Lung cancer remains one of the deadliest cancers worldwide, largely due to late-stage diagnosis and the complex, often asymptomatic progression of the disease. The study presents a new noise-aware Computer-Aided Diagnosis (CAD) framework for lung cancer detection in CT scans, addressing the critical challenge of image noise that can obscure vital diagnostic details. Thus, the proposed work uses a multilayer perceptron-based classifier that uses texture descriptors from the Gray Level Co-Occurrence Matrix (GLCM) and Local Binary Pattern (LBP), integrates Inception V3 for feature extraction, and introduces a high-level adaptive Gaussian filter with Multi-spatial Channel Attention (MSCA) convolutional segmentation. Finally, classification was achieved via a multilayer perceptron (MLP) using a novel Adaptive Osprey Optimization Algorithm (AOOA). This architecture enables effective feature learning, segmentation, and classification through modular integration of CNN, attention, and statistical texture extraction. The experimental results on the IQ-OTH/NCCD dataset show a classification accuracy (0.9894), specificity (0.9917), sensitivity (0.9846), and AUC metrics. This framework holds strong potential for real-world clinical integration, offering improved early diagnosis and supporting radiologists in lung cancer assessment.
Pahlevani et al. (Fri,) studied this question.