Detection of Lung Nodules from CT Images using Image Processing Techniques

Conference: McGraw-Hill International Conference on Signal, Image Processing Communication and Automation
Author(s): Swathi H K, Nanda S Year: 2017
Grenze ID: 02.MH-ICSIPCA.2017.1.50 Page: 317-323

Abstract

Lung cancer is the common cause of death among people throughout the world. So, early detection of lung cancer\ncan increase the chance of survival among the people. The overall 5-year survival rate for lung cancer patients increases from\n14 to 49% if the disease is detected in time. Computed Tomography (CT) can be more efficient than X-ray in detecting the\nlung cancer. In this paper, an algorithm is developed for the segmentation of lung from CT images using optimal\nthresholding technique. Then Vector quantization (VQ) is performed on Lung CT images to segment lung nodule. In vector\nQuantization codebook is generated for each image and the training process is done according to the input data. The\nperformance evaluation of segmentation is done using measures like Area, Sensitivity, DICE co-efficient, Jaccard Index,\nHausdroff distance and Mahalanobis distance. Texture features, , morphological features and wavelet based features are\nextracted from the segmented nodule. K-nearest neighbor (KNN), Support Vector Machine and Decision tree are used for the\nclassification of lung nodules as benign or malignant. The performances of these classifiers are compared.

<< BACK

MH-ICSIPCA - 2017