Classification of Early Childhood Caries (ECC) Severity Using Spectroscopic Analysis and Artificial Intelligence

Early childhood caries (ECC) continues to pose a significant challenge for preschoolers worldwide, with a global prevalence of 46.2% in 2020. In the modern world of telehealth prominence, addressing ECC screening efficiently can be achieved through the integration of AI technology and saliva sample analysis. This framework allows parents and teachers to collect saliva samples from the children remotely to reduce the need for in-person healthcare visits. These samples will then be sent to healthcare facilities for analysis and attention will be given to screening results that indicate medium to high caries risk. In this study, a cohort of 104 kindergarten students have voluntarily provided saliva samples, along with additional parameters like pH, viscosity, quantity of saliva and hydration. The saliva was analyzed using the laser-induced breakdown spectroscopy (LIBS) and the results were used for artificial neural network (ANN) development to classify ECC severity. Two ANN models were developed with Model I used multivariate inputs consisting of pH, viscosity, hydration and spectroscopic saliva results whilst model II is developed based on spectroscopic saliva results only. Both ANN models are capable of effectively predicting ECC risk categories. The first model with multivariate inputs achieved performance accuracy of 91.8% whilst the second model, which relies solely on saliva spectroscopic data, exhibited performance accuracy of 92.7%. This study concluded that the use of telehealth and modern technologies such as LIBS and AI for ECC screening is helpful in ensuring a more accessible and efficient healthcare for young children by revolutionizing the healthcare approach.