Feature selection and regional labeling of significant detection for pulmonary nodules in CT images

Abstract

The function of visual attention mechanism is to acquire the useful visual information at the fastest speed. The Itti visual attention model commonly used at present has achieved good effects in natural image. In order to find the region of interest as soon as possible, this paper attempts to introduce visual attention mechanism into pulmonary nodules detection. However, the Itti model is more to detect significant regions in image as a whole, and it does not reflect size and shape of significant goal. In order to improve detection accuracy, this paper attempt to detect pulmonary nodules by Itti model combined with features of pulmonary nodules. Some primary features such as gray, direction, corner point, edge and local entropy were chosen to generate saliency map. This paper compares emphatically their respective effects and marks the significant areas that they have detected in original image.