Recognition of Lung Nodules in Computerized Tomography Lung Images using a Hybrid Method with Class Imbalance Reduction

Lung cancer is among the deadly diseases affecting millions globally every year. Physicians' and radiologists' manual detection of lung nodules has low efficiency due to variety shapes nodule locations. The paper aims recognize in computerized tomography (CT) images utilizing a hybrid method reduce problem space at step. First, suggested uses fast robust fuzzy c-means clustering (FRFCM) algorithm segment CT extract two lungs, followed by convolutional neural network (CNN) identify sick for use next Then, adaptive thresholding detects suspected regions interest (ROIs) all available objects lung. Next, some statistical features are selected from ROI, then restricted Boltzmann machine (RBM) considered feature extractor that extracts rich features. After that, an artificial (ANN) employed classify ROIs determine whether ROI includes or non-nodules. Finally, cancerous localized Otsu algorithm. Naturally, more than healthy ones, leading class imbalance substantially decreases ANN ability. To solve this problem, reinforcement learning (RL)-based used, which states sampled. agent receives larger reward/penalty correct/incorrect classification examples related minority class. proposed model compared with state-of-the-art methods on image database consortium collection (LIDC-IDRI) dataset standard performance metrics. results experiments demonstrate outperforms its rivals.