Fusion of CNN and 2DCNN features in Lung nodule Malignancy Classification

Abstract

Pulmonary cancer is one of the diseases in the people of the whole world. Doctors use different methods for recognition and diagnosis of pulmonary cancer, such as CT, MRI, isotope, and X-ray. Such manual methods failed to correctly classify the lung nodule malignancy. That causes lung-infected patients rate of survival get reduced due to late detection of cancerous lung nodule malignancy. To solve this classification problem and early detection of lung nodule malignancy, the feature fusion of CNN and 2DCNN deep learning algorithms are proposed to obtain the different feature vectors. Which are fused using concatenation fusion. Softmax use fused features to classify the malignant nodule. To locate the cancerous pulmonary nodules from the provided real input pulmonary CT images. The YOLO8 algorithm is proposed at last to find the location of lung nodules in the classified lung malignant image. The implemented system is tested on a real data set collected from Latur Super Specialty Hospital. From the implemented experiment result, the overall accuracy of the system is improved than the existing system. The CNN and 2DCNN features fusion attain 98.56% accuracy.