Adaptive Handcrafted Features Convolutional Neural Network for Lung Cancer Detection Using CT Images

Abstract

In this study, an Adaptive Handcrafted Features Convolutional Neural Network (AHFCNN) for lung cancer detection from CT images is developed. The database is first built using data from web resources. After that, a pre-processing step is created to purge the photos of undesirable information. The pre-processing procedure is also taken into account while enhancing the photos. Because they work on various images with excellence and simplicity, the pixel intensity assessment and histogram techniques are used to improve the image quality. The Grey level cooccurrence matrix (GLCM) and local binary pattern (LBP) are then used to extract the necessary features. Finally, the CT image of the lung cancer is classified using the retrieved features. Convolutional neural networks (CNN) and hybrid meta-heuristic approaches (HMHA) are combined in the suggested classifier. The HMHA was applied to the CNN to select the best gain values. The Coati Optimisation Algorithm (COA) and Honey Badger Optimisation (HBO) are combined to create the HMHA. The HBO was used in the COA to improve the coatis' updating procedure. The proposed methodology was put into practise in Python, and its effectiveness was assessed by taking into account performance indicators including sensitivity, specificity, recall, sensitivity, and F-Score. Recurrent Neural Network- Whale Optimisation Algorithm (RNN-WOA), Deep Belief Neural Network- Remora Optimisation Algorithm (DBNN-ROA), and CNN- Grey Wolf Optimisation (CNN-GWO) are traditional methodologies that are compared to the proposed methodology.