Radiographs are the most accurate diagnostic aid for the detection of osseous abnormalities in the maxilla and the mandible. Density and gray-scale changes in radiographs are important visual features the clinicians use to evaluate changes in bone pattern. In this work we present a quantitative study on different regions of periapical images by means of statistical textural features for classification purposes. We employed the Co-occurrence Matrices method for extracting features, and a multilayer perceptron as classifier. Our analysis has been performed on a database of 54 images; from every image two Regions Of Interest (ROIs) were selected, corresponding to regions where the periapical lesion was visible or not. Two different combinations of learning and testing sets were used; classification performance has been evaluated with ROC analysis. The obtained results show the effectiveness and robustness of this representation and, at the same time, encourage the development of this approach in order to obtain a followup system for supporting the decision-making process by clinicians.
Analysis of Periapical Lesion Using Statistical Textural Features and Neural Networks / Caputo, B.; Gigante, Giovanni Ettore. - In: PHYSICA MEDICA. - ISSN 1120-1797. - 17:2(2001), pp. 67-70.
Analysis of Periapical Lesion Using Statistical Textural Features and Neural Networks
GIGANTE, Giovanni Ettore
2001
Abstract
Radiographs are the most accurate diagnostic aid for the detection of osseous abnormalities in the maxilla and the mandible. Density and gray-scale changes in radiographs are important visual features the clinicians use to evaluate changes in bone pattern. In this work we present a quantitative study on different regions of periapical images by means of statistical textural features for classification purposes. We employed the Co-occurrence Matrices method for extracting features, and a multilayer perceptron as classifier. Our analysis has been performed on a database of 54 images; from every image two Regions Of Interest (ROIs) were selected, corresponding to regions where the periapical lesion was visible or not. Two different combinations of learning and testing sets were used; classification performance has been evaluated with ROC analysis. The obtained results show the effectiveness and robustness of this representation and, at the same time, encourage the development of this approach in order to obtain a followup system for supporting the decision-making process by clinicians.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.