A Neural Network and Optimization Based Lung Cancer Detection System in CT Images

Area of Science:

• Medical Imaging
• Computational Biology
• Oncology

Background:

• Lung cancer is a leading cause of cancer-related mortality in both men and women.
• Early detection of lung nodules through effective screening is crucial for timely treatment and improved patient survival rates.
• Existing methods for lung nodule segmentation and classification often lack the required accuracy and reliability.

Purpose of the Study:

• To propose an effective lung cancer screening procedure to rapidly identify lung cancer lesions and increase diagnostic accuracy.
• To develop a robust framework for accurate segmentation and classification of lung nodules.
• To enhance the reliability of lung cancer detection systems.

Main Methods:

• Utilized Otsu thresholding for precise segmentation of lung areas of interest.
• Employed the cuckoo search algorithm to identify optimal features for nodule characterization.
• Extracted relevant lesion features using a local binary pattern descriptor.
• Designed a Convolutional Neural Network (CNN) classifier for malicious vs. non-malicious lung lesion identification.

Main Results:

• The proposed framework achieved a high accuracy of 96.97% in classifying lung lesions.
• The integration of Otsu thresholding, cuckoo search, and LBP features improved nodule detection and characterization.
• Results demonstrated enhanced accuracy compared to traditional methods.

Conclusions:

• The developed screening procedure effectively identifies lung cancer lesions with improved accuracy.
• The combination of segmentation, feature extraction, and CNN classification offers a reliable approach for lung nodule analysis.
• Further validation using Particle Swarm Optimization and genetic algorithms supports the enhanced performance.