Polynomial Characterization Clustering Method for Bayès Syndrome Detection

Abstract

Several studies have established a strong association between Bayès Syndrome and multiple cardiovascular and neurological conditions. As it represents a potential risk for patients, its detection at an early stage is considered relevant. The morphology of the electrocardiogram (EKG) P wave is related to the detection of this syndrome. There are numerous works that detect and characterize the P wave. However, they are grounded in pure mathematical techniques. While effective, such methods often entail substantial costs in terms of computational time, particularly for screening applications in big data contexts. Our approach combines the robustness of wave characterization through the use of polynomial regression to identify relevant aspects of morphology. Clustering methods have been analyzed to classify the P wave: K-Means++ and fast autonomous unsupervised multidimensional (FAUM) using the polynomial coefficients as features. The obtained clustering and classification metrics for 49 P-waves show that the F1 Score = 1.00 with k = 3 and polynomial degree = 3. On the other hand, FAUM significantly enhances temporal efficiency compared to other implementations of K-Means++, making it particularly suitable for applications involving large sample sizes.