Proficient lungs nodule detection and classification using machine learning techniques

Abstract

Amongst all cancerous diseases, lung cancer is the most fatal disease, if it is not detected at initial stage. Lots of image processing techniques have been used for its detection but still there is room for improvement. In this paper a novel region based ACM with Otsu preprocessing technique is proposed for the segmentation phase of the lung nodule detection. From segmented images features are extracted and for classification SVM classifier is used. We have used the dataset of 12 images taken from LIDC. The proposed technique provides very proficient results. For comparative analysis we have also tested it with other different segmentation techniques. It is evaluated that proposed techniques has several advantages over the others. It can efficiently detect the boundaries of lung part from images, irrespective of contour initial position. It takes less computation time and no of iterations than others. Moreover it can also detect weak and blur boundary in images. Experimental analysis on the lung images reveals the advantages of the proposed technique over geodesic active contours (GAC), Chan–Vese (C–V) active contours, simple ACM and DRLSE in terms of both effectiveness and accuracy. With this view, the present study will facilitate the radiologist as well as the physicians.