Abstract

Lung Cancer is one of the most deadly cancer in the world. It has a high demise rate and its cases are increasing day by day in throughout world due to people lifestyle. Cancer diagnosis and treatment has been one of the most difficult problems that humanity has encountered in past few year. If it has not been detected at early stage then the patient survival chance will be very less. Detecting lung cancer at early stage Computer Aided Diagnosis (CAD) has become an indispensable tool for supporting radiologists' CT interpretations. Its speedup the whole process of analysis of cancer. This paper describes a automatic method for classifying tumours seen on computed tomography scans of lung cancer as malignant or benign using a Convolutional Neural Network (CNN). For classification of computed tomography scan we use thresholding segmentation as initial segmentation to segment out lung tissue from computed tomography in CNN but this seems to be inadequate so that we modified U-Net trained to predict nodule. CT scans with segmented lungs were fetch into Convolutional Neural Networks to prove the CT scan as positive or negative for lung cancer based on the location. CAD system performance depend on various factor several training and testing stage, required lots of label data for training.