Falcon Optimization-based region growing segmentation for intracerebral hemorrhage (ICH) detection from CT images

Intracerebral hemorrhage (ICH) segmentation from CT images remains challenging due to intensity heterogeneity, low contrast, and the dependency of traditional region-growing methods on manually selected seed points and fixed thresholds. Existing deep learning and hybrid optimization approaches, although effective, often require large, annotated datasets and high computational resources, limiting their applicability in real-time clinical environments. To address this gap, this study proposes a Falcon Optimization Algorithm (FOA)-guided Region Growing segmentation framework that enables adaptive and automated selection of optimal seed points and similarity thresholds. The novelty of the approach lies in leveraging the exploration–exploitation balance of FOA to dynamically guide region expansion, thereby improving boundary delineation without increasing computational complexity. The proposed method is evaluated on the PhysioNet CT-ICH dataset and compared against conventional techniques including Otsu Thresholding, K-Means Clustering, Watershed Segmentation, and Morphological Active Contour models. Experimental results demonstrate that the proposed method achieves a Dice coefficient of 0.91 and Jaccard Index of 0.85, outperforming the best competing method (Morphological Active Contour) by approximately 13 and 18%, respectively. The method also attains superior sensitivity (0.93) and pixel accuracy (92.4%) while maintaining low computational cost with CPU-level execution. These results confirm that the proposed FOA-guided framework provides a robust, interpretable, and computationally efficient solution for accurate hemorrhage segmentation in clinical CT imaging.