WO2023163670A1

WO2023163670A1 - A dental health application

Info

Publication number: WO2023163670A1
Application number: PCT/TR2022/050220
Authority: WO
Inventors: Sanaz SADRY; Esra PAMUKÇU GÜVEN; Tosun Tosun; Sajad EINY; Esra SOMTÜRK
Original assignee: T.C. Istanbul Aydin Univeristesi
Priority date: 2022-02-24
Filing date: 2022-03-11
Publication date: 2023-08-31
Also published as: TR2022002554A2

Abstract

The invention relates to an artificial intelligence supported dental health application (A) developed for oral and dental health care in the field of dentistry in the health technology sector.

Description

DESCRIPTION

A DENTAL HEALTH APPLICATION

Technical Field

The invention relates to a dental health application.

In particular, the invention relates to an artificial intelligence supported dental health application developed for oral and dental health care in the field of dentistry in the health technology sector.

Background of the Invention

The importance of oral and dental health is very important because it is the mouth, which is the beginning of the digestive system. Since the teeth carry out the process of disintegration of the food and sending it to the stomach, missing, unhealthy, diseased or rotten teeth disrupts the digestive state.

Dental or oral health; relates to our teeth, gums and mouth. The aim is to prevent complications such as tooth decay (cavities) and gum disease and to maintain the overall health of your mouth.

Good dental practice is important for a healthy mouth, teeth and gums. It also helps your appearance and quality of life.

Your dental health team consists of you and your dental hygienists. Together, you can prevent many dental problems from reduced quality of life and possible medical complications.

Diseased, crooked, missing teeth or a deformed jaw can interfere with your speech, make chewing food difficult and painful, and lead to expensive corrections. In the current technique, there are mobile applications that only aim to create a patient record base for dentists and contain content that facilitates the keeping of patient records. Or there are game-based applications specific to users.

These practices do not provide professional protection, improvement and maintenance of oral and dental health.

The most important problem for the patient is the absence of applications that will solve the problem completely, and the fact that the ones that are not reliable and sustainable are encountered.

Regarding the subject in the US patent application numbered US2020411167 in the literature “A first machine learning model is trained to output a patient ID, study ID, and/or image view ID. A final layer of the first model is removed to obtain an encoder that outputs feature vectors that may be used to characterize input images. Images with matching patient ID, study ID, and/or image view ID may be identified by comparing feature vectors. The first machine learning model may be a CNN with two fully connected layers, one of which is removed after training. The encoder may also be trained by evaluating triplet loss, comparing feature vectors for matching and nonmatching images, or by training an encoder to reproduce a vector used to generate a synthetic image by a generator as part of an adversarial learning routine.” statements are included.

In said application, automatic dental patient identification and duplicate content extraction using contradictory learning is disclosed.

In the Australian patent application numbered AU2020104159 in the literature, " In recent years, multi fold improvement is viewed in the field of Artificial Intelligence hence plays a significant role in image classification especially classification of medical images. In specific Convolutional Neural Networks (CNN) belonging to Artificial Intelligence performs well in detection of several diseases such as heart disease, Dental diseases, Malaria and Parkinson's disease. CNN has significant vision in detection of lung disease utilizing the medical images of the patient such as X-rays. Lung disease is the basic symptom of the global pandemic disease COVID-19. This invention proposes a CNN model for the detection of lung disease where the model involves four layers namely input layers, convolutional layers, fully connected layers and output layers. The three layered two dimensional convolutional layers involves ReLu activation function along with Max pooling making the detection process easier by training the model using dataset.” statements are included.

In the aforementioned patent application, intelligent intelligence-based detection of lung disease from chest X-ray is disclosed.

Also in the literature, in the US patent application numbered US2021256696, " A method for designing a drilling template, wherein a dental situation is measured by means of a 3D surface measuring device and a 3D surface model of the dental situation is produced and/or measured by means of an X-ray device or an MRI device, wherein the dental situation is measured and a volume model of the dental situation is produced, the method comprising the steps of: applying an artificial neural network for machine learning (convolutional neural network; CNN) to the 3D surface model of the dental situation and/or the volume model of the dental situation and/or to an initial 3D model of the drilling template: and automatically producing a ready made 3D model of the drilling template.” statements are included.

In said embodiment, a method of measuring a tooth condition by means of a dental camera or a laboratory scanner and producing a 3D surface model of the tooth condition and/or designing a drilling template by means of an X-ray device is disclosed.

Again in the literature, in the Australian patent application numbered AU2021100684, “A MATLAB App for Caries Detection and Diagnosis from Dental X-rays In the aeon of deep learning, CNN outperform significant part in medical image analysis. So, Software Applications for caries detection can utilize significant features to detect and diagnose different types of caries. Now a day CNNs based application software are worth popular due to automatic relevant features extraction. CNNs can be trained from ground up for medical images but due to finite number of medical images transfer learning and data augmentations are used for training. Dental X-Rays can contain different types of caries in dentin, enamel, proximal and root surface of tooth anatomy. In this invention we have developed a MATLAB app for Caries detection and diagnosis like Dentin, Enamel and Pulp based on Hybrid CNN framework. This MATLAB app performs different steps to enhance contrast, to preprocess the dental X-ray. At the final step classify carious lesions into Dentin, Enamel and Pulp. It has been trained and examined on primary Dental X-rays available from a Dental Clinic”.

In the aforementioned patent, a MATLAB application is described for the detection and diagnosis of caries from dental x-rays.

Due to the disadvantages mentioned above, there was a need to introduce a new artificial intelligence-supported dental health application.

Disclosure of the Invention

Based on this position of the technique, the aim of the invention is to present a new artificial intelligence supported dental health application that eliminates the existing disadvantages.

Another aim of the invention is to present a structure that makes artificial intelligencebased diagnostic planning and provides feedback to the patient and the physician.

Another aim of the invention is for potential dental patients, children; is to present an application that appeals to a wide audience such as kindergartens, love houses, NGOs, nursing homes, associations, dentists, hospitals, polyclinics and practices, dental companies.

Another aim of the invention is to present a structure that realizes information sharing on an academic basis, thanks to the informative videos and animations in its content.

Another aim of the invention is to present an application that contains personal content.

Another aim of the invention is to present an application that allows self-care. Another aim of the invention is to present a structure that allows remote physician support in order to direct patients to the right branch, choose a physician, and find support in emergencies.

Another aim of the invention is to present a structure that enables users to be motivated by following their own intraoral transformation.

Another aim of the invention is to present a structure that makes reminders by following the tooth brushing processes of the users.

Explanation of Figures

Figure - 1 A schematic view of the dental health application subject to the invention

Reference Numbers

A- Dental Health Application

1. Input Module

2. Preprocessing Module

3. Processing Unit

4. Output Module

5. Artificial Intelligence Module

6. Dental Detection Unit

Detailed Description of the Invention

In this detailed explanation, the innovation that is the subject of the invention is only explained with examples that will not have any limiting effect for a better understanding of the subject.

The invention relates to an intelligence-supported dental health application (A), was developed in the health technology sector, in dentistry, for oral and dental care, to evaluate dental radiography, congenital tooth defects, periodontal diseases, orthodontic manipulations, oral tumors, endodontic treatments, oral trauma and any condition where an abnormality is suspected, characterized in that; comprises, input module (1 ) that allows two-dimensional images of panoramic dental x-rays, which use very small doses of ionizing radiation to capture the entire mouth in a single image, to the mobile phone as an input with the help of web service, Bluetooth and wireless communication systems, the processing unit (3), which analyzes the two-dimensional images of the said x-ray and detects the types of disease in the model teeth and sends them to the doctor for verification, artificial intelligence module (5) that provides classification of the radiographic image and gives more detailed information about dental abnormalities using a threshold value of 80% and above and a dental detection unit (6) formed to separate and analyze each tooth separately.

Figure - 1 shows a schematic view of the dental health application (A) subject to the invention.

The dental health application (A) according to invention consists main parts of, input module (1 ) that allows two-dimensional images of panoramic dental x-rays, which use very small doses of ionizing radiation to capture the entire mouth in a single image, to the mobile phone as an input with the help of web service, Bluetooth and wireless communication systems, preprocessing module (2) with histogram equalization, which adjusts the contrast of the image using its histogram to enhance the x-ray image of said panoramic dental x-ray, the processing unit (3), which analyzes the two- dimensional images of the said x-ray and detects the types of disease in the model teeth and sends them to the doctor for verification, output module (4) that emails the scan results to the client after the disease type has been identified and verified by the doctor, artificial intelligence module (5) that provides classification of the radiographic image and gives more detailed information about dental abnormalities using a threshold value of 80% and above and a dental detection unit (6) formed to separate and analyze each tooth separately.

The aforementioned dental health application (A) is used to evaluate dental radiography, congenital dental defects, periodontal diseases, orthodontic manipulations, oral tumors, endodontic treatments, oral trauma and any condition where an abnormality is suspected. Said dental health application (A) detects such abnormalities and generates a report to the dentist.

The report sent to the mentioned doctor includes general information such as the patient's age, gender and the last visit of the doctors. In addition to such general information, the artificial intelligence module (5) provides more detailed information about dental abnormalities with the help of the image processing unit (3) that analyzes the radiographic image. First, using supervised learning, pre-labeled images are learned by the model. For this learning, the artificial intelligence module (5) mentioned in order to teach and label the diseases around the teeth to the machine includes the YOLO algorithm.

In the mentioned dental health application (A), edge computing infrastructure platform is used. The deep learning architecture is based on the hybrid internet of things platform. Edge computing is used to reduce the transaction cost on the server. Software architecture consists of two main parts. While the model is being trained, image segmentation is performed in edge calculation. It also includes a dental detection unit (6) to separate and analyze each tooth separately.

In the second step, each separated and fragmented tooth is sent to the server for analysis processing. Finally, the deep learning algorithm detects the abnormality or disease type from the input image and reports the disease type for the dentist.

The report created is reviewed by the doctor once again, and if there is an error, it is labeled correctly. As the data that was predicted incorrectly and then labeled correctly by the doctor accumulates, it is used to train the model at certain intervals. In this way, it is aimed to design a sustainable continuous learning system. The results are available to the doctor and client.

This hybrid loT can be applied to any edge, such as mobile phones and embedded systems.

In the implementation of the invention, the radiographic image is uploaded by the customer with the help of the mobile dental health application (A). Then, the image sent to the central server with the help of the mobile processor unit is passed through the artificial intelligence module (5) and classified. A treshold value of 80% and above is used in this classification. The values below are saved in the system to teach the model and no report is presented to the user.

Presicion, recall and F1 score values will be decisive in the model outputs. Our main purpose here is to use a model with F1 score0.8 and above. The response time of this system for the result is about one minute and the accuracy of the classification of teeth is over 80%.

Because the X-ray dataset comes from a variety of sources, there are many variations in the size, shape, and color of these images. In this case, Histogram equalization is used for standardization of these images (CLAHE (Contrast Limited Adaptive Histogram Equalization)).

Opencv library is used for histogram equalization technique which allows to increase contrast (CLAHE) locally while limiting noise amplification. A window size of 8.8 is a good ratio for our contrast between all images.

Tensorflow Object Detection API is used to extract each tooth from the radiographic image. Model ZOO allows you to select a pre-trained model and train it easily on your dataset. Transfer learning method is used to increase the accuracy rate in object detection.

Faster rcnn Mobilnet V1 coco was selected and used. Transfer learning is helpful for two main reasons. First of all, a pre-trained model makes it easy to learn shapes and specific objects. Second, the faster RCNN Mobilnet V1 uses higher image sizes than its predecessors.

The dental health application (A), which is the subject of the invention, is a new application based on edge computing systems such as the information processing systems of smart phones. This deep learning-based application includes caries and dental health detection on panoramic dental x-ray images. This application includes deep learning-based anomaly detection in panoramic dental x-ray in smartphone systems for the first time. It also reduces server-side transaction costs and improves response time due to the use of smartphone distribution systems.

Panoramic dental x-rays use very small doses of ionizing radiation to capture the entire mouth in a single image. It is widely performed by dentists and oral surgeons in daily practice and is used to plan the treatment of dentures, braces, extractions and implants. These 2D images are sent to the mobile phone as an input with the help of web service, Bluetooth and wireless communication systems.

Histogram equalization is applied in the preprocessing module (2) to improve the X- ray image. Histogram Equalization is an image processing technique that adjusts the contrast of an image using its histogram. It spreads the most common pixel density values or extends the intensity range of the image to improve the contrast of the image. By doing this, histogram equalization ensures that lower contrast areas of the image gain higher contrast. This image processing tool is processed in the edge section.

The main process part can be changed according to the application scenario of the dental health application (A) which is the subject of the invention. The detection process is stored on the client side, which can also be set on the server side. Deep learning architecture is based on smart phone smart artificial systems. After pretreatment, the processing unit (3) detects the types of disease in the model teeth. Furthermore, the processing unit (3) sends these results to the doctor for verification.

After the disease type is determined and confirmed by the doctor, the scan results are sent to the customer via e-mail via the output module (4).

Claims

CLAIMS The invention relates to an intelligence-supported dental health application (A), was developed in the health technology sector, in dentistry, for oral and dental care, to evaluate dental radiography, congenital tooth defects, periodontal diseases, orthodontic manipulations, oral tumors, endodontic treatments, oral trauma and any condition where an abnormality is suspected, characterized in that; comprises, input module (1 ) that allows two-dimensional images of panoramic dental x-rays, which use very small doses of ionizing radiation to capture the entire mouth in a single image, to the mobile phone as an input with the help of web service, Bluetooth and wireless communication systems, the processing unit (3), which analyzes the two-dimensional images of the said x- ray and detects the types of disease in the model teeth and sends them to the doctor for verification, artificial intelligence module (5) that provides classification of the radiographic image and gives more detailed information about dental abnormalities using a threshold value of 80% and above and a dental detection unit (6) formed to separate and analyze each tooth separately. A dental health application (A) according to claim 1 , consists of preprocessing module (2) with histogram equalization, which adjusts the contrast of the image using its histogram to enhance the x-ray image of said panoramic dental x-ray. A dental health application (A) according to any preceding claims; consists of output module (4) that emails the scan results to the client after the disease type has been identified and verified by the doctor. A dental health application (A) according to any preceding claims; the artificial intelligence module (5) includes the YOLO algorithm with the aim of teaching and labeling the diseases around the teeth to the machine. A dental health application (A) according to any preceding claims; It includes an edge computing infrastructure platform that enables deep learning-based anomaly detection.