A Multi-Feature Fuzzy c-Means Framework for Enhanced Microarray Image Segmentation

Abstract

Induction Accurate segmentation of microarray images is critical for reliable gene expression analysis, yet conventional methods often struggle with challenges such as noise, irregular spot morphology, and overlapping signals. This study presents a Multi-Feature Fuzzy c-Means (MF-FCM) framework that integrates intensity, texture, and spatial proximity features to enhance the clustering and delineation of microarray spots. By leveraging fuzzy membership functions, the proposed framework effectively addresses uncertainty and partial pixel memberships, reducing misclassification at spot boundaries. The MF-FCM algorithm is implemented with optimized feature weighting and iterative centroid refinement, enabling robust performance across diverse image conditions. Experimental evaluation demonstrates significant improvements in segmentation accuracy, precision, and recall compared to traditional FCM, single-feature clustering, and evolutionary algorithm-based approaches. The results indicate that MF-FCM not only mitigates the effects of noise and spot variability but also provides computational efficiency suitable for high-throughput microarray analysis. This research contributes a theoretically grounded, practically implementable methodology that advances automated microarray image segmentation, improving the fidelity of downstream genomic analyses and offering potential applications in biomedical research, diagnostics, and systems biology.