Comparative Evaluation of Spectrophotometric and Artificial Intelligence-Based Tooth Shade Determination

Abstract

Aim: The aim of the present study was to evaluate the agreement between spectrophotometric tooth shade determination and an artificial intelligence-based method using standardized digital images of maxillary central incisors. Materials and Methods: A comparative clinical study was conducted on 25 patients aged between 18 and 30 years with intact natural dentition in the anterior maxillary region. A total of 50 teeth were examined, including the maxillary right central incisor 11 and the maxillary left central incisor 21. Tooth shade was determined using SpectroShade Micro II/SpectroShade according to the VITA Classical and VITA 3D-Master shade systems. Standardized digital photographs were obtained using a Canon EOS SL3 camera in manual mode with fixed exposure parameters. A grey card was included in each image for white balance standardization. The images were processed in Adobe Lightroom Classic software and subsequently analyzed using ChatGPT based on the GPT-5.5 Thinking model. Agreement between SpectroShade and ChatGPT was assessed by calculating the percentage of complete agreement and the unweighted Cohen’s kappa coefficient. Results: Complete agreement between SpectroShade and ChatGPT was observed in 32 out of 50 teeth for both shade systems, corresponding to an agreement rate of 64.0%. The Cohen’s kappa value was κ = 0.402 for VITA Classical and κ = 0.494 for VITA 3D-Master, indicating moderate agreement between the two methods.  Conclusion: Within the limitations of the present study, the artificial intelligence-based method demonstrated moderate agreement with spectrophotometric shade determination. ChatGPT may be considered a potential auxiliary tool for tooth shade assessment when standardized photographic protocols are applied. However, it should not be regarded as a substitute for spectrophotometric evaluation.