A combination of FractalNet and CNN for Lung Nodule Classification

Lung Cancer is the fastest growing cancer around the world and is mostly diagnosed at an advanced stage. In order to prevent the delay in diagnosis, several medical imaging modalities like CT (Computed Tomography) has been used. So far researchers have used several machine learning architectures to classify the lung nodule captured in CT scans into benign or cancerous. In this paper, we use a deep learning architecture which combines fractalnet and CNN to classify the lung nodule into benign and malignant. The lung nodule classification was validated on LUNA dataset on which it achieved an accuracy of 94.06%, sensitivity of 97.52%, specificity of 86.76% and area under receiver operating characteristic (AUC) score of. 98, using the proposed architecture.