Fully automated CAD system for lung cancer detection and classification using 3D residual U-Net with multi-region proposal network (mRPN) in CT images

Lung cancer is one of the leading causes of mortality worldwide. The survival rate of lung cancer depends on its timely detection and diagnosis. For pulmonary cancer detection, numerous Computer-Assisted Diagnosis (CADx) systems have been developed that use the CT scan imaging modality. Recent advancement in deep learning techniques has enabled these CADx to automatically model high-level abstractions in CT-Scan images using a multi-layered Convolutional Neural Network (CNN). Our proposed CAD system comprises 3D residual U-Net for nodule detection. Initially, the 3D residual U-Net resulted in false positive results; therefore, a multi-Region Proposal Network (mRPN) was proposed for the improvement of nodule detection. The detected nodules are assigned a probability of malignancy. Furthermore, each detected nodule is classified into four classes based on its respective malignancy score. Extensive experimental results illustrate the effectiveness of our 3D residual U-Net model. These results demonstrate the exceptional detection performance achieved by our proposed model with a sensitivity of 97.65% and an average classification accuracy of 96.37%. Performance analysis demonstrates the potential of the proposed CAD system for the detection and classification of lung nodules with high efficiency and precision.