Analysis of Lung Cancer by Using Deep Neural Network

Abstract

AbstractLung cancer is one of the world’s deadliest cancers and one of the highest mortality rates. There has been a recent increase in the prevalence of lung cancer. The key aim of this research was to create a computer-aided diagnostic (CAD) method for classifying histopathological photographs of lung tissues. For the creation and validation of CAD, we used a publicly available dataset (15,000 samples) of histopathological photographs of lung adenocarcinoma, lung squamous cell carcinoma, and benign lung tissue from three different types. In order to extract image features, multi-scale processing was used. Finally, the comparison study has been made based upon seven pre-trained convolution neural network (CNN) models, including MobileNet, VGG-19, ResNet 101, DenseNet 121, DenseNet 169, Inception V3, InceptionResNet V2, and MobileNetV2 for classification of lung cancer. All pre-trained models are hyper-tuned by considering several factors such as batch size, learning rate, number of epochs, and model accuracy. Among all, ResNet 101 had the best accuracy of these CNN version, at 98.67%. This research will aid researchers in the development of more successful CNN-based lung cancer detection models.KeywordsTransfer learningLung cancerClassificationCNNMulti-scale processing