Análise facial automatizada através de inteligência artificial: validação do algoritmo

Abstract

The acquisition of facial measurements plays a crucial role in dentistry, particularly in the planning phase of dentures. Measurement methods are based on observing specific anatomical features and manually measuring using calipers and specialized rulers. Aiming to create a more accurate and practical method of obtaining facial measurements employing cutting-edge digital technologies, this study intended to validate the application's ability to measure distances between facial anatomical landmarks. Additionally, the agreement between the results generated by the application on different smartphones and the agreement between each smartphone's results and measurements taken using calipers were assessed. Through the use of artificial intelligence, the application identifies multiple facial anatomical points, allowing for the automatic measurement of various dimensions from a single frontal image of the patient. This study involved a sample of 105 participants and the measurements evaluated were: mouth width between commissures, facial width at the level of the zygomatic arch, nose width, intercanthal width, height between subnasal and menton, and between nasion and menton points. Statistical analysis employed the Intraclass Correlation Coefficient (ICC) and Bland-Altman plots. Results demonstrated good to excellent agreement (ICC between 0.849 and 0.918) when comparing results across smartphones. It was possible to observe agreement between 70% and 90% between smartphones and calipers, after establishing a clinical tolerance limit of 10% of the mean for each dimension evaluated. The application successfully measured distances between facial anatomical points, indicating the promising use of digital technologies, such as smartphone applications, in dentistry. The results obtained by the automated facial measurement app demonstrate clinical relevance, providing a viable and effective alternative to calipers.