Application of Several Transfer Learning Approach for Early Classification of Lung Cancer

Abstract


 INTRODUCTION: Lung cancer, a fatal disease characterized by abnormal cell growth, ranks as the second most lethal worldwide, as observed in recent research conducted in India and other regions. Early detection is crucial for effective treatment, and manual differentiation of nodule types in CT images poses challenges for radiologists.
 OBJECTIVES: To enhance accuracy and efficiency, deep learning algorithms are proposed for early lung cancer detection. Transfer learning-based computer recognition algorithms have shown promise in providing radiologists with additional insights.
 METHODS: The dataset used in this study comprises 1000 CT scan images representing lung large cell carcinoma, lung adenocarcinoma, lung squamous cell carcinoma, and normal lung cases. A preprocessing phase, including picture rescaling and modification, is applied to the input CT scan images of the lungs, followed by the utilization of a specific transfer learning model to develop a lung cancer detection system.
 RESULTS: The performance of various transfer learning strategies is evaluated using measures such as accuracy, precision, recall, specificity, area under the curve, and F1-score.
 CONCLUSION: Comparative analysis indicates that VGG16 outperforms other models in accurately categorizing different types of lung cancer.