Efficient lung tumour detection and Isotropic volume reconstruction With gesture interaction in ct images

Abstract

Lung cancer is the leading cause of cancer-associated deaths in the newlineworld, and often requires surgical intervention for tumour therapy. The main newlineobjective of the research is to develop a powerful and useful system that newlineprovides a novel technique for lung tumour segmentation and isotropic newlinevolume reconstruction of the tumour. The thesis is also directed towards newlineproviding sterile, contactless, gesture-based, computer-aided guidance for newlineaccessing medical data during surgery. newlineThe quality of lung CT image is directly proportional to the newlinepatient s absorbed dose. An increase in the absorbed radiation consequently newlineincreases the absorbed dose by the patients, thereby, includes noise in the newlinereconstructed image. Also, the lung CT image includes soft tissues which newlineshare indistinct boundaries and have similar intensity, thus denoising newlineprocedure is necessary for further processing. The difficulty lies in reducing newlinethe noise in the input lung CT image without losing the visual features of the newlineimage such as edges, contrast, and other sharp corner structures. newlineInitially, the lung CT image is pre-processed using a denoising newlinealgorithm. Uniformly, the input lung CT images are degraded by adding newlineadditive white Gaussian noise of level 0.06dB to every image. Transformbased newlinealgorithms like Curvelet and Contourlet denoising methods are used to newlinedenoise the degraded CT slices. The multi-resolution property of the newlineContourlet transform is achieved through the four-level Laplacian pyramidal newlinedecomposition of a corrupted image into a low pass and bandpass output. The newlinedirectionality property of the Contourlet is achieved through the multilevel newlinedecomposition of the band-pass output by the directional filter bank. Thus, newlineContourlet coefficients of the noisy input image are obtained at each level of newlinethe directional filter bank. newline