Development of machine learning model for the field of dentistry
In recent years, oral health problems have gained increasing attention, and oral examinations have become a part of people's daily lives. When facing complex dental problems, dentists usually choose to use dental radiographs to assist in diagnosis. However, the use of dental radiographs for dia...
Saved in:
Main Author: | |
---|---|
Other Authors: | |
Format: | Thesis-Master by Coursework |
Language: | English |
Published: |
Nanyang Technological University
2023
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/169098 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
Summary: | In recent years, oral health problems have gained increasing attention, and oral examinations have become a part of people's daily lives. When facing complex dental problems, dentists usually choose to use dental radiographs to assist in diagnosis. However, the use of dental radiographs for diagnosis relies on the experience of the dentist. Additionally, prolonged observation of dental radiographs with the naked eye can cause fatigue and even lead to incorrect judgments.
In recent years, with the development of the artificial intelligence industry, object recognition technology has made significant progress. The application of related technologies in the field of dentistry has also emerged. It has become common to use algorithms to detect object in dental radiographs. YOLO, as a single-stage algorithm, has the advantages of small memory requirements, fast processing speed, and high detection accuracy, making it very suitable for use in dental hospitals that cannot deploy large-scale computer equipment.
The main purpose of this dissertation is to use YOLO to detect objects in dental radiographs. According to the FDI tooth numbering system, 32 teeth in the adult oral cavity are numbered as 11-18, 21-28, 31-38, and 41-48. In addition to teeth crowns, root canals, and implants, a total of 35 categories of targets are detected. This dissertation chooses to use YOLOv5, which is implemented in Python programming language, and is trained on a dataset of 565 dental radiographs, achieving an accuracy rate of over 85% for prediction results. |
---|