User-individual age prediction method based on dental images and related devices

Abstract

Embodiments of the present application provide a user-individual age prediction method based on dental images and related equipment, belonging to the field of artificial intelligence. The user-uploaded image to be detected is obtained; the content scene compliance of the image to be detected is detected; in the case of compliance, the visual statistical features such as saturation, contrast and average brightness of the image to be detected are extracted, and whether the image to be detected is a dental medical image is determined based on the visual statistical features; in the case of being determined as a dental medical image, the image to be detected is preprocessed, and an age prediction model based on deep learning is called to output an age prediction value and an uncertainty. The present application can intercept invalid images in the open upload scene for users, improve the age prediction efficiency and the stability of online prediction service.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a user-oriented individual age prediction method and related equipment based on dental images. Background Technology

[0002] In forensic identification, criminal investigation, and identification of unidentified corpses, individual age estimation has significant legal implications. Tooth development is highly correlated with age and is less affected by nutrition and environment compared to bone age; therefore, dental age assessment has become the mainstream technical approach for forensic age estimation. Currently, the mainstream dental age assessment methods include the Demirjian staging system, the Willems scoring system, and the Cameriere root opening ratio method. All of these methods are based on manually observing the stage of tooth mineralization and manually converting scores to determine an individual's age.

[0003] In related technologies, deep learning algorithms can be used to evaluate dental images to determine an individual's age. However, deep learning models are currently only in the background testing stage and have not yet been truly deployed to users as applications for individual age prediction based on deep learning analysis of dental images. Furthermore, if a trained deep learning model were deployed to users, it would face problems such as malicious attacks, resulting in low efficiency of user-facing individual age prediction methods. Summary of the Invention

[0004] The main objective of this application is to propose a user-oriented individual age prediction method and related equipment based on dental images, aiming to improve the efficiency of user-oriented individual age prediction.

[0005] To achieve the above objectives, one aspect of this application proposes a user-individual age prediction method based on dental images, comprising the following steps: Obtain the uploaded image to be detected; The content scene compliance is checked on the image to be detected to obtain the content detection result; If the content detection result indicates compliance, visual features are extracted from the image to be detected to obtain image visual features; The image to be detected is classified and determined as a dental medical image based on the visual features of the image, and the classification result is obtained. If the classification result indicates that the image to be detected is a dental medical image, a deep learning-based age prediction model is used to predict the age of the image to be detected, thereby obtaining an age prediction result. In some embodiments, the user-individual age prediction method based on dental images further includes the following steps: A unique identifier is randomly generated for the image to be detected; The image to be detected is renamed according to the unique identifier and the original file extension of the image to be detected to obtain an image file name, and then the image file name is used to manage the image to be detected.

[0006] In some embodiments, the user-individual age prediction method based on dental images further includes the following steps: If the content detection result indicates non-compliance, the image to be detected is deleted and an error message is returned to the client.

[0007] In some embodiments, the image visual features include image saturation, image contrast, and average image brightness. The step of classifying the image to be detected as a dental medical image based on the image visual features to obtain a classification result includes the following steps: Determine whether the preset detection conditions are met based on the visual features of the image; If any of the preset detection conditions are met, the image to be detected is determined to be a non-dental medical image. The preset detection conditions include an average image saturation greater than a first threshold, a standard deviation of image contrast less than a second threshold, and an average image brightness greater than a third threshold.

[0008] In some embodiments, the first threshold, the second threshold, and the third threshold are determined by the following steps: Collect dental medical imaging samples and non-dental imaging samples; Statistical analysis of the visual features of the dental medical image samples and the non-dental image samples was performed respectively to obtain the first data distribution features of the dental medical image samples and the second data distribution features of the non-dental image samples. Based on the first data distribution characteristics and the second data distribution characteristics, a first threshold, a second threshold, and a third threshold are determined.

[0009] In some embodiments, the step of using a deep learning-based age prediction model to predict the age of the image to be detected and obtaining the age prediction result includes the following steps: The image to be detected is input into an age prediction model based on a convolutional neural network to obtain the predicted age value and uncertainty. The age prediction interval is determined based on the predicted age value and the uncertainty. An age prediction result is formed based on the predicted age value, the uncertainty, and the predicted age range.

[0010] In some embodiments, the age prediction model is trained through the following steps: Obtain a dental medical image sample dataset, which includes normal dental samples and dental disease samples; The dental medical image sample dataset is divided into a first sample dataset and a second sample dataset according to gender. The initialized age prediction model was trained using the first sample dataset and the second sample dataset respectively, resulting in two age prediction models for different genders.

[0011] To achieve the above objectives, another aspect of this application proposes a user-oriented individual age prediction system based on dental images, comprising: The image acquisition module is used to acquire the uploaded image to be detected; The content compliance detection module is used to perform content scene compliance detection on the image to be detected and obtain the content detection result; The visual feature extraction module is used to extract visual features from the image to be detected, provided that the content detection result indicates compliance, to obtain the image visual features; The dental X-ray authenticity determination module is used to classify and determine whether the image to be detected is a dental medical image based on the visual features of the image, and obtain the classification and determination result; An age prediction module is used to predict the age of the image to be detected using a deep learning-based age prediction model when the classification result indicates that the image to be detected is a dental medical image, and to obtain an age prediction result.

[0012] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0013] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0014] The embodiments of this application include at least the following beneficial effects: This application provides a method, system, electronic device, and program product for user-oriented individual age prediction based on dental images. After the platform obtains the uploaded image to be detected, it first performs content scene compliance detection on the image to be detected. After passing the compliance detection, visual features are extracted from the image to obtain image visual features. Then, based on the image visual features, the image to be detected is classified to determine whether it is a dental medical image, and a classification result is obtained. If the classification result indicates that the image to be detected is a dental medical image, a deep learning-based age prediction model is used to predict the age of the image to obtain the age prediction result. This solution analyzes the legality and authenticity of the images to be detected uploaded to the platform in advance. If it is a legal dental medical image, then the deep learning model is used for age prediction. This can intercept uploaded invalid non-dental images from calling the age prediction model, reducing the risk of misjudgment caused by non-target images entering the age prediction model, thereby improving the efficiency of user-oriented individual age prediction and the stability of online prediction services. It is suitable for application scenarios where users can upload their own images. Attached Figure Description

[0015] Figure 1 This is a flowchart of a user-individual age prediction method based on dental images provided in an embodiment of this application; Figure 2 This is a schematic diagram of the image upload interaction interface provided in an embodiment of this application; Figure 3 This is a schematic diagram of the EfficientNet-B4 network structure provided in the embodiments of this application; Figure 4 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0018] Before providing a detailed description of the embodiments of this application, some of the nouns and terms involved in the embodiments of this application will be explained first. The nouns and terms involved in the embodiments of this application are subject to the following interpretations.

[0019] Oral panoramic imaging (OPG): refers to the panoramic X-ray image of the oral cavity, which can simultaneously display all teeth, roots and part of the jawbone structure of the upper and lower jaws. It is the main input image type processed in the embodiments of this application.

[0020] Deep learning: A machine learning method based on multi-layer neural networks that automatically learns data feature representations.

[0021] Convolutional Neural Network (CNN): A deep learning network structure suitable for processing image data, which automatically extracts image features through convolution, pooling, and nonlinear transformations.

[0022] EfficientNet-B4: A convolutional neural network model, which is a specific model version in the EfficientNet series.

[0023] UUID: Universally Unique Identifier, used in this invention for anonymous renaming of uploaded files.

[0024] In related technologies, the degree of tooth development is highly correlated with age and is less affected by factors such as nutrition, environment, and endocrine function compared to bone age. Therefore, dental age assessment has become one of the important technical routes in forensic age estimation. With the development of medical imaging digitization and artificial intelligence technology, using deep learning models to directly extract age-related features from dental images and automatically output age prediction results has become an important development direction for dental age assessment technology. However, current technologies still generally have the following problems: First, most research work is mainly at the stage of model experimental verification, lacking an online service implementation path that integrates with the client; Second, many methods only perform basic format judgment on images at the model input end, lacking compliance detection and dental radiograph authenticity discrimination mechanisms for online systems, which can easily lead to invalid images or non-dental radiographs entering the model inference process, affecting the model inference efficiency; Third, current models only output a single age value, resulting in poor interpretability and reliability.

[0025] In view of this, this application provides a method and related equipment for predicting individual user age based on dental images. This solution is an automatic dental age assessment system method for practical online application scenarios. It mainly improves the efficiency of age prediction by reducing the probability of invalid input consuming inference resources through dental authenticity verification of images uploaded by the client. In addition to outputting the predicted age value to the user, this solution also outputs the uncertainty of the prediction, which can improve the interpretability and reliability of the prediction results. This solution integrates client image uploading, platform content security detection, local dental X-ray authenticity judgment, image preprocessing, deep learning model prediction, and result return into a complete engineering technology chain, thereby providing users with a stable age prediction platform.

[0026] The method for predicting individual user age based on dental images provided in this application relates to the field of artificial intelligence technology. This method can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or in-vehicle terminal, but is not limited to these. The server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing the method for predicting individual user age based on dental images, but is not limited to the above forms.

[0027] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0028] Figure 1 This is an optional flowchart of a user-individual age prediction method based on dental images provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S101 to S104.

[0029] S101, Obtain the uploaded image to be detected; S102, Perform content scene compliance detection on the image to be detected to obtain the content detection result; S103, if the content detection result indicates compliance, visual features are extracted from the image to be detected to obtain image visual features; S104, classify the image to be detected as a dental medical image based on its visual features, and obtain the classification result; S105, if the classification result indicates that the image to be detected is a dental medical image, a deep learning-based age prediction model is used to predict the age of the image to be detected, and the age prediction result is obtained.

[0030] In step S101 of some embodiments, the solution of this application is implemented based on a server-side and client-side architecture. The method for predicting individual user age based on dental images is mainly implemented on the server side, which can be a Flask server, a lightweight backend web service framework based on the Python language. It is understood that the solution of this application can be deployed on platforms such as WeChat mini-programs, apps, or browsers. Subsequent embodiments will use WeChat mini-programs as an example. The server can provide the client with, for example... Figure 2 The image upload interface shown is used to acquire the image to be detected. Exemplarily, the system's image acquisition module includes a mini-program interface unit, an image selection unit, a user parameter input unit, and a submission control unit. The user selects the dental image file to be evaluated via the WeChat mini-program, fills in or inputs the necessary parameters, and then submits it to the server. In the current embodiment, in addition to the image file, the prediction interface can also receive openid and gender parameters, where openid is used for security review request association, and gender is used as an auxiliary parameter for model invocation. The image acquisition module's input is the dental image file selected by the user and related form fields, and its output is a packaged upload request.

[0031] In step S102 of some embodiments, the compliance detection of the content scene of the image to be detected can be achieved through a deep learning image classification model for detecting illegal content built by the system detection model, or by requesting an external interface to achieve illegal content detection. Performing authenticity detection on the image only after the image to be detected is found to be legal can improve the stability and security of the online prediction service. In this embodiment, platform content security detection is performed first, followed by local dental X-ray authenticity determination; only after both pass does the model prediction begin.

[0032] In steps S103 to S104 of some embodiments, in online deployment scenarios, the files uploaded by users may not be panoramic dental images, but rather webpage screenshots, chat screenshots, personal photos, or other non-compliant images. Directly inputting these images into the model for inference would severely impact the reliability of the results and the efficiency of the model's inference. Therefore, in this embodiment, before inputting the image to be detected into the model for age prediction, visual features are extracted from the image to be detected to obtain visual features. Then, based on these visual features, the image to be detected is classified to determine whether it is a dental medical image, resulting in a classification result. Specifically, the visual features can be basic visual statistical features such as contrast, brightness, and saturation, or features such as shape extracted based on edge operators. These visual features can be input into a trained classification model, or threshold comparison can be used to classify whether it is a dental medical image.

[0033] In step S105 of some embodiments, if the classification result indicates that the image to be detected is a dental medical image, a deep learning-based age prediction model is then used to predict the age of the image to be detected, thereby improving the stability of the online prediction service and increasing the efficiency of age prediction for users. The age prediction model can employ networks such as ResNet, Inception, Vision Transformer, and EfficientNet-B4.

[0034] In some embodiments, the method further includes verifying and saving the uploaded image file. This function is implemented by a file verification and upload module, which includes a file format verification unit, a request field verification unit, a file size limit unit, and an upload and save unit. When the server receives a predict request, it first checks whether the request contains a file field; if the file field does not exist, an error message is immediately returned. If the file name is empty, "No file selected" is returned. If the file extension is not in the whitelist set, "Disallowed file type" is returned, such as only allowing png / jpg / jpeg / gif files to be uploaded, and directly rejecting other types to reduce the risk of malicious files. In this embodiment, allowed extensions include png, jpg, jpeg, and gif. In addition, the server can configure Flask to limit the maximum size of uploaded files to 16MB to prevent abnormally large files from consuming resources. This embodiment can ensure that the uploaded object is an image file that the system can process and complete basic legality screening before entering subsequent processes.

[0035] In some embodiments, the user-individual age prediction method based on dental images according to this application may also include, but is not limited to, the following steps: S201, Randomly generate a unique identifier for the image to be detected; S202: Rename the image to be detected according to the unique identifier and the original extension of the image to be detected to obtain the image file name, and then use the image file name to manage the image to be detected.

[0036] In this embodiment, the renaming process described above can be implemented by a file anonymization module, which includes a file renaming unit, a path concatenation unit, and a temporary storage unit. After the file passes type verification, the server generates a new filename using a random unique identifier combined with the original extension. In this embodiment, the image file (i.e., the image to be detected) is renamed by concatenating the original extension with a UUID (Universally Unique Identifier), and the original filename is processed using `secure_filename`, thereby avoiding filename conflicts, path injection, and the direct exposure of privacy information by the original filename. The renamed file is stored in a preset upload directory on the server. This embodiment improves the anonymity, security, and project manageability of uploaded files.

[0037] In some embodiments, the user-individual age prediction method based on dental images according to this application may also include, but is not limited to, the following steps: S301: If the content detection result indicates non-compliance, delete the image to be detected and return an error message to the client.

[0038] In this embodiment, the image is only tested for authenticity if it is found to be valid; otherwise, the image is deleted from the system and an error message is returned to the client, which further improves the stability and security of the online prediction service.

[0039] For example, the method in this embodiment can be a lightweight mobile application on the WeChat platform, used to upload dental images and receive prediction results. The WeChat platform uses an OpenID as a unique identifier for users; in this embodiment, it is used to associate with the content security detection interface request. The system's compliance detection is implemented by the platform's content security detection module, which includes an interface call credential acquisition unit, a media address construction unit, a content security review request unit, and a review result processing unit. The server obtains an access_token by calling the WeChat interface using the application identifier and application key; then, it constructs an accessible media_url for the saved image and sends a review request to the WeChat content security detection interface carrying the user's OpenID. If the platform returns success, it continues to the subsequent steps; if it returns failure, it immediately deletes the saved file and terminates the prediction process, while simultaneously returning the corresponding error information to the client. This module can realize compliance detection of uploaded content, preventing image content that does not conform to platform specifications from entering the prediction process.

[0040] This embodiment first performs platform content security checks, then verifies the authenticity of local dental X-rays. Only when both checks pass does the model proceed to prediction, as detailed below: Platform content security detection: After saving the image, the server constructs an accessible media_url, carries the openid, and calls the WeChat wxa / media_check_async asynchronous content security detection interface; if the return errcode!=0, the detection is deemed to have failed and the process is terminated, while the saved file is deleted and an error message is returned; if successful, the trace_id can be obtained for audit tracing.

[0041] Local dental X-ray authenticity verification: After the platform approves the image, a "is it a dental X-ray?" verification is performed: the image is scaled to 128×128 and the average saturation, grayscale contrast standard deviation, and average brightness of the RGB channel differences are calculated; if saturation > threshold / contrast < threshold / brightness > threshold, it is determined to be a non-dental image and the file is deleted, and an "image verification failed" message is returned.

[0042] The execution order of the dual detection is as follows: first, the WeChat content security detection is completed, and then the local dental X-ray authenticity detection is performed. If either step fails, the prediction process is interrupted and an error code is returned, thus forming a dual threshold protection for compliance and authenticity.

[0043] In some embodiments, image visual features include image saturation, image contrast, and average image brightness. Step S103 may include, but is not limited to, the following steps: S401, determine whether the preset detection conditions are met based on the visual features of the image; S402, under any preset detection condition, determine the classification result of the image to be detected as a non-dental medical image; S403, multiple preset detection conditions include image average saturation greater than the first threshold, image contrast standard deviation less than the second threshold, and image average brightness greater than the third threshold.

[0044] In this embodiment, the authenticity detection of teeth in the image to be detected is implemented by a local dental X-ray authenticity determination module. This module includes an image scaling unit, a color statistics unit, a grayscale statistics unit, a rule-based determination unit, and a failure rollback unit. After passing the platform's content security check, the uploaded image is further rapidly determined locally to possess the visual statistical features of a dental X-ray. This determination process is not based on the user's subjective choice, nor is it simply inferred from the file extension; instead, it constructs a rule-based determination logic based on the imaging features of the dental X-ray. Specifically, the image to be detected is first uniformly scaled to 128×128 pixels to reduce the subsequent statistical calculation overhead; then, the image is converted to RGB format and the pixel matrix of each channel is extracted. Next, the following image visual features are extracted: (1) Image saturation, specifically image average saturation: the average difference between the maximum and minimum values ​​of the RGB three channels of each pixel is used to characterize the overall color intensity of the image; (2) Image contrast, specifically grayscale contrast standard deviation: the standard deviation of grayscale pixels is calculated after grayscale conversion, which represents the richness of grayscale levels and structural information of the image; (3) Average brightness of the image: The average value of the gray matrix represents the overall brightness level of the image.

[0045] The discrimination rules are based on the typical characteristics of panoramic dental X-rays, which usually have "approximate grayscale, weak color information, certain grayscale levels, and should not be too bright overall." The specific preset detection conditions are as follows: (1) When the average saturation of the image is greater than the preset threshold (e.g., 30), it indicates that the image has strong comprehensive color information and is more likely to be a life photo, page screenshot or other color image, which does not conform to the characteristics of dental X-ray film; setting this preset detection condition can exclude input samples that deviate significantly from the distribution of medical grayscale images by color dispersion.

[0046] (2) When the standard deviation of image grayscale contrast is less than the preset threshold (e.g., 30), it indicates that the image lacks sufficient grayscale levels and structural information, and may be a blank image, an excessively blurred image, or other invalid images; setting this preset detection condition can use the global grayscale dispersion of the image to eliminate low-quality inputs that lack effective tooth structure information.

[0047] (3) When the average brightness of the image is higher than the preset threshold (e.g., 200), it indicates that the image is too bright overall and is more likely to be a white-background document screenshot, chat screenshot, or an image with abnormal exposure. Setting this preset detection condition can reduce misjudgments caused by white-background interface images, document screenshots, and overexposed images by excluding images with significantly higher overall brightness.

[0048] If any of the above conditions are met, the image is determined not to be a dental X-ray suitable for dental age assessment, the prediction process is terminated, the temporary file on the server is deleted, and a "Image verification failed" message is returned.

[0049] In the embodiments of this application, the first threshold, the second threshold, and the third threshold used in the process of determining the authenticity of dental medical images can be determined based on the statistical distribution of samples and the needs of engineering screening.

[0050] In one example, the threshold can be determined through pre-experimental statistical analysis and engineering optimization based on the statistical distribution of images in the training dataset and actual uploaded samples. That is, first, a judgment approach is pre-defined based on the image characteristics of dental X-rays; then, observation and adjustments are made on existing samples; finally, a threshold that can effectively distinguish between "valid dental X-rays" and "non-dental X-rays / screenshots / color images" is selected. Threshold selection methods can include sample statistical methods, ROC / optimal split point methods, and quantile methods, as detailed below: The sample statistics method involves statistically analyzing the distribution of indicators from a valid set of dental X-ray samples and a non-dental sample set (life photos, screenshots, web page images, interface images, etc.). The following are calculated separately: average saturation distribution, grayscale contrast distribution, and average brightness distribution. Then, the overlap between the two distributions is observed, and a threshold that maximizes the discriminative power is selected.

[0051] The ROC / optimal split point method involves setting up a binary classifier based on the aforementioned preset detection rules, and calculating sensitivity, specificity, Youden index, and AUC for different threshold combinations. The threshold combination with the best overall performance is then selected.

[0052] The quantile method involves taking the following from the set of valid dental radiograph samples: the 95th percentile for saturation as the upper limit threshold, the 95th percentile for luminance as the upper limit threshold for luminance, and the 5th percentile for contrast as the lower limit threshold.

[0053] The detection method for determining whether an image to be detected is a dental medical image in this embodiment does not require the additional collection of a large amount of real and fake dental X-ray data, nor does it require retraining a pre-classification network. It only requires a small amount of pixel statistical calculation, resulting in extremely low computational cost and high efficiency, with almost no additional inference cost on the server side. Furthermore, the preset detection conditions make this detection method interpretable.

[0054] In some embodiments, the first threshold, the second threshold, and the third threshold may be determined by, but not limited to, the following steps: S501, collects dental medical imaging samples and non-dental imaging samples; S502, Perform statistical analysis on the visual features of the dental medical image samples and the non-dental image samples respectively to obtain the first data distribution characteristics of the dental medical image samples and the second data distribution characteristics of the non-dental image samples. S503, determine the first threshold, the second threshold, and the third threshold based on the first data distribution characteristics and the second data distribution characteristics.

[0055] In this embodiment, the first, second, and third thresholds can be determined through pre-experimental statistics and engineering optimization. Specifically, dental medical image samples and non-dental image samples are collected separately, and their average saturation, grayscale contrast, and average brightness distribution characteristics are statistically analyzed. A threshold combination that can effectively distinguish between the two types of samples is then selected as the rule parameters. The thresholds can be further adaptively calibrated under different devices or deployment environments. In this way, the saturation threshold, contrast threshold, and brightness threshold used to distinguish whether an image is a tooth can be quickly and accurately determined.

[0056] In some embodiments, step S104 may include, but is not limited to, the following steps: S601, Input the image to be detected into the age prediction model based on the convolutional neural network to obtain the age prediction value and uncertainty; S602, Further determine the age prediction interval based on the predicted age value and uncertainty; S603, based on the predicted age value, uncertainty, and age prediction interval, forms the age prediction result.

[0057] In some embodiments, before inputting the image to be detected into the age prediction model, it can be preprocessed. This preprocessing is implemented by an image preprocessing module, which includes a fixed-coordinate cropping unit, a uniform-size scaling unit, a grayscale three-channel conversion unit, and a tensor input unit. After the uploaded image passes content detection and authenticity assessment, it enters the image preprocessing module. First, the image to be detected is cropped within a preset coordinate range, such as using the top-left corner coordinates (50, 50) and the bottom-right corner coordinates (2650, 1350) as the cropping boundaries to preserve the main dental arch area and remove most irrelevant background information. Then, the cropped image is uniformly scaled to 380×380 pixels to match the input size of the subsequent deep learning model. Next, the image is converted into a three-channel grayscale image, i.e., the grayscale image is copied into three channels to adapt to the input format of the convolutional neural network pre-trained structure. Finally, the image is converted into a Tensor for model inference. Understandably, in the actual age prediction process, the preprocessing mainly includes cropping, scaling, grayscale three-channel conversion, and tensor transformation. During model training, data augmentation strategies such as normalization, rotation, flipping, brightness perturbation, and noise enhancement can be further added on top of this. This embodiment can reduce background noise in the original panoramic image, improve the model's focus on key features of the dental arch region, and ensure a standardized input format.

[0058] Furthermore, this embodiment addresses the issue of high resolution in panoramic dental images and the loss of detail due to direct scaling. It designs various cropping strategies, including cropping to the original image proportion, cropping by extracting the dental region, scaling while maintaining aspect ratio, and cropping with partial dental enhancement. The optimal cropping strategy and model combination are selected based on validation set performance. After determining the optimal cropping strategy, it is embedded at the inference end for online service, ensuring consistent input distribution between training and deployment and reducing domain bias.

[0059] In some embodiments, age prediction of the image to be detected is achieved by a deep learning prediction module, including an EfficientNet-B4 backbone network, a dual-output regression head, and an uncertainty constraint unit. This is achieved using... Figure 3 The EfficientNet-B4 network shown is used as the feature extraction backbone, and the number of output categories is set to 2. After the model's forward propagation, the first dimension of the output tensor is used as the age prediction value μ, and the second dimension is used as the uncertainty parameter σ, reflecting the dispersion of the prediction result. To ensure that σ is always positive, this embodiment uses the Softplus function to transform it and adds a minimum bias to improve numerical stability. Furthermore, the age prediction interval can be determined as [μ-kσ, μ+kσ] based on the age prediction value and uncertainty, where k is a preset coefficient. The system outputs a composite result of the age prediction value, uncertainty, and age prediction interval to the user, improving the judicial interpretability of the age prediction.

[0060] In some embodiments, the model can divide prediction uncertainty into two parts: cognitive uncertainty and random uncertainty. Cognitive uncertainty stems from insufficient model parameter learning, while random uncertainty arises from individual differences in tooth development and annotation noise. During the inference phase, multiple forward propagations are performed using Monte Carlo Dropout to obtain multiple sets of μ output sequences, whose standard deviation is used as an estimate of cognitive uncertainty. Simultaneously, the mean of the model output σ is used as an estimate of random uncertainty, thereby obtaining an interpretable and auditable legal-grade uncertainty structure. The specific process is as follows: During the model inference phase, Dropout activation is maintained and forward propagation is performed N times (e.g., 100 times) to obtain the age prediction sequence {μ1…μ n} and uncertainty sequence {σ1…σ n}; The estimated age value output is: ; Cognitive uncertainty is: ; () represents the standard deviation of the age prediction series; The random uncertainty is: ; This represents the median of the uncertainty series; The uncertainty of the output is: .

[0061] In some embodiments, the age prediction model in step S104 can be trained through, but is not limited to, the following steps: S701, Obtain the dental medical image sample dataset, which includes normal dental samples and dental disease samples; S702, the dental medical image sample dataset is divided into the first sample dataset and the second sample dataset according to gender; S703, the initialized age prediction model is trained based on the first sample dataset and the second sample dataset respectively, resulting in two age prediction models for different genders.

[0062] In this embodiment, two age prediction models are trained for men and women. After receiving the dental images and gender parameters uploaded by the client, the server selects the corresponding deep learning model weight file to perform inference based on the gender parameters. Specifically, male samples are loaded with the male dental age assessment model weight file, and female samples are loaded with the female dental age assessment model weight file. The male and female models are trained independently based on samples from their respective genders to accommodate the differences in tooth mineralization, root closure, and developmental timing between different genders.

[0063] This embodiment incorporates images of dental disease cases during training and validation. By comparing the error distribution of normal samples with those of dental disease samples, it verifies that the model still possesses stable predictive performance in cases of dental disease. Compared to traditional dental age assessment methods, which are only applicable to individuals with relatively complete dentition development and no obvious lesions, this embodiment significantly expands the scope of applicability, enabling the system to cover examinees with caries, missing teeth, abnormal development, or other dental diseases, thereby improving the usability of forensic age determination in complex cases.

[0064] In some embodiments, the negative log-likelihood (NLL) loss function or the heteroscedastic regression loss function can be used to jointly optimize μ and σ during the model training phase. The negative log-likelihood (NLL) loss function is expressed as: ; Where y is the true age, this optimization strategy can achieve an adaptive balance between fitting error and confidence calibration.

[0065] In some embodiments, the overall process of model training is as follows: the dataset is randomly divided into training set, validation set and test set according to gender, with a ratio of 7:1.5:1.5, to ensure that the distribution of samples of different genders is consistent; The training set employs random scaling (0.9–1.1), random rotation (±30°), and brightness and contrast perturbations to enhance generalization ability, while the validation / test set is only standardized. An early stopping strategy is introduced: training stops when the validation set performance no longer improves within a preset number of rounds to prevent overfitting. K-fold cross-validation is used to collect all test outputs for uncertain frameworks to statistically measure coverage (e.g., ±1σ / ±2σ / ±3σ) and calibrate performance, forming legally interpretable evidence of predicted interval coverage.

[0066] In some embodiments, the method further includes a function for calling the age prediction model for inference and result encapsulation and return. This function is implemented by a server inference and result return module, which includes a device selection unit, a model loading unit, an inference execution unit, a result encapsulation unit, an exception handling unit, and a temporary file cleanup unit. The system preferably automatically detects whether an available GPU exists in the current environment; if so, GPU inference is used; otherwise, it reverts to CPU mode. Before model inference, the server places the model in evaluation mode and disables gradient calculation. The preprocessed single image is expanded in batch dimension and then input into the model to obtain the age prediction value and uncertainty result, which are then rounded to two decimal places. The server finally returns the following standardized JSON structure: (1) On success, return: code=0, and include the predicted age and uncertainty in the data field; (2) If it fails, return code=1 and return the corresponding error message.

[0067] In some embodiments, regardless of whether the prediction succeeds or fails, the server attempts to delete the uploaded temporary image file within the code block; if content security detection fails or local authenticity determination fails, deletion will also be performed in advance, thus forming a closed loop of file destruction under the principle of minimum retention. This embodiment improves resource management capabilities and privacy protection levels in online deployment scenarios.

[0068] According to some embodiments of this application, the overall process of the method in this application is as follows: S1. The mobile device obtains the user's openid through the / getOpenid interface; S2. The user selects a dental image and submits a / predict request in the mini-program. S3. The server verifies the file fields, file name, file type, and file size; S4. The server renames and saves the file using a UUID. S5. The server calls the WeChat content security detection interface to perform compliance review on the image; S6. If the content security check passes, proceed with the local dental X-ray authenticity verification. S7. If the authenticity check passes, then perform fixed ROI cropping, 380×380 scaling, grayscale three-channel conversion, and Tensorization. S8. Input the processed image into the EfficientNet-B4 model and output the age prediction value and uncertainty parameter. S9. Return the results to the mobile device in JSON format; S10. Delete temporary image files on the server.

[0069] Taking an age prediction WeChat mini-program as an example, a complete embodiment of the solution in this application is as follows: The server uses the Flask framework to deploy two HTTP interfaces: / getOpenid and / predict. The / getOpenid interface obtains the openid by exchanging a WeChat login code; if the code parameter is missing, an error message is returned.

[0070] The / predict interface receives image files and form parameters, and performs security checks, authenticity determination, preprocessing, model inference, and result return.

[0071] After a user uploads an image, the server first obtains the interface call credential based on the application identifier and application key, and then calls the WeChat content security detection interface to review the uploaded image. After the review is passed, the image is then used to determine the authenticity of the local dental X-ray. The authenticity determination steps include scaling the image to 128×128 pixels, calculating the average saturation, grayscale contrast standard deviation, and average brightness, and comparing them with preset thresholds. If any indicator does not meet the statistical characteristics of a dental X-ray, the image is determined to be an invalid dental X-ray, the file is deleted, and a prompt message is returned.

[0072] If the image passes the authenticity check, the following image preprocessing operations are performed: (1) Crop the original tooth image according to the preset coordinates (50, 50, 2650, 1350); (2) Scale the cropped image to 380×380 pixels; (3) Convert the image to a three-channel grayscale image; (4) Convert the image into a Tensor and use it as input to the model.

[0073] The server prioritizes CUDA devices for inference; otherwise, it uses the CPU. The model uses EfficientNet-B4 as the backbone and outputs two values: the first is the predicted age, and the second, after processing with Softplus, is used as an uncertainty parameter. The server returns both the predicted age and the uncertainty parameter to two decimal places. The current deployment loads the optimal model weight file for the corresponding gender based on the gender parameter passed from the client; it loads the male model weight file when "gender" indicates male and the female model weight file when "gender" indicates female.

[0074] When a prediction is successful, the server returns code=0, and includes age and sigma in the data field. If an upload error, review failure, authenticity determination failure, or model prediction anomaly occurs, it returns code=1 and an error message. Regardless of success or failure, the server deletes temporary image files at the end of the process to prevent images from being permanently stored on disk.

[0075] This application also provides a user-specific age prediction system based on dental images, which may include, but is not limited to: The image acquisition module is used to acquire the uploaded image to be detected; The content compliance detection module is used to perform content scene compliance detection on the image to be detected and obtain the content detection result; The visual feature extraction module is used to extract visual features from the image to be detected, provided that the content detection result indicates compliance, to obtain the image visual features; The dental X-ray authenticity determination module is used to classify and determine whether the image to be detected is a dental medical image based on the visual features of the image, and obtain the classification and determination result; An age prediction module is used to predict the age of the image to be detected using a deep learning-based age prediction model when the classification result indicates that the image to be detected is a dental medical image, and to obtain an age prediction result.

[0076] Furthermore, the system in this application embodiment may also include an image preprocessing module, an image preprocessing module, and a temporary file deletion module, etc.

[0077] It is understood that the methods described in the above method embodiments are applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0078] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0079] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0080] Please see Figure 4 , Figure 4 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 901 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 902 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called and executed by the processor 901. The input / output interface 903 is used to implement information input and output; The communication interface 904 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 905 transmits information between various components of the device (e.g., processor 901, memory 902, input / output interface 903, and communication interface 904); The processor 901, memory 902, input / output interface 903, and communication interface 904 are connected to each other within the device via bus 905.

[0081] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0082] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0083] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0084] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented by the embodiments of this program product are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0085] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0086] The method and related equipment for predicting individual user age based on dental images provided in this application have at least one of the following beneficial effects: (1) By directly extracting age-related features from panoramic dental images and outputting prediction results through deep learning models, the reliance on artificial dental staging and artificial scoring conversion processes is reduced.

[0087] (2) By combining WeChat mini-programs with backend model services, an automatic dental age assessment system that can be deployed online and called by mobile devices has been formed, which has improved the usability and promotion of the technology.

[0088] (3) A dual input interception mechanism was constructed at the input end, which includes platform content security detection and local dental X-ray authenticity judgment. This mechanism can filter out non-compliant content and exclude obviously non-dental X-ray inputs, thereby improving the reliability of online prediction results.

[0089] (4) A local fast discrimination method based on average saturation, gray-scale contrast standard deviation and average brightness is adopted. The authenticity screening of dental films can be achieved without adding an additional pre-deep learning classification model. It has the advantages of low computational cost, simple implementation and strong interpretability.

[0090] (5) Preprocessing methods such as fixed ROI cropping, uniform input size and grayscale three-channel conversion are adopted to standardize the input image and reduce irrelevant background interference.

[0091] (6) The model outputs the predicted age value and uncertainty parameter, which provides a basis for further generation of age range or credibility analysis.

[0092] (7) By anonymizing and renaming files and automatically deleting temporary files throughout the process, the system’s data security and privacy protection level have been improved.

[0093] (8) It has a complete engineering implementation chain and can be directly deployed in actual online service scenarios, rather than just staying in the offline algorithm verification stage.

[0094] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0095] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0096] The system embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0097] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0098] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or modules is not necessarily limited to those steps or modules explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0099] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0100] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0101] The modules described above as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0102] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0103] If the integrated modules are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0104] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A user-specific age prediction method based on dental images, characterized in that, Includes the following steps: Obtain the uploaded image to be detected; The content scene compliance is checked on the image to be detected to obtain the content detection result; If the content detection result indicates compliance, visual features are extracted from the image to be detected to obtain image visual features; The image to be detected is classified and determined as a dental medical image based on the visual features of the image, and the classification result is obtained. If the classification result indicates that the image to be detected is a dental medical image, an age prediction model based on deep learning is used to predict the age of the image to be detected, and an age prediction result is obtained.

2. The method for predicting individual user age based on dental images according to claim 1, characterized in that, The user-specific age prediction method based on dental images also includes the following steps: A unique identifier is randomly generated for the image to be detected; The image to be detected is renamed according to the unique identifier and the original file extension of the image to be detected to obtain an image file name, and then the image file name is used to manage the image to be detected.

3. The method for predicting individual user age based on dental images according to claim 1, characterized in that, The user-specific age prediction method based on dental images also includes the following steps: If the content detection result indicates non-compliance, the image to be detected is deleted and an error message is returned to the client.

4. The method for predicting individual user age based on dental images according to claim 1, characterized in that, The image visual features include image saturation, image contrast, and average image brightness. The process of classifying the image to be detected as a dental medical image based on these visual features to obtain a classification result includes the following steps: Determine whether the preset detection conditions are met based on the visual features of the image; If any of the preset detection conditions are met, the image to be detected is determined to be a non-dental medical image. The preset detection conditions include an average image saturation greater than a first threshold, a standard deviation of image contrast less than a second threshold, and an average image brightness greater than a third threshold.

5. The method for predicting individual user age based on dental images according to claim 4, characterized in that, The first threshold, the second threshold, and the third threshold are determined through the following steps: Collect dental medical imaging samples and non-dental imaging samples; Statistical analysis of the visual features of the dental medical image samples and the non-dental image samples was performed respectively to obtain the first data distribution features of the dental medical image samples and the second data distribution features of the non-dental image samples. Based on the first data distribution characteristics and the second data distribution characteristics, a first threshold, a second threshold, and a third threshold are determined.

6. The user-individual age prediction method based on dental images according to any one of claims 1 to 5, characterized in that, The step of using a deep learning-based age prediction model to predict the age of the image to be detected, and obtaining the age prediction result, includes the following steps: The image to be detected is input into an age prediction model based on a convolutional neural network to obtain the predicted age value and uncertainty. The age prediction interval is determined based on the predicted age value and the uncertainty. An age prediction result is formed based on the predicted age value, the uncertainty, and the predicted age range.

7. The user-oriented individual age prediction method based on dental images according to any one of claims 1 to 5, characterized in that, The age prediction model is trained through the following steps: Obtain a dental medical image sample dataset, which includes normal dental samples and dental disease samples; The dental medical image sample dataset is divided into a first sample dataset and a second sample dataset according to gender. The initialized age prediction model was trained using the first sample dataset and the second sample dataset respectively, resulting in two age prediction models for different genders.

8. A user-specific age prediction system based on dental images, characterized in that, include: The image acquisition module is used to acquire the uploaded image to be detected; The content compliance detection module is used to perform content scene compliance detection on the image to be detected and obtain the content detection result; The visual feature extraction module is used to extract visual features from the image to be detected, provided that the content detection result indicates compliance, to obtain the image visual features; The dental X-ray authenticity determination module is used to classify and determine whether the image to be detected is a dental medical image based on the visual features of the image, and obtain the classification and determination result; An age prediction module is used to predict the age of the image to be detected using a deep learning-based age prediction model when the classification result indicates that the image to be detected is a dental medical image, and to obtain an age prediction result.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method according to any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images