Evaluation of VITA shade-based tooth color categories using deep learning

With the increasing interest in dental aesthetics, more patients are seeking tooth shade evaluations and whitening treatments. However, traditional methods of visually assessing tooth color with a commercial shade guide are often subjective, emphasizing the need for objective and reliable approaches in clinical practice. This study aimed to develop and validate a deep learning model for assessing tooth shade, comparing its accuracy with that of experienced dental practitioners. Seventy adult participants underwent tooth shade evaluation using the VITA Classic shade guide, with high-resolution intraoral images analyzed using deep learning techniques to detect, segment, and classify tooth shades. The performance metrics-namely, accuracy, precision, recall, and the F1 score-were employed to evaluate the tooth shade classification of deep learning models derived from six modified CNN architectures. The ResNet model demonstrated the most optimal results. A subsequent McNemar's test, which was utilized to make a comparison between the deep learning model and experienced dentist, yielded statistically significant results. The deep learning model provides a reliable and efficient alternative for assessing tooth shade assessment, minimizing subjectivity and outperforming experienced clinicians in accuracy.