Teeth/palate and interdental segmentation using artificial neural networks

Abstract

We present a computational system that combines Artificial Neural Networks and other image processing techniques to achieve teeth/palate segmentation and interdental segmentation in palatal view photographs of the upper jaw. We segment the images into teeth and non-teeth regions. We find the palatal arch by adjusting a curve to the teeth region, and further segment teeth from each other. Best results to detect and segment teeth were obtained with Multilayer Perceptrons trained with the error backpropagation algorithm in comparison to Support Vector Machines. Neural Networks reached up to 87.52% accuracy at the palate segmentation task, and 88.82% at the interdental segmentation task. This is an important initial step towards low-cost, automatic identification of infecto-contagious oral diseases that are related to HIV and AIDS. © 2012 Springer-Verlag.