Detection of rounded opacities on chest radiographs using convergence index filter

This paper presents a method to detect rounded opacities on digital chest radiographs. Based on a model of rounded opacities, three filters that evaluate the convergence degree of gradient vectors in the neighborhood of the pixel of interest are investigated to detect rounded opacities in the complex lung area. These filters do not depend on the contrast of the rounded opacity to its background. Their outputs depend on only the distribution of the gradient vector orientation. Experiments to test their performance have been performed. Their results show the effectiveness of the proposed filters. Especially, the iris filter and the adaptive ring filter are shown to be very effective in detecting cancerous tumor candidates. Another experiment to compare the performance of the iris filter and the Min-ΦDD filter which is one of the filters effective in detecting rounded opacities have been performed and the superiority of the proposed filters has been shown