Classification of Lung Images Using Deep Convolutional Neural Network

Abstract

The history of medical imaging clearly portrays numerous computer aided diagnosis system (CAD) which was successfully used and implemented to assist radiologists about their patients. Medical image analysis had taken great hike over two decades using Artificial Neural Network for its task but since recent past it is being taken over by the Convolutional Neural Network and has also gained high popularity in medical imaging. CNN has mainly been developed as medical images possess high semantic features. In this paper, the tasks proposed ideology is on novel deep convolution neural network (DCNN) based method for lung normality classification. The extracted deep features from computer tomography (CT) images of the lungs are widely further used to classify the lungs abnormality i.e either as malignant or benign. Suitable modifications are performed to produce an acceptably high accuracy rate, thereby reducing the computational complexity rate. The proposed methodology involves the role of fully connected layer. While nearing to the outcome before which this layer plays a vital role in acquiring the desired classified images as per the requirement once the convolution process is finished. Therefore, this methodology is likely to be found only supportive to the system formed and thus improvising the accuracy level.