Marking Early Lesions in Labial Colored Dental Images using a Transfer Learning Approach

摘要

Dental caries, usually known as tooth decay, are prevalent in patients. According to the World Health Organization (WHO) report, a significant chunk of the overall world population suffers from dental caries. It is essential to trace out the dental cavity at every early stage to treat dental disease. It is difficult for physicians to check out the progress of growing early lesions. X-rays are good enough for severe caries but not for very early-stage lesions. Hence optical images play an essential role in the detection of early lesions. Localizing the carious dental region is also crucial as it reduces the physician recognition activity and can be used in computer-aided learning for new medical students. But unfortunately, accurate localization of the carious region from optical images is challenging. In this research, we used transfer learning to detect the carious enamel regions and locate each carious region found in optical tooth images. We consider photographic images to detect and localize carious regions. Our approach gives 95% accurate results. The proposed methodology is not harmful to health, helpful for physicians, and can be used in computer-aided learning.