Convolutional Neural Networks for Classification of Lung Cancer Based on Histopathological Images

Abstract

Lung cancer is one of the deadliest types of cancer characterized by the uncontrolled growth of cancer cells in the lung tissue due to the accumulation of carcinogens. Lung cancer ranks second in the most cases with 2.206 million new cases and ranks first in deaths. This lung cancer often does not cause symptoms in the early stages, because it only appears after the tumor is large enough or the cancer has spread to surrounding tissues or organs, so it is necessary to have early detests to prevent severity and determine follow-up treatment. This study aims to classify lung cancers using digital pathology images with data of 15000 images obtained from the LC25000 dataset containing 5,000 images for each class. The method used in this classification process uses convolutional neural networks (CNN) which is one of the implementations of Deep Learning used for digital image processing. Using this method, the doctor can diagnose and find out the type of lung cancer quickly without spending much time. Thus, the faster the prediction results received by the doctor / health expert, the faster the next action or handler will be, this study produces a fairly accurate accuracy value even though it uses a shallow CNN architecture because it only consists of 5 layers with 3 convolution layers and 2 fully connected layers, with the resulting accuracy value of 98.53%.