Dynamic And Volumetric Approach For Segmentation And Classification Of Lung Ct Images

Abstract

Lung cancer is one of the dangerous deadly diseases for individuals around the world. Thus, the survival rate is low due to difficultly in detecting lung cancer at advanced stages like symptoms and thus prominence for early diagnosis is important. Research on tumor identification in various parts of the body is of challenging task. Amongst all Lung tumor detection is of major concern. Various screening techniques are available like X-Ray, CT, Sputum Cytology, here CT images are considered for identification of Lung tumor. Computed Tomography has been widely exploited for various clinical applications. The early diagnosis and treatment of lung tumor can help to improve the survival rate and CT scan is the best modality for imaging lung tumor. In many cases, when the nodules are identified it might be either more advanced or too large to be effectively cured. Physical characteristics of the nodules such as the size, tumor type and type of borders are very significant in the examination of nodules. The detection and treatment of lung cancer will be of great importance for early diagnosis. Machine learning classification can benefit greatly from the wealth of research on the use of image processing for detecting lung cancer. Image segmentation for dividing malignant nodules has made it more easier to identify tumor subtypes and assess disease progression. Using watershed segmentation and region growth segmentation for processing CT images of lung cancer yields positive results but better segmentation could help to increase the accuracy of the result for which a two level segmentation algorithm is implemented. In addition, support vector machines (SVMs) and neural networks (NNs) are crucial in determining the subtype of lung cancer. We propose VGG-16-T as a sophisticated deep convolutional network architecture. VGG-16-T makes extensive use of small convolution kernels of size [3, 3] to support and depict profundity.