Classification of Lung CT scan images using Image Processing and CNN

ABSTRACT Lung cancer is the second most common disease in men and women.The mortality rate of lung tumour is the highest among all other types of tumour. Early detection of lung cancer can increase the possibility of survival among people. Lung cancer is also found by imaging tests such as chest computed tomography scan because it provides more elaborate picture. Computed Tomography (CT) are said to be more effective than plain chest X-ray in detecting and diagnosing the lung cancer. To classify the samples of lung CT scan images into normal, benign and malignant categories, a method is developed by implementing image processing and soft computing techniques. In this work, new algorithm is developed using image processing and machine learning technique to observe the cancer at early stage with additional accuracy. Image processing involves the pre-processing that is image smoothing, enhancement and segmentation. Once pre-processing of images is done, they are provided to Convolutional Neural Network classifiers for classification into cancerous or non-cancerous image.The implemented system gives an accuracy of 98.6 % for classification of samples into normal and abnormal classes and an accuracy of 99.5 % for the classification of samples intobenign and malignant categories. Keywords: Image Processing, Convolutional Neural Network classifier, CT scan, Lung Cancer Detection