Lung Cancer Detection from Computed Topography Images using Marker controlled Watershed

Abstract

The death count due to lung cancer is increasing day by day. According to the statistics calculated by world health organization (WHO), the estimated deaths are around 2,28,150 (116,440 in men and 111,710 in women). The statistics shows the lung cancer consists of about 14 percent among other cancers. It ranks second place in each woman and men. This research is carried out using lung cancer CT scan images as input data. In this research, the proposed methodology is implemented in three stages. In stage one, preprocessing is done using gaussian filter and Gabor filter. The image smoothing is accomplished with gaussian filter and image enhancement is carried out using Gabor filter. In stage two, image segmentation is done using marker-controlled watershed algorithm it segments the lung portion only. In stage three, binarization is used for detection and classification is carried out based on the black and white pixels. In the binarization method, if black pixel count is less than 17179, then lung cancer is detected. The adopted Navies Bayes classifier shows an accuracy of 94.6 percent. Gabor filter gives best in terms of texture analysis and intensity when compare with a gaussian filter. Gabor filter increases the contrast in nodule areas that are very helpful for cancer detection.