An Optimized Transformative Approach To Detect Lung Cancer Based On Global & Local Feature Elements

Abstract

This paper studies deep learning focused lung cancer detection using the IQ-OTHNCCD lung cancer database. This study aims to improve the classification accuracy for patients with severe and non-severe cancers by Includes additional datasets Pre-trained model and complex statistical methods The data is pre-processed for many tasks. This includes scaling, style, and geometric parameters such as color intensity, aspect ratio, and size distribution of image objects. The frame evolution model was built using ResNet50 as a basis, with convolution and branch transformation combined with a new multivariate algorithm. These ecosystems combine global and local elements to effectively maintain diversity. Advanced techniques such as L2 preparation at the teaching and learning level. and planning the teaching rate It is used to reduce excessive consistency and improve performance. Test results showed high discrimination accuracy. The approach got an F1 score of 0.99, demonstrating that it can reliably detect lung cancer. Confusion matrix visualization ROC curve analysis and classification measures also confirmed the robustness of the model. This study provides to the application of deep learning in complex medical imaging tasks. This has contributed to advances in automatic cancer diagnosis.