Novel Lung Nodule Detection Strategy using Hybrid Optimization tuned Deep Convolutional Neural Network

Abstract

newline Abstract newlineLung cancer acts as a universal wellness concern ascribed by its lower survival rate; hence, early newlinedetection is significant for improving the life span of the cancer patient. Despite the advancement newlinein medical imaging, detecting lung nodules (LN) more precisely using Computed Tomography newline(CT) scans is crucial owing to variations in image intensity and noise. This research presents two newlinehybrid optimization approaches in the direction of efficient detection and LNs classification using newlineDeep Learning (DL) approaches. The first methodology employs an Elephant-Based Bald Eagle newlineOptimization (EBEO) algorithm for the segmentation process and classification of LNs, by using newlinea Deep Convolutional Neural Network (DCNN). The EBEO algorithm, stimulated by elephant newlineherding behavior, enhances segmentation accuracy and tunes the DCNN s hyperparameters, newlineimproving detection performance. The optimization algorithm employed in the second newlinemethodology is Eyie Flock, where flocking characteristics of birds and tuskers are combined to newlineoptimize segmentation and classification tasks. Hence, this method accelerates convergence and newlineenhances the LN s precision of detection. The evaluation results show that the comparative analysis newlineof the proposed model facilitates superior performance against other existing models in terms of newlinesegmentation and classification of LN. newlineThe DCNN model utilizes Eyrie Flock optimization technique to regulate the parameters, such as newlineweights and bias, to gain accurate classification. The Eyrie Flock-DCNN model gained an accuracy newlineof 92.62%, sensitivity of 93.52%, and specificity of 97.19% in terms of training percentage, where newlinethe model correctly predicted LN and reduced false predictions. Likewise, the Eyrie Flock-DCNN newlinemodel outperformed all other models with an accuracy of 94.12%, sensitivity of 95.21%, and newlinespecificity of 98.96% in terms of K-fold. This, Eyrie Flock-DCNN s consistently higher specificity, newlineindicates its superior ability to classify true negatives correctly, t