A powerful Transfer learning technique for multiclass classification of lung cancer CT images

Abstract

Lung cancer is a lethal disease caused by unusual cell growth in the lungs. Early cancer detection leads to potent treatment planning. Precise identification of different types of nodules in CT images through naked eyes becomes arduous for radiologists. Transfer learning-based computer-aided detection system has shown effectual results in providing a second opinion to the radiologist. This paper proposes an EfficientNet-based transfer learning model for multi-class classification of benign, normal and malignant CT images. Experimental results revealed that the proposed model obtains accuracy, precision, recall, the area under curve and F1-score of 100% each. The classification model excelled over the different variants of EfficientNet and other pre-trained networks. Thus, the proposed multi-class EfficientNet model is felicitous for early lung cancer detection.