An efficient lung nodule detection model using multiscale hybrid 2D-3D dilation assisted adaptive Trans-Res-Unet++ with hybrid heuristic mechanism

Abstract

As lung nodules are only more widely detectable once they have moved to other lung sections, it is highly challenging to anticipate the incidence of lung cancer at the beginning stages. Even before radiologists and other specialized medical professionals determine whether a lung nodule exists or not, the danger increases as time increases. The demand for Computer-Aided Detection (CAD) systems to accurately classify lung nodules into malignant or benign types has arisen as a result of improvements in medical imaging techniques like Computed Tomography (CT) scans, as this prevents delays in lung nodule diagnosis. In order to identify the lung nodule and deliver better and more prompt timely therapy, a new deep learning-based Lung Nodule Detection Model (LNDM) is created in this paper. Lung nodule detection and image collection make up the ideal functions of the created model. The lung nodule images are initially gathered from the common standard datasets. Then, the acquired images are used with Multiscale Hybrid 2D-3D Dilation assisted Adaptive Trans-Res-UNet++ (MHD-ATRes-UNet++) approaches for lung nodule diagnosis. This method combines ResNet and UNet++ with a Multiscale mechanism, a 2D-3D dilation layer, and an adaptive transformer layer. The Golden Eagle Optimizer (GEO) and the Levy Flight Distribution Algorithm (LFD) are combined to generate a Population-based Levy Flight Distribution-Golden Eagle Optimization (PLFD-GEO) algorithm, which is used in this instance to detect lung nodules. Eventually, the constructed network produces lung nodule detection results. Using a variety of performance indicators, the created lung nodule detection model's effectiveness is examined.