A Fine Tuned Deep Learning Model for Accurate Cancer Cell Classification

Abstract

Lung cancer is one of the deadly disease around the world. In order to cure the disease appropriately, it is essential to identify the type of cancer lungs have. Therefore, the pathological analysis of cancer cells has been performed to identify the category of cancer. This categorization requires an expert and misclassification can impact the entire process of dealing with the disease. In this context, proposed work involve a deep learning model for identifying the cancer cell’s type in to three categories namely Lung benign tissue, Lung adenocarcinoma, and Lung squamous cell carcinoma. In order to simulate the proposed model utilizes a dataset obtained from kaggle. This dataset is consist of a total of 15k images of three cancer classes, each class has a total of 5k images. There are two deep learning architectures sequential convolutional neural network (SCNN) and 2D-CNN has been trained. In this experiment, 2D-CNN model was found promising but the hyper parameter tuning has been performed. Based on the performed parameter tuning of the 2D-CNN model the following parameters are found optimal for classifying the cancer cell accurately.