Deep Learning Aided Dynamic Clustering for White Blood Cell Segmentation Using Enhanced Fuzzy C-Means Optimization

Segmentation of white blood cells represents a critical component of medical image analysis, indispensable for diagnosing conditions such as infections, leukemia, and immune-related pathologies. Nonetheless, precise segmentation poses significant challenges attributable to cell overlaps, image noise, low contrast, and irregular boundaries. In this paper, we introduce a hybrid frame-work that integrates deep learning with enhanced fuzzy C-means clustering to augment segmentation efficacy. Deep learning enables robust feature extraction, while the enhanced FCM employs refined distance metrics for superior clustering refinement. The framework was trained and validated on the Kaggle Blood Cell Images dataset, with performance evaluated via accuracy, precision, recall, and F1-score metrics. Comparative evaluations encompassed baseline machine learning approaches, conventional FCM, standalone deep learning, and the proposed hybrid method. Empirical outcomes indicate that the hybrid model markedly enhances segmentation accuracy and robustness relative to standalone techniques. It effectively mitigates noise interference and expedites cluster convergence, positioning it as viable for real-time clinical deployment. Further-more, this integration addresses inherent limitations of traditional FCM, such as sensitivity to initial conditions and susceptibility to local optima, through the application of advanced optimization techniques.