Classification of Benign and Malignancy in Lung Cancer Using Capsule Networks with Dynamic Routing Algorithm on Computed Tomography Images

There is a widespread agreement that lung cancer one of the deadliest types cancer, affecting both women and men. As result, detecting at an early stage crucial to create accurate treatment plan forecasting reaction patient adopted treatment. For this reason, development Convolutional Neural Networks (CNNs) for task classification has recently seen trend in attention. CNNs have great potential, but they need lot training data struggle with input alterations. To address these limitations CNNs, novel machine-learning architecture capsule networks been presented, it potential completely transform ares deep learning. Capsule networks, which are focus work, interesting because can withstand rotation affine translation relatively little data. This research optimizes performance CapsNets by designing new allows them perform better on challenge classification. The findings demonstrate proposed Network method outperforms challenge. CapsNet single convolution layer 32 features (CN-1-32), 64 (CN-1-64), double (CN-2-64) three Capsulel developed Lung nodules, benign malignant, classified using CT images. LIDC-IDRI database was utilized assess those networks. Based testing results, CN-2-64 network performed out tested, specificity 98.37%, sensitivity 97.47% accuracy 97.92%.