Tooth segmentation using dynamic programming-gradient inverse coefficient of variation

Abstract

Teeth provide meaningful clues of an individual. The growth of the teeth is correlated with the individual age. This correlation is widely used to estimate age of an individual in applications like conducting forensic odontology, immigration, and differentiating juveniles and adolescents. Current forensic dentistry largely depends on laborious investigation process that is performed manually and can be influenced by human factors like fatigue and inconsistency. Digital panoramic radiograph dental images allow noninvasive and automatic investigation to be performed. This paper presents analyses on third molar tooth segmentation for the population in Malaysia, ranging from persons age of 5 years old to 23 years old. Two segmentation techniques: gradient inverse coefficient of variation with dynamic programming (DP-GICOV) and Chan-Vese (CV) were employed and compared. Results demonstrated that the accuracy of DP-GICOV and CV were 95.3%, and 81.6%, respectively.