Un algoritmo híbrido adaptativo de metaheurística para el cáncer de pulmón en la segmentación de imágenes patológicas

Área de la Ciencia:

• Digital pathology
• Medical image analysis
• Computational intelligence

Sus antecedentes:

• Histopathological images are crucial for lung cancer diagnosis and subtyping.
• Automated segmentation of these images is challenging due to high resolution, color diversity, and complexity.
• Accurate segmentation is vital for reliable AI-assisted diagnostic systems.

Objetivo del estudio:

• To develop a novel hybrid metaheuristic approach for multilevel image thresholding.
• To enhance the accuracy and computational efficiency of lung cancer histopathology image segmentation.
• To address the challenges posed by complex image characteristics in automated analysis.

Principales métodos:

• An adaptive hybrid metaheuristic algorithm, SCSOWOA, was developed by integrating Sand Cat Swarm Optimization (SCSO) and Whale Optimization Algorithm (WOA).
• The SCSOWOA algorithm sequentially and adaptively combines SCSO's exploration with WOA's exploitation.
• Performance was evaluated on the LC25000 lung cancer dataset using Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), and Feature Similarity Index Measure (FSIM).

Principales resultados:

• SCSOWOA achieved high-quality segmentation with average PSNR of 27.9453 dB, SSIM of 0.8048, and FSIM of 0.8361.
• Optimal performance was observed at T=12, yielding SSIM of 0.9340 and FSIM of 0.9542.
• The algorithm demonstrated a 40% improvement in computational efficiency, with an average execution time of 1.3221 s.

Conclusiones:

• SCSOWOA effectively balances exploration and exploitation for robust image segmentation.
• The algorithm provides high accuracy, low variance, and computational efficiency in histopathology image analysis.
• SCSOWOA shows significant potential for AI-assisted lung cancer diagnosis systems.