A Deep Learning Network for Classification of Lung Cancer from Computer Tomography Images Using Fine-Tuned Visual Geometric Group-16

Abstract

Many deaths from cancer are caused by lung cancer, of the most prevalent and deadly forms of the illness. Although lung cancer remains a significant health issue, improvements in research, early detection techniques, and current treatments give hope for improved outcomes. To diagnosis a wide range of diseases, numerous Computer Aided Diagnosis (CAD) systems have been created recently. Early lung cancer detection is now crucial and simple thanks to deep learning and image processing methods. For radiologists, identifying cancerous lung nodules is a difficult and time-consuming process that involves computed tomography (CT) scans. Kaggle was used to gather the lung cancer CT scans. Two essential techniques used in DL to improve the quality, diversity, and generalization capabilities of images during training are image pre-processing and augmentation. This article presents an improved Deep Learning (DL) model that effectively classifies lung cancer from CT scans. The foundation model was a modified VGG-16 model. By applying the fine-tuning process, the suggested model's efficacy is greatly enhanced. In the categorization report, metrics such as F1-score, recall, accuracy, and precision are frequently provided. After adjustments, the suggested model's accuracy rose dramatically from 81.91% to 96.78%. According to the experimental findings, the suggested approach outperformed the current methods for classifying lung cancer.