Document Type


Publication Date


Publication Title

AMA-IEEE Medical Technology Conference on Individualized Healthcare


Atrial Fibrillation (AF) is a significant clinical problem and the complications of cardiovascular postoperative AF often lead to longer hospital stays and higher heath care costs. The literature showed that AF may be preceded by changes in electrocardiogram (ECG) characteristics such as premature atrial activity, heart rate variability (HRV), and P-wave morphology. We hypothesize that the limitations of statistics-based attempts to predict AF occurrence may be overcome using a hybrid neuro-fuzzy prediction model that is better capable of uncovering complex, non-linear interactions between ECG parameters. We created a neuro-fuzzy network that was able to classify the patients into the control and AF groups with the performances: 99.42% sensitivity, 99.89% specificity, and 99.74% accuracy for 30 minutes just before AF onset.


Also presented at Cleveland Clinic Research Day, October, 2010.


This work was supported in part by the Grant 10CRP2600305 AHA.

Original Citation

M. Ovreiu, M. Petre, D. Simon, D. Sessler, and C. Bashour, “Classification of Atrial Fibrillation prone Patients using Electrocardiographic Parameters in Neuro-Fuzzy Modeling,” AMA-IEEE Medical Technology Conference on Individualized Healthcare, Washington, DC, March 2010


Publisher's PDF