Extraction of lung nodules in HRCT images using partial differential equation based model

Abstract

Detection of a lung nodule in a chest CT favors in understanding the malignant behavior of nodules. The size of lung nodule which reflects the malignant nature helps in early diagnosis and treatment of lung cancer. False detection of lung nodule will misinterpret healthy patient as lung cancer patient which may lead to wrong medication by the clinician. Existing algorithms for detection of lung nodule like Active Contour Model, Markov Gibbs Random Field model, Expectation Maximization algorithm, Active Appearance Model etc which solves the problem for certain extent, but have drawbacks such as large parameterization, re-initialization, blurring of image, failure to enter concavities, intensity in homogeneity and inability to differentiate different types of nodules. To overcome these drawbacks, a realistic approach is proposed by using partial differential equation based technique. The proposed method adopts Perona-Malik model for enhancement procedure and geometric active contour model for segmentation. Proposed method can efficiently deblur the contours of the original image containing weak or blurred edges. Pre-processed CT images are successfully segmented using geometric active contour model. The contribution of the proposed approach has been evaluated by estimating the efficiency and sensitivity for various values of control parameters with the existing system for various variances.