Attention-guided deep neural network with a multichannel architecture for lung nodule classification

• DOI: 10.1016/j.heliyon.2023.e23508
• Publication Date: 2023-12-10T15:41:15Z
• Authors: Rong Zheng; Hongqiao Wen; Feng Zhu; Weishun Lan
• Keywords: Nodule (geology); RGB color model; Computer-Aided Diagnosis

Detecting and accurately identifying malignant lung nodules in chest CT scans a timely manner is crucial for effective cancer treatment. This study introduces deep learning model featuring multi-channel attention mechanism, specifically designed the precise diagnosis of nodules. To start, we standardized voxel size images generated three RGB varying scales each nodule, viewed from different angles. Subsequently, applied submodels to extract class-specific characteristics these images. Finally, nodule features were consolidated model's final layer make ultimate predictions. Through utilization an could dynamically pinpoint exact location without need prior segmentation. proposed approach enhances accuracy efficiency classification. We evaluated tested our using dataset 1018 sourced Lung Image Database Consortium Resource Initiative (LIDC-IDRI). The experimental results demonstrate that achieved classification 90.11 %, with area under receiver operator curve (AUC) score 95.66 %. Impressively, method this high level performance while utilizing only 29.09 % time needed by mainstream model.