Lung Cancer Detection Using Image Segmentation by means of Various Evolutionary Algorithms

Area of Science:

• Medical Imaging
• Computational Intelligence
• Oncology

Background:

• Manual interpretation of large medical images like CT scans is time-consuming and challenging for physicians.
• Accurate and fast image segmentation is crucial for effective computer-aided diagnosis, especially for early lung cancer detection.
• Existing segmentation algorithms face challenges in precision and convergence time, necessitating novel approaches.

Purpose of the Study:

• To develop and evaluate an expedient image segmentation algorithm for medical CT scans.
• To reduce the burden on physicians during the interpretation of complex radiological images.
• To enhance early lung cancer detection through improved tumor segmentation.

Main Methods:

• Comparative analysis of five optimization algorithms: k-means, k-median, particle swarm optimization (PSO), inertia-weighted PSO, and guaranteed convergence PSO (GCPSO).
• Evaluation of preprocessing techniques including median, adaptive median, and average filters, with adaptive median filter selected for optimal performance.
• Application of adaptive histogram equalization for image contrast enhancement.
• Validation using MATLAB on 20 lung CT sample images.

Main Results:

• The adaptive median filter demonstrated superior performance in the preprocessing stage for medical CT images.
• Adaptive histogram equalization effectively enhanced image contrast.
• Among the tested algorithms, GCPSO achieved the highest accuracy of 95.89% for lung tumor extraction.
• The study successfully verified the practical results on sample lung images.

Conclusions:

• GCPSO is a highly accurate and efficient algorithm for segmenting lung tumors in CT images.
• The proposed segmentation approach can significantly aid physicians in the early diagnosis of lung cancer.
• The integration of adaptive filtering and contrast enhancement further optimizes the segmentation process.