A Segmentation Based Technique for Detection of Lung Cancer using Deep Learning

Abstract

The identification of lung cancer has emerged as a significant problem in medical research in the past few decades. This study used 8000 CT lung pictures of lung images, categorized into four separate classes: adenocarcinoma, Large-cell carcinoma, normal and squamous cell carcinoma. Deep learning algorithms designed for lung cancer detection are evaluated against five classifier methods employing the SKlearn (Scikit-learn) framework: Single Layer CNN, Multi-layer CNN, VGG16, ResNet50, and Artificial Neural Network (ANN). The outcomes of the suggested five classifiers have been analysed. A model based on a convolutional neural network (CNN) was developed using the ReLU activation functions. Construct a fundamental CNN model and evaluate it with regularization and augmentation methods to enhance accuracy. The accuracy of the Multi-layer CNN is 99%. Logistic regression yields 80%, whereas Single Layer CNN, VGG16 and ResNet50 achieved 97%. We got the least accuracy with 39% from ANN. In the comparison of deep learning algorithms with Multi Layer CNN, the latter had a superior accuracy of 99% in the detection of cancer in CT lung pictures.