Enhancing Lung Nodule Classification: A Novel CViEBi-CBGWO Approach with Integrated Image Preprocessing

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Accurate lung nodule classification is critical for cancer treatment and survival.
• Early-stage diagnosis of malignant lung nodules significantly improves patient outcomes.
• Existing methods face challenges in achieving high classification accuracy.

Purpose of the Study:

• To propose a novel model, CViE-CBGWO, for enhanced lung nodule classification accuracy.
• To improve the early detection and diagnosis of malignant lung nodules.
• To overcome limitations of existing classification methods.

Main Methods:

• Utilized LUNA16 and LIDC-IDRI datasets for CT image analysis.
• Applied preprocessing techniques including normalization, data augmentation, and filtering.
• Extracted local features using Convolutional Neural Network (CNN) and global features using Vision Transformer (ViT).
• Combined features and used Elman Bidirectional Long Short-Term Memory (EBiLSTM) for classification.
• Optimized model parameters using a hybrid Grey Wolf Optimization (GWO) with crossover (CBGWO).

Main Results:

• The proposed CViEBi-CBGWO model achieved a superior classification accuracy of 98.72%.
• Comparative analysis showed significant improvement over existing methods.
• Effective feature extraction and parameter optimization were demonstrated.

Conclusions:

• The CViEBi-CBGWO model offers a highly accurate solution for lung nodule classification.
• This AI-driven approach holds promise for improving early lung cancer diagnosis.
• The integration of CNN, ViT, EBiLSTM, and GWO optimization enhances diagnostic capabilities.