Cell Segmentation Based on FOPSO Combined With Shape Information Improved Intuitionistic FCM

Area of Science:

• Biomedical image analysis
• Computational biology
• Computer vision

Background:

• Fuzzy c-means (FCM) clustering is effective for image segmentation but struggles with noise, initialization sensitivity, and complex cell images.
• Existing FCM methods fail to accurately segment cell images with non-uniform gray values and touching cells.

Purpose of the Study:

• To develop an advanced cell image segmentation algorithm overcoming limitations of traditional Fuzzy c-means (FCM).
• To improve segmentation accuracy for challenging cell images, including those with inhomogeneous gray values and complex touching cells.

Main Methods:

• A novel cell image segmentation algorithm combining fractional-order velocity based particle swarm optimization (FOPSO) with shape information improved intuitionistic FCM (SI-IFCM) clustering.
• Iterative optimization process involving FOPSO and SI-IFCM for robust cell image segmentation.

Main Results:

• The proposed FOPSO-SI-IFCM algorithm demonstrated superior performance in cell image segmentation.
• Achieved the highest recall rate of 90.25% and the lowest false discovery rate of 0.28% compared to state-of-the-art methods.

Conclusions:

• The integrated FOPSO and SI-IFCM approach significantly enhances cell image segmentation accuracy.
• This method offers a robust solution for segmenting complex cell images, outperforming existing algorithms.