A method for monitoring oral health using image analysis

Abstract

The present invention relates to the technical field of oral health monitoring, and discloses a method for monitoring oral health using image analysis, comprising the following steps: step one, collecting two-dimensional images and three-dimensional images of a patient's oral cavity at different time periods, and performing preprocessing; step two, classifying all images and extracting feature data to form a two-dimensional feature data set and a three-dimensional feature data set; step three, calculating and generating a dynamic model Dtmx according to the two-dimensional feature data set and the three-dimensional feature data set; step four, substituting the two-dimensional feature data set and the three-dimensional feature data set into a deep learning algorithm model, calculating and generating an initial model Csmx and a prediction model Ycmx, and deeply analyzing the image simulation diagnosis to be more accurate; step five, judging and generating a management report according to the initial model Csmx, the dynamic model Dtmx and the prediction model Ycmx, and achieving high oral health management and treatment efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of oral health monitoring, and in particular to a method for monitoring oral health using image analysis. Background Art

[0002] Oral health monitoring using image analysis is a non-invasive diagnostic method that uses advanced computing technology to analyze pictures of the inside of the mouth to identify and evaluate the health status of teeth, gums, and other oral tissues. This method can assist dentists in detecting early signs of diseases such as caries, periodontal disease, and oral cancer, and track the effectiveness of treatment. Image analysis technology usually includes steps such as image acquisition, preprocessing, feature extraction, classification, and interpretation. Modern intraoral cameras and x-ray equipment can provide high-resolution two-dimensional or three-dimensional images. The acquired images need to be preprocessed, including denoising, contrast enhancement, and edge detection, to improve the accuracy of subsequent analysis. Feature extraction is the screening of features such as color, texture, shape, and edges in oral image analysis. Selecting appropriate features is crucial to improving the performance of the classifier, and then the deep learning algorithm model is used to identify important data related to specific diseases, identifying tiny cracks on the tooth surface, dental plaque, tartar, gingival inflammation, and bleeding spots. With the advancement of science and technology, the field of digital medicine has made significant progress. Especially in the field of oral health, the application of image analysis technology has brought revolutionary changes to traditional oral examinations. Through high-resolution cameras and advanced image processing algorithms, doctors can more accurately monitor patients' oral health conditions and provide more personalized and efficient treatment plans.

[0003] At present, the traditional oral health monitoring method using image analysis can only provide static two-dimensional or three-dimensional images, and cannot deeply simulate the tooth displacement process and trend. The image analysis results are easily affected by the experience of the interpreting doctor and have a certain degree of subjectivity, which may lead to misdiagnosis or missed diagnosis. The traditional monitoring method lacks the automation and accuracy of image analysis. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] In view of the shortcomings of the prior art, the present invention provides a method for monitoring oral health using image analysis, which has the advantages of in-depth analysis of images, more accurate simulation prediction, and high efficiency in oral health management and prevention. It solves the problem that traditional methods for monitoring oral health using image analysis cannot deeply simulate the tooth displacement process and trend, are highly subjective, and lack automation and accuracy.

[0006] (II) Technical solution

[0007] To achieve the above object, the present invention provides the following technical solution: a method for monitoring oral health using image analysis, comprising the following steps:

[0008] Step 1: Collect two-dimensional and three-dimensional images of the patient's oral cavity at different time periods using an X-ray device, and pre-process all images;

[0009] Step 2: After classifying all images, extract the feature data of tooth color, tooth shape, gum margin, tongue coating and oral mucosa, and form a two-dimensional feature data set and a three-dimensional feature data set;

[0010] Step 3: Calculate and generate a dynamic model Dtmx based on the two-dimensional feature data set and the three-dimensional feature data set, and display the dynamic model from multiple angles in three-dimensional space through computer software;

[0011] Step 4: Substitute the two-dimensional feature data set and the three-dimensional feature data set into the deep learning algorithm model, calculate and generate the initial model Csmx and the prediction model Ycmax, and display the initial model and prediction model from multiple angles in the three-dimensional space through computer software;

[0012] Step 5: Based on the initial model Csmx, the dynamic model Dtmx and the prediction model Ycmx, analyze and compare the changes in the patient's oral status in different time periods, and determine and generate the corresponding management report.

[0013] Preferably, in step one, the preprocessing process includes image denoising, contrast enhancement and edge detection steps. First, a Gaussian filter is used for balanced denoising, then a histogram equalization method is used to adjust the grayscale distribution of the image, and finally a Canny edge detection algorithm is used to retain the clarity of the two-dimensional image edge and the three-dimensional image edge.

[0014] Preferably, in the step 2, the preprocessed two-dimensional images and three-dimensional images are classified in chronological order from early to late, the two-dimensional images collected in the same time period are grouped as a group, and the three-dimensional images collected in the same time period are grouped as a group, and the tooth color, tooth shape, gum margin, tongue coating and oral mucosa feature data are extracted from each group of images in chronological order from early to late.

[0015] Preferably, in step 2, the two-dimensional feature data set expression is t represents the time data of the two-dimensional image in the same time period, 2D represents the feature data of the two-dimensional image in the same time period, to Corresponding to 1 to n groups of oral 2D images collected at different time periods, and each group of 2D images contains multiple oral 2D images collected at the same time period, the 3D feature data set expression is: j represents the time data of the three-dimensional image in the same time period, 3D represents the feature data of the three-dimensional image in the same time period, to They correspond to groups 1 to n of oral three-dimensional images collected at different time periods, and each group of three-dimensional images includes multiple oral three-dimensional images collected at the same time period.

[0016] Preferably, in step 3, SV3D technology is used to analyze the two-dimensional feature data set and the three-dimensional feature data set, and the dynamic model Dtmx is calculated and generated, and the calculation formula is as follows:

[0017]

[0018] In the formula, Dtmx represents the dynamic model, It means that the feature data of the two-dimensional image and the feature data of the three-dimensional image in the latest time period are substituted into the potential video diffusion model provided by the SV3D technology, and the multi-view data for constructing the dynamic model is calculated and generated.

[0019] Preferably, in step 4, the two-dimensional feature data set and the three-dimensional feature data set are substituted into the deep learning algorithm model F, and the initial model Csmx is calculated and generated, and the calculation formula is as follows:

[0020]

[0021] In the formula, Csmx represents the initial model, and YZ represents the uniform standard tooth centerline in the initial model calculation formula. It means that the tooth midline is used as the axis of the initial model, and the feature data of the two-dimensional image and the feature data of the three-dimensional image in the earliest time period are substituted into the deep learning algorithm model to calculate and generate multi-view data for constructing the initial model.

[0022] Preferably, in step 4, the two-dimensional feature data set and the three-dimensional feature data set are substituted into the deep learning algorithm model F, and the prediction model Ycmx is calculated and generated, and the calculation formula is as follows:

[0023]

[0024] In the formula, Ycmx represents the prediction model, α is the weight index for the two-dimensional feature data set, and β is the weight index for the three-dimensional data set. Indicates the difference change of two-dimensional image features from the earliest time period to the latest time period. Indicates the difference change of 3D image features from the earliest time period to the latest time period. It means that under the support of different weights, based on the initial model, the two-dimensional image feature difference and the three-dimensional image feature difference are substituted into the deep learning algorithm model to calculate and generate multi-view data for building a prediction model.

[0025] Preferably, in step five, the evolving playback video is analyzed and generated based on the initial model Csmx and the dynamic model Dtmx, and then the disease management report is generated based on the oral disease diagnosis criteria provided by the database, and the disease management report includes oral disease prediction results and prevention recommendations.

[0026] Preferably, in step five, a trend prediction video is analyzed and generated based on the dynamic model Dtmx and the prediction model Ycmx, and then a health management report is generated based on the oral health management cases provided by the database. The health management report includes relevant oral health management cases and professional suggestions.

[0027] Preferably, when there is only feature data of a single time period in the two-dimensional feature data set and the three-dimensional feature data set, only the initial model Csmx is calculated and generated, and then the disease management report is generated based on the oral disease diagnosis standard provided by the database.

[0028] Compared with the prior art, the present invention provides a method for monitoring oral health using image analysis, which has the following beneficial effects:

[0029] 1. The present invention uses an x-ray device to collect two-dimensional images and three-dimensional images of the patient's oral cavity at different time periods, and pre-processes all images to improve visual effects and make image details clearer. All images are classified in chronological order from early to late, which is crucial for understanding the dynamic characteristics of the data. The tooth color, tooth shape, gingival margin, tongue coating and oral mucosa feature data in the image are extracted, and a two-dimensional feature data set and a three-dimensional feature data set are formed to facilitate subsequent rapid identification of oral symptoms, analysis of the evolution process and future trends of the oral state. According to the two-dimensional feature data set and the three-dimensional feature data set, a dynamic model Dtmx is calculated and generated, and a dynamic model of multiple angles in three-dimensional space is displayed by computer software, a 1:1 dynamic model under the current oral state is constructed, key features are intuitively displayed and highlighted, the two-dimensional feature data set and the three-dimensional feature data set are substituted into the deep learning algorithm model, and the initial model Csmx and the prediction model Ycmx are calculated and generated, and in-depth analysis of image simulation prediction is more accurate.

[0030] 2. The present invention generates an evolving playback video through analysis of the initial model Csmx and the dynamic model Dtmx, and then generates a disease management report through the oral disease diagnosis standard provided by the database. The oral disease diagnosis standard is the theoretical basis for determining whether a patient suffers from oral disease. The disease management report includes oral disease prediction results and prevention suggestions. According to the dynamic model Dtmx and the prediction model Ycmx, a trend prediction video is generated through analysis. Then, through the oral health management cases provided by the database, the oral health management cases are the practical basis for providing reference for the patient's subsequent treatment plan. The health management report includes relevant oral health management cases and professional suggestions, and the oral health management prevention efficiency is high. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a step diagram of the method of the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] See also Figure 1 , a method for monitoring oral health using image analysis, comprising the following steps:

[0034] Step 1: Collect two-dimensional and three-dimensional images of the patient's oral cavity at different time periods using an X-ray device, and pre-process all images;

[0035] The oral health management cycle of patients consists of multiple time periods, which are set to 2-4 weeks or 3-4 months. Different patients have different oral health management cycles. 2-4 weeks are suitable for patients who have just started orthodontic treatment, and 3-4 months are suitable for patients with common oral diseases, such as caries, abnormal wisdom tooth growth, etc.

[0036] The preprocessing process includes image denoising, contrast enhancement and edge detection steps. First, Gaussian filter is used to balance denoising to improve the visual effect. Then the histogram equalization method is used to adjust the grayscale distribution of the image to improve the overall or local contrast of the image and make the image details clearer. Finally, the Canny edge detection algorithm is used to retain the clarity of the edges of two-dimensional and three-dimensional images.

[0037] Step 2: After classifying all images, extract the feature data of tooth color, tooth shape, gum margin, tongue coating and oral mucosa, and form a two-dimensional feature data set and a three-dimensional feature data set;

[0038] The preprocessed 2D and 3D images are classified in chronological order from early to late. The 2D images collected in the same time period are grouped together, and the 3D images collected in the same time period are grouped together. The tooth color, tooth shape, gingival margin, tongue coating, and oral mucosa feature data in each group of images are extracted in chronological order from early to late. Analyzing the image features in time series is crucial to understanding the dynamic characteristics of the data.

[0039] The expression of the two-dimensional feature data set is t represents the time data of the two-dimensional image in the same time period, 2D represents the feature data of the two-dimensional image in the same time period, to It corresponds to 1 to n groups of oral 2D images collected at different time periods, and each group of 2D images contains multiple oral 2D images collected at the same time period. Each group of 2D images can provide tooth color features, tooth shape features, and tooth texture features, which is convenient for subsequent rapid identification of oral symptoms. The expression of the 3D feature data set is: j represents the time data of the three-dimensional image in the same time period, 3D represents the feature data of the three-dimensional image in the same time period, to It corresponds to 1 to n groups of oral 3D images collected at different time periods, and each group of 3D images contains multiple oral 3D images collected at the same time period. Each group of 3D images can provide tooth size and position characteristics, which is convenient for subsequent rapid analysis of the evolution process and future trends of the oral state.

[0040] Step 3: Calculate and generate the dynamic model Dtmx based on the two-dimensional feature data set and the three-dimensional feature data set, and display the dynamic model from multiple angles in the three-dimensional space through computer software. The calculation formula is as follows:

[0041]

[0042] In the formula, Dtmx represents the dynamic model, It means that the feature data of the two-dimensional image and the feature data of the three-dimensional image in the latest time period are substituted into the potential video diffusion model provided by the SV3D technology to calculate and generate multi-view data for building a dynamic model. The latest time period is specifically the data collected in the last time period of the two-dimensional feature data set and the three-dimensional feature data set. According to the dynamic model Dtmx, a 1:1 dynamic model of the current oral state can be quickly built, and key features can be intuitively displayed and highlighted, so that subsequent decision-making efficiency is higher;

[0043] Step 4: Substitute the two-dimensional feature data set and the three-dimensional feature data set into the deep learning algorithm model. The deep learning algorithm model is specifically a convolutional neural network commonly used in the field of image analysis. The initial model Csmx and the prediction model Ycmx are calculated and generated, and the initial model and prediction model from multiple angles in three-dimensional space are displayed through computer software. The calculation formula is as follows:

[0044]

[0045] In the formula, Csmx represents the initial model, and YZ represents the uniform standard tooth centerline in the initial model calculation formula. It means that the tooth midline is used as the axis of the initial model, and the feature data of the two-dimensional image and the feature data of the three-dimensional image in the earliest time period are substituted into the deep learning algorithm model to calculate and generate multi-view data for building the initial model. The earliest time period is specifically the data collected in the first time period of the two-dimensional feature data set and the three-dimensional feature data set. According to the initial model Csmx, combined with the patient's personal characteristics, the healthy oral state based on the standard tooth midline is restored to facilitate the subsequent analysis and comparison of the state gap between the initial model Csmx and the dynamic model Dtmx. If the current tooth midline of the patient's dynamic model cannot coincide with the tooth midline of the initial model, it means that the patient's oral disease is more serious, and in-depth analysis of image simulation prediction is more accurate;

[0046]

[0047] In the formula, Ycmx represents the prediction model, α is the weight index for the two-dimensional feature data set, and β is the weight index for the three-dimensional data set. Indicates the difference change of two-dimensional image features from the earliest time period to the latest time period. Indicates the difference change of 3D image features from the earliest time period to the latest time period. It means that under different weights, based on the initial model, the two-dimensional image feature difference and the three-dimensional image feature difference are substituted into the deep learning algorithm model. The deep learning algorithm model is specifically a convolutional neural network commonly used in the field of image analysis. It calculates and generates multi-view data for building a prediction model. According to the prediction model Ycmx, potential oral disease risks are discovered in time to avoid worsening of the disease. In-depth analysis of image simulation diagnosis is more accurate.

[0048] Step 5: According to the initial model Csmx, the dynamic model Dtmx and the prediction model Ycmx, the changes in the oral status of the patient in different time periods are analyzed and compared, and the corresponding management report is generated to determine the high efficiency of oral health management and prevention;

[0049] According to the initial model Csmx and the dynamic model Dtmx, the evolving playback video is analyzed and generated, and then the oral disease diagnosis standard provided by the database is used to judge whether the patient has oral disease. The oral disease diagnosis standard is the theoretical basis for judging whether the patient has oral disease, and the disease management report, oral disease prediction results and prevention suggestions are generated to advise patients to prevent related oral diseases in advance;

[0050] According to the dynamic model Dtmx and the prediction model Ycmx, the trend prediction video is analyzed and generated, and then the oral health management cases provided by the database are used to judge and generate the health management report, which includes relevant oral health management cases and professional suggestions.

[0051] When there is only feature data of a single time period in the two-dimensional feature data set and the three-dimensional feature data set, only the initial model Csmx is calculated and generated, and then the disease management report is generated based on the oral disease diagnosis standards provided by the database. It is suitable for emergency patients with oral diseases or patients who are undergoing oral health management for the first time.

[0052] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring oral health using image analysis, characterized in that: The following steps are involved: Step 1: Collect two-dimensional images and three-dimensional images of the patient's oral cavity at different time periods using an X-ray device, and pre-process all the images; Step 2: After classifying all images, extract the feature data of tooth color, tooth shape, gum margin, tongue coating and oral mucosa, and form a two-dimensional feature data set and a three-dimensional feature data set. The expression of the two-dimensional feature data set is t represents the time data of the two-dimensional image in the same time period, 2D represents the feature data of the two-dimensional image in the same time period, to Corresponding to 1 to n groups of oral 2D images collected at different time periods, and each group of 2D images contains multiple oral 2D images collected at the same time period, the 3D feature data set expression is: j represents the time data of the three-dimensional image in the same time period, 3D represents the feature data of the three-dimensional image in the same time period, to The three-dimensional images of the oral cavity collected at different time periods correspond to groups 1 to n in sequence, and each group of three-dimensional images includes multiple three-dimensional images of the oral cavity collected at the same time period; Step 3: Calculate and generate a dynamic model Dtmx based on the two-dimensional feature data set and the three-dimensional feature data set, and display the dynamic model from multiple angles in three-dimensional space through computer software; Step 4: Substitute the two-dimensional feature data set and the three-dimensional feature data set into the deep learning algorithm model, calculate and generate the initial model Csmx and the prediction model Ycmx, and display the initial model and prediction model at multiple angles in the three-dimensional space through computer software; substitute the two-dimensional feature data set and the three-dimensional feature data set into the deep learning algorithm model F, and calculate and generate the initial model Csmx. The calculation formula is as follows: In the formula, Csmx represents the initial model, and YZ represents the uniform standard tooth centerline in the initial model calculation formula. It means that the tooth midline is used as the axis of the initial model, and the feature data of the two-dimensional image and the feature data of the three-dimensional image in the earliest time period are substituted into the depth algorithm model to calculate and generate multi-view data for constructing the initial model; Substitute the two-dimensional feature data set and the three-dimensional feature data set into the deep learning algorithm model F, and calculate and generate the prediction model Ycmx. The calculation formula is as follows: In the formula, Ycmx represents the prediction model, α is the weight index for the two-dimensional feature data set, and β is the weight index for the three-dimensional data set. Indicates the difference change of two-dimensional image features from the earliest time period to the latest time period. Indicates the difference change of 3D image features from the earliest time period to the latest time period. It means that under different weights, based on the initial model, the two-dimensional image feature difference and the three-dimensional image feature difference are substituted into the deep learning algorithm model to calculate and generate multi-view data for building a prediction model; Step 5: Based on the initial model Csmx, the dynamic model Dtmx and the prediction model Ycmx, analyze and compare the changes in the patient's oral status in different time periods, and determine and generate the corresponding management report.

2. A method for monitoring oral health using image analysis according to claim 1, characterized in that: In step 1, the preprocessing process includes image denoising, contrast enhancement and edge detection steps. First, a Gaussian filter is used for balanced denoising, then a histogram equalization method is used to adjust the grayscale distribution of the image, and finally a Canny edge detection algorithm is used to retain the clarity of the edges of the two-dimensional image and the three-dimensional image.

3. A method for monitoring oral health using image analysis according to claim 2, characterized in that: In the step 2, the preprocessed two-dimensional images and three-dimensional images are classified in chronological order from early to late, the two-dimensional images collected in the same time period are grouped as one group, and the three-dimensional images collected in the same time period are grouped as one group, and the tooth color, tooth shape, gingival margin, tongue coating and oral mucosa feature data are extracted in chronological order from early to late in each group of images.

4. The method for monitoring oral health using image analysis according to claim 1, characterized in that: In step 3, SV3D technology is used to analyze the two-dimensional feature data set and the three-dimensional feature data set, and the dynamic model Dtmx is calculated and generated, and the calculation formula is as follows: In the formula, Dtmx represents the dynamic model, It means that the feature data of the two-dimensional image and the feature data of the three-dimensional image in the latest time period are substituted into the potential video diffusion model provided by the SV3D technology, and the multi-view data for constructing the dynamic model is calculated and generated.

5. The method for monitoring oral health using image analysis according to claim 1, characterized in that: In the step five, the evolving playback video is analyzed and generated based on the initial model Csmx and the dynamic model Dtmx, and then the disease management report is generated based on the oral disease diagnosis criteria provided by the database. The disease management report includes oral disease prediction results and prevention recommendations.

6. The method for monitoring oral health using image analysis according to claim 1, characterized in that: In the step five, a trend prediction video is analyzed and generated based on the dynamic model Dtmx and the prediction model Ycmx, and then a health management report is generated based on the oral health management cases provided by the database. The health management report includes relevant oral health management cases and professional suggestions.

7. The method for monitoring oral health using image analysis according to claim 1, characterized in that: When there is only feature data of a single time period in the two-dimensional feature data set and the three-dimensional feature data set, only the initial model Csmx is calculated and generated, and then the disease management report is generated based on the oral disease diagnosis standard provided by the database.

Images