Recognition of Lung Tumor Area Based on Watershed Segmentation and CNN

Abstract

Early recognition of tumor would assist in saving an enormous number of lives over the globe frequently. Study and remedy of lung tumor have been one of the greatest troubles faced by humans over the latest few decades. Effective recognition of lung tumor is a vital and crucial aspect of image processing. Several Segmentation methods were used to detect lung tumor at an early stage. An approach is presented in this paper to diagnose lung tumor from CT scan images. The input image (CT scan image) will be preprocessed initially using median filter to remove the noise. After applying preprocessing technique, the Dual-Tree Complex Wavelet Transform (DTCWT) segmentation technique is used for the edge detection. The Gray-Level Co-occurrence Matrix (GLCM) features are calculated based on the pixel values of the extracted image. These features can be compared with database images using Convolutional Neural Network (CNN) which facilitates in categorizing it as tumorous. After confirming that the affected area is tumorous, watershed segmentation algorithm is used to get the color features of the tumor.