Trophallactic optimization algorithm with markov random field refinement for stroke lesion segmentation

Abstract

Cerebrovascular accidents (strokes) represent a critical medical emergency re quiring rapid and accurate diagnosis. Automated segmentation of stroke lesions from computed tomography (CT) images remains challenging due to low con trast, image noise, and high anatomical variability between ischemic and hem orrhagic subtypes. This paper introduces a novel hybrid approach combining the trophallactic optimization algorithm (TOA), inspired by cooperative nectar exchange in bee colonies, with markov random fields (MRF) for spatial coher ence modeling. The proposed TOA-MRF method operates semi-automatically from a single user-defined seed point, leveraging bio-inspired collective intel ligence to progressively explore and refine regions of interest. The algorithm simulates the enzymatic transformation of nectar into honey through iterative information exchange between virtual bees, followed by MRF-based regulariza tion to ensure anatomical consistency. Evaluated on a clinical CT dataset from [Hospital Name], the method achieves a Dice similarity coefficient of 87.3% for ischemic strokes and 91.2% for hemorrhagic strokes, with an overall detection accuracy exceeding 89%. Comparative analysis demonstrates the complemen tary strengths of TOA exploration and MRF refinement, offering a robust and efficient solution for clinical stroke assessment with minimal user intervention.