Detecting Dental Caries through Captured Images Using the Machine Learning Technology Teachable Machine
Tran Tuan Anh *
Becamex International Hospital, Vietnam.
Nguyen The Huy
Becamex International Hospital, Vietnam.
Nguyen Tien Phat
Becamex International Hospital, Vietnam.
Nguyen Thi Hoai Nhi
Becamex International Hospital, Vietnam.
Vo Truong Nhu Ngoc
School of Odonto-Stomatology, Hanoi Medical University, Vietnam.
Tran Hoang Anh
Binh Duong Provincial General Hospital, Vietnam.
*Author to whom correspondence should be addressed.
Abstract
Objective: The objective is to utilize open-source artificial intelligence tool Teachable Machine to detect dental caries.
Subjects and Research Methods: cross-sectional description. The research involved analyzing a total of 2,063 digital images, comprising 1,563 images showing dental caries and 500 images depicting teeth without any signs of dental caries.
Results: Among the 1,563 images featuring dental caries, the recognition tool accurately identified 1,512 images (96%), while 51 images remained undetected, representing 4% of the total. Among the entire set of 2,063 images, encompassing both those with and without dental caries, 1,512 images were accurately identified (73.3%), while 551 images (26.7%) were not detected to have dental caries.
Conclusion: The study on 1,563 images with dental caries using the Teachable Machine learning tool yielded promising results, achieving a high accuracy rate of 96%. However, when applied to the mixed dataset of 2,063 images, the accuracy rate for identifying images with dental caries dropped to only 73.3. The variance is ascribed to the initial stages of dental caries, which closely resemble the color of healthy tooth enamel. Therefore, the research team proposes the necessity for additional data on this form of decay to enhance classification and identification accuracy.
Keywords: Artificial intelligence, machine learning, dental caries, teachable machine, dental cariestooth enamel, tooth enamel, X-ray