ML-Based Tooth Shade Assessment to Prevent Metamerism in Different Clinic Lights

Abstract

The aesthetic understanding has found its place in dental clinics and prosthetic dental treatment. Determining the appropriate prosthetic tooth color between the clinician, patient and technician is a difficult process due to metamerism. Metamerism, known as the different perception of the color of an object under different light sources, is caused by the lighting differences between the laboratory and the dental clinic. The traditional trial-error color determination method, coupled with the high cost of instrumental color value determination, has prompted the need for alternative technologies. The integration of AI technologies into dental practices aims to minimize errors in tooth shade assessment, reduce equipment usage, eliminate the impact of clinic lighting on color detection, and decrease costs for patients, dentists, and laboratories. In this study, a machine learning (ML) based approach that can correctly detect tooth shade even under different clinical lights has been developed. A dataset consisting of 580 dental images taken under four different clinical lights and with five repetitions was created using the Vita color shade guide. Experimental studies were performed using the HSV color space, 6 different ML algorithms and color histograms. As a result, 97.93% accuracy rate was achieved by using cross-validation (cv = 5) in the classification of 29 color values independent of clinical lights. It has been shown that the tooth colors can be determined with high accuracy using ML algorithms and metamerism can be prevented.