A Bio-Inspired Antigen-Antibody Optimization Approach for MRI-Based Stroke Lesion Segmentation
S Sylvia Irish, J Dhalia Sweetlin, P Horsley SolomonObjective:
Ischemic strokes constitute a significant global health issue, contributing to high death rates and chronic disability, which creates an urgent need for precise and efficient detection methods. Accurate segmentation and classification of ischemic stroke lesions from magnetic resonance imaging (MRI) remains a challenging task due to lesion heterogeneity, intensity variations, and limited annotated datasets. This study investigates the feasibility of integrating an antigen-antibody-inspired optimization strategy into a lesion segmentation framework for ischemic stroke MRI analysis.
Methods:
The proposed proof-of-concept framework combines Cellular Automata (CA)-based brain masking, Simple Linear Iterative Clustering (SLIC) based superpixel segmentation, and an antigen-antibody-inspired optimization mechanism for segmentation refinement. The optimized lesion regions are subsequently analyzed using a ResNet-50 deep learning model for feature extraction and classification. MRI data from the ISLES-2015 dataset were used for methodological evaluation under heterogeneous lesion conditions.
Results:
The proposed framework achieved a classification accuracy of 93.09%, pixel accuracy of 96.96%, precision of 0.723, recall of 0.76, F1-score of 0.71, Intersection over Union (IoU) of 0.56, and Dice coefficient of 0.71.
Discussion:
The antigen-antibody-inspired optimization stage has contributed to improved lesion-region refinement compared with baseline segmentation output, thereby improving the classification results. These findings indicate that biologically inspired optimization can effectively complement conventional image segmentation techniques in handling ischemic stroke lesions.
Conclusion:
The proposed framework effectively utilizes biologically inspired optimization strategies for segmentation refinement in ischemic stroke MRI analysis. However, the present work should be interpreted as an exploratory methodological proof-of-concept rather than a clinically validated diagnostic framework. Further evaluation using clinically representative datasets, external validation cohorts, and multicenter studies is required to establish robustness and clinical applicability.