Quantitative Evaluation and Comparison of Lung Contour Extraction Methods

• Author(s): José Silvestre Silva ; Augusto Silva ; Beatriz Sousa Santos
• Date: 2012
• Origin: Electrónica e Telecomunicações
• Subject(s): pulmonary contours; Lung Contour; automated methods; CT images; pulmonary contours; Lung Contour; automated methods; CT images
• Document Type: Article
• Language: English

Description

Pulmonary contours in chest CT images are closed curves that may, in some cases, exhibit multiple concave and convex irregularities. In this paper we assess the quality of pulmonary contours obtained by several fully automated methods. A comparison among these contours and corresponding manually drawn reference contours was performed and several figures of merit were assigned to each one of the automated contours extraction algorithms.Likeness between a computed contour and reference contourwas measured through the Pratt figure of merit, similarityindex, mean error and fraction of errors greater than 5 pixels. For each of these figures we present an exploratorystatistical analysis and discuss its sensibility to thelongitudinal location of the CT slice. Pulmonary contours in chest CT images are closed curves that may, in some cases, exhibit multiple concave and convex irregularities. In this paper we assess the quality of pulmonary contours obtained by several fully automated methods. A comparison among these contours and corresponding manually drawn reference contours wasperformed and several figures of merit were assigned to eachone of the automated contours extraction algorithms.Likeness between a computed contour and reference contourwas measured through the Pratt figure of merit, similarityindex, mean error and fraction of errors greater than 5pixels. For each of these figures we present an exploratorystatistical analysis and discuss its sensibility to thelongitudinal location of the CT slice.