Lung Cancer: Detection and Classification of Malignancies

Abstract

AbstractLung cancer is one of the severe diseases nowadays and one of the major causes of growing deaths. So, medical experts believe that diagnosing malignant cells in the early phase can reduce the risk of patients’ death. Present-time Deep learning is contributing lots of services in different sectors for automating human tasks with ease. As the introduction of many deep learning models such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and various other models. With little advancement in these models, humans are capable of solving complex tasks such as image recognition, natural language processing, object detection, object segmentation, and language translation and achieved state-of-the-art performance. However, problems like lack of viewpoint invariance lead to the introduction of a supervised deep learning model known as Capsule Network (CapsNet) with the intent of achieving human-level efficiency. Our proposed methodology is based on CapsNet to detect lung cancer nodules and further, we apply pre-trained FixEfficientNet to classify CT images as it requires fewer parameters. We train these models using publicly available datasets: LUNA-16 and LIDC-IDRI and validate on these benchmarks. We achieved an accuracy of 98.47% in nodule detection and 93.59% in nodule classification.KeywordsLung cancerNoduleCTComputer-Aided diagnosis(CAD)DetectionClassificationCapsule network(CapsNet)MalignantBenign