A Deep Learning Approach to Submerged Teeth Classification and Detection

Objectives: The study aimed to compare the success and reliability of an artificial intelligence application in the detection and classification of submerged teeth in orthopantomography (OPG).

Methods: Convolutional neural networks (CNN) algorithms were used to detect and classify submerged molars. The detection module, which is based on the state-of-the-art Faster R-CNN architecture, processed the radiograph to define the boundaries of submerged molars. A separate testing set was used to evaluate the diagnostic performance of the system and compare it to the expert level.

Results: The success rate of classification and identification of the system is high when evaluated according to the reference standard. The system was extremely accurate in performance comparison with observers.

Conclusions: The performance of the proposed computer-aided diagnosis solution is comparable to that of experts. It is useful to diagnose submerged molars with an artificial intelligence application to prevent errors. Also, it will facilitate pediatric dentists' diagnoses.

Completed

Looking for future studies?

Notify Me

Key information

Conditions

Artificial Intelligence Submerging Tooth

Age range

5 year–12 year

Sex eligibility

All sexes

Study type

Observational

Primary location

Seçil Çalışkan

Eskişehir, 26040, Turkey (Türkiye)

About this study

Pre-processing, Training, and Classification The study was conducted with balanced data sets. The case and control data sets were randomly divided into two parts, the training group (27 case group/27 control group) and the test group (10 case group/10 control group) to prevent the use of the visuals in the training group for retesting. The testing data set was not seen by the system during the training phase.

All 2943-by-1435 pixel images in the data set were resized to 971 by 474 pixels prior to training. All OPG images used include the whole dentitions. The training and test data sets were used to estimate and generate weight factors for the optimal CNN algorithm. An arbitrary sequence was generated using open-source Python programming (Python 3.6.1, Python Software Foundation, Wilmington, DE, USA, https://www.python.org/) language and OpenCV, NumPy, Pandas, and Matplotlib libraries. In this study, Tensorflow for model development was used to classify submerged primary molars. InceptionV3 architecture was used as transfer learning, and the transfer values were saved in the cache. Then, fully connected layer and softmax classifiers were combined to form the final model layers. The training was carried out using 7000 steps with 16G RAM and a PC equipped with NVIDIA GeForce GTX 1050.

Who can participate

Healthy volunteers accepted: Yes

Only the study team can determine whether someone qualifies for participation.

Inclusion criteria

Exclusion criteria

OPG images of poor quality (metal artifact, artifacts due to position errors during shooting, etc.) were excluded.

Treatment and study plan

deep learning

Diagnostic Test

the deep learning method is a field of study involving artificial neural networks and similar machine learning algorithms with many hidden layers.

Primary outcomes

Submerged Tooth Detection

Time frame: 6 months

The detection module, which is based on the state-of-the-art Faster R-CNN architecture, processed the radiograph to define the boundaries of submerged molars.

Sponsors and collaborators

Lead sponsor

Eskisehir Osmangazi University

Other

Registry information

Official study title

A Deep Learning Approach to Submerged Deciduous Teeth Classification and Detection

Important dates

Study start: 2019
Primary completion: 2020
Study completion: 2020
First posted: Mar 16, 2020
Registry last updated: Mar 16, 2020

OpenTrials presents study information sourced from ClinicalTrials.gov. The official registry record should be consulted for the latest information.

View the official ClinicalTrials.gov record (opens in a new tab)

This listing is for discovery and informational purposes only. It is not medical advice, does not guarantee that a study is recruiting, and does not determine eligibility. Contact the study team and a qualified healthcare professional when considering participation.

Published trials that share one or more normalized conditions with this study.

Completed