Date of Publication :20th March 2024

Abstract: The prevalence of lung cancer has increased significantly, attributed to lifestyle modifications, alcohol consumption, hereditary factors, and exposure to environmentally-dispersed hazardous gases. The disease is affected by genetic abnormalities and exposures from industries like construction, manufacturing, and mining. The identified research challenges highlight limitations in current approaches, emphasizing the importance of developing an early detection framework, reducing false-positive rates, and ensuring accurate classification. However, there are concerns regarding convolutional neural networks (CNNs) in the proposed model, including increased processing latency, particularly with deep processing needed for high-resolution data. Efforts to mitigate this issue involve minimizing GPU usage when handling high-definition images. The proposed approach integrates transfer learning, employing the ensemble concept with the VGG16 network known as TLA-VGGN. The system's performance is evaluated based on accuracy and error rate, achieving 99.02% accuracy with a validation loss of 0.920. The training utilized a dataset of 2901 images across four classes, successfully identifying 568 images representing different cancer types.

Reference :