Assessment of Stroke Risk Using Soft Computing

Stroke risk assessment is complex in biomedical engineering due to the interaction of multiple clinical and lifestyle-related risk factors, which may not be consistently or comprehensively evaluated in routine clinical practice. To address this challenge, this study proposes a non-learning soft computing framework based on fuzzy cognitive mapping for stroke risk assessment. The proposed model integrates expert knowledge from three neurologists to construct the fuzzy cognitive map and assigns individual risk levels into low, moderate, and high categories. Model performance was evaluated using 10-fold cross-validation on a dataset of 110 individuals and benchmarked against the fuzzy c-means clustering algorithm and logistic regression. Experimental results demonstrate that the proposed FCM-based system outperforms the comparative methods, achieving an overall classification accuracy of 90.7%. These findings indicate that the proposed approach provides an interpretable and effective decision-support tool for stroke risk assessment.