Dental tissue classification using computational intelligence and digital image analysis

Abstract:

Detailed visual inspection is a fundamental skill at the core of routine procedures in clinical dentistry. In this study we explore the possibility to achieve a more precise visual inspection process by combining computer vision and intelligence techniques. The aim of this paper is to propose a dental tissue classification system based on digital image analysis and computational intelligence techniques. A four-step computational model consisting of image acquisition, segmentation, feature extraction and tissue classification is defined. A set of 300 samples (N =300) were acquired from a dental picture database consistent of occlusal photographs of vital teeth. Each sample was segmented by a non- parametric Mean-Shift algorithm. Color and texture descriptors were extracted from each segmented region. A group of five professional dentists identified and labeled each one of the segmented regions as the desired classification targets. Finally, a multilayer perceptron for pattern recognition was trained using extracted features as inputs, and labeled regions as outputs. The proposed dental tissue classification procedure achieves high global performance rates: Average sensitivity =79.91%, average specificity =75.2%, average accuracy =81.4%, average AOC =0.84. Mean execution time was ∼12 s. The proposed model can be used to provide objective and measurable data of visually assessed dental structures and its surroundings in clinical procedures. © 2014 Taylor & Francis Group.

Año de publicación:

2014

Keywords:

Fuente:

scopus