Detection and classification of lung cancer using CNN and Google net. (December 2022)

• Record Type: Journal Article
• Authors: Pandian, R.; Vedanarayanan, V.; Ravi Kumar, D.N.S.; Rajakumar, R.
• Is Part Of: Measurement. Volume 24(2022)
• Publication Date: 2022-12
• Subjects: Cancer -- CNN -- Google net -- VGG-16, CT -- Accuracy -- Image; CNN Convolutional Neural Network -- CT computerized tomography; Detectors -- Periodicals; Measurement -- Periodicals; 530.7
• DOI: 10.1016/j.measen.2022.100588
• Languages: English
• ISSNs: 2665-9174

Abstract

Abstract: Lung cancer is a serious disease occurring in human being. Medical treatment process mainly depends on cancer types and its location. It is possible to save many precious human lives by detecting cancer cells as early as possible. Developing an automated tool is essential to detecting malignant states at the earliest possible stage. The accuracy of prediction has always been a challenge, despite the many algorithms proposed in the past by many researchers. Using artificial neural networks, this study proposes a methodology to detect abnormal lung tissue growth. In order to achieve great accuracy, a tool with a higher probability of detection is taken into account. The manual interpretation of results is incapable of avoiding misdiagnoses. During the course of this research, lung images from both healthy and malignant individuals were analyzed. Data bases have also been developed for the various views of the CT scanning system, such as axial, coronal, and sagittal. A neural network, based on the textural characteristics of the images, makes it feasible for classification of the normal images, identifying away from the malignant ones. In order to overcome this problem, CNN and Google Net deep learning algorithms have been proposed to detect Cancer. Both the region proposal network and the classifier network use the VGG-16 architecture as their base layer. The algorithm achieves a precision of 98% in detection and classification. Based on confusion matrix computation and classification accuracy results, a quantitative analysis of the proposed network has been conducted.