unit for analyzing the predicted fluorescence feature to obtain a predicted risk index and forming a target oral risk prediction for the target user.

[0120] In some implementation manners, the execution steps of the predicted fluorescence feature calculation unit in the oral risk decision-making module 15 include: Performing polynomial fitting on the first predetermined amplification curve to obtain a first fitting formula.

[0121] Inputting the target time into the first fitting formula to obtain a first fitting value.

[0122] Use the first fitting value as the first predicted amplification coefficient to perform weighted adjustment on the first fluorescence feature to obtain the predicted fluorescence feature.

[0123] In some embodiments, the first fluorescence feature at least includes a first fluorescence coverage rate and a first fluorescence dispersion degree.

[0124] Embodiment 3: The present application provides a computer-readable storage medium, which can be used to store software programs, computer-executable programs, and modules, such as the program instructions / modules corresponding to the oral risk assessment method based on image recognition in the embodiments of the present invention. By running the software programs, instructions, and modules stored in the computer-readable storage medium, the above-mentioned oral risk assessment method based on image recognition can be implemented.

[0125] It should be understood that the key point of the embodiments mentioned in this specification lies in their differences from other embodiments. The specific embodiments in the foregoing Embodiment 1 are equally applicable to the oral risk assessment system based on image recognition described in Embodiment 2. For the sake of brevity of the specification, no further elaboration will be made here.

[0126] It should be understood that the disclosed embodiments of the present invention and the above descriptions can enable those skilled in the art to implement the present invention using the present invention. At the same time, the present invention is not limited to the above-mentioned part of the embodiments. It should be understood that those of ordinary skill in the art can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. An oral risk assessment method based on image recognition, characterized in that Including: Performing block segmentation on the target oral image of the target user acquired by a medical imaging device to obtain a set of target image blocks; Activating a lesion recognizer to analyze and recognize the first image features of the first image block to obtain a first lesion recognition result, where the first image block refers to any one image block in the set of target image blocks; If the first lesion recognition result meets a predetermined constraint, matching a first oral disease and traversing in an oral disease flora database to obtain a first flora type set corresponding to the first oral disease; Performing fluorescence imaging acquisition on the first flora in the first flora type set through a fluorescence microscope to obtain a first fluorescence image; Collecting the first fluorescence features of the first fluorescence image and combining with the first oral disease to form a target oral risk decision for the target user.

2. The oral risk assessment method based on image recognition according to claim 1, wherein, Before activating the lesion recognizer to analyze and recognize the first image features of the first image block to obtain a first lesion recognition result, including: Determining a first oral site corresponding to the first image block; Constructing a first image sample set of the first oral site, where the first image sample set includes a first healthy image and a first lesion image; Constructing a training data set based on the first healthy image and the first lesion image and performing supervised machine learning on the training data set to obtain the lesion recognizer; Among them, constructing a training data set based on the first healthy image and the first lesion image includes: Obtaining first feature information of the first healthy image and forming a first data set; Obtaining second feature information of the first lesion image and forming a second data set; The first data set and the second data set construct the training data set.

3. The oral risk assessment method based on image recognition according to claim 1, wherein If the first lesion recognition result meets a predetermined constraint, matching a first oral disease and traversing in an oral disease flora database to obtain a first flora type set corresponding to the first oral disease, including: Obtaining a healthy oral sample, and the healthy oral sample has a set of healthy flora types; Obtaining a first oral sample of the first oral disease, and the first oral sample has a set of disease flora types; Comparing the set of healthy flora types with the set of disease flora types to obtain the first flora type set of the first oral disease; Constructing the oral disease flora database according to the corresponding relationship between the first oral disease and the first flora type set.

4. The oral risk assessment method based on image recognition according to claim 1, wherein Performing fluorescence imaging acquisition on the first flora in the first flora type set through a fluorescence microscope to obtain a first fluorescence image, including: Obtaining a first antibody corresponding to the first flora; Performing fluorescence labeling on the first antibody to obtain a first labeling result; Obtaining the first fluorescence image through the fluorescence microscope, where the first fluorescence image refers to the combined imaging of the first labeling result and the first flora.

5. The oral risk assessment method based on image recognition according to claim 1, characterized in that, Collecting the first fluorescence features of the first fluorescence image and combining with the first oral disease to form a target oral risk decision for the target user, including: Performing normalized weighted calculation on the first fluorescence features to obtain a first fluorescence index; Taking the first fluorescence index as the first risk index of the first oral disease; Form the target oral risk decision according to the first correspondence between the first oral disease and the first risk index.

6. The oral risk assessment method based on image recognition according to claim 5, wherein The first fluorescence feature at least includes a first fluorescence coverage rate and a first fluorescence dispersion degree.

7. The oral risk assessment method based on image recognition according to claim 1, wherein After collecting the first fluorescence feature of the first fluorescence image and combining the first oral disease to form the target oral risk decision of the target user, it further includes: Obtain the first predetermined amplification curve of the first bacterial flora; Taking the first predetermined amplification curve as a constraint, combine the first fluorescence feature to obtain the predicted fluorescence feature of the first bacterial flora at the target time; Analyze the predicted fluorescence feature to obtain a predicted risk index, and form the target oral risk prediction of the target user.

8. The oral risk assessment method based on image recognition according to claim 7, wherein Taking the first predetermined amplification curve as a constraint, combine the first fluorescence feature to obtain the predicted fluorescence feature of the first bacterial flora at the target time, including: Perform polynomial fitting on the first predetermined amplification curve to obtain a first fitting formula; Input the target time into the first fitting formula to obtain a first fitting value; Use the first fitting value as a first predicted amplification coefficient to perform weighted adjustment on the first fluorescence feature to obtain the predicted fluorescence feature.

9. An oral risk assessment system based on image recognition, characterized in that, For implementing the oral risk assessment method based on image recognition according to any one of claims 1 to 8, it includes: An image block segmentation module, configured to perform block segmentation on the target oral image of the target user collected by a medical imaging device to obtain a set of target image blocks; A lesion recognition module, configured to activate a lesion recognizer to analyze and recognize the first image feature of the first image block to obtain a first lesion recognition result, where the first image block refers to any one of the set of target image blocks; An oral disease matching module, configured to, if the first lesion recognition result meets a predetermined constraint, match a first oral disease, and traverse in an oral disease flora database to obtain a set of first bacterial flora types corresponding to the first oral disease; A fluorescence imaging acquisition module, configured to perform fluorescence imaging acquisition on the first bacterial flora in the set of first bacterial flora types through a fluorescence microscope to obtain a first fluorescence image; An oral risk decision module, configured to collect the first fluorescence feature of the first fluorescence image and combine the first oral disease to form the target oral risk decision of the target user.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the oral risk assessment method based on image recognition according to any one of claims 1 to 8.

Images