Systolic and diastolic blood pressure estimation during exercise stress test using GK-MARS fuzzy function approach

Exercise test is a screening test which is used to investigate and diagnosis cardiovascular disease. Measuring Blood Pressure (BP) during exercise test is an important issue. Because an individual experiences a great amount of physical stress and the vital signs should be carefully monitored. Using cuff and barometer leads to low accuracy of measuring BP. Moreover, vessel crush encounters the continuous measurements with crucial limitations. Hence, measuring systolic and diastolic BP values with desired accuracy in a cuff-less approach is of challenges in this context. This could be achieved by using features extracted from ECG and PPG signals. BP is highly correlated with parameters such as PTT and HR, but not necessarily linear. The correlation could be nonlinear, multimode and vague. Therefore, using intelligent approaches could be very effective in such modeling issue. In this context, using Fuzzy Function (FF) technique is proposed in this paper. Gustafson-Kessel (GK) clustering method instead of the FCM, because of considering correlation between data points, and MARS, due to its great ability in multimode and nonlinear modeling, are used to produce the antecedent and consequence of the rules respectively. GK-MARS structure, especially in cases with limited datasets as this study, could be very effective. In order to evaluation, the performance of this structure in BP estimation is compared with other intelligent methods including GRNN, fuzzy systems based on GD and RLS training methods and other structures of FF. Results indicate better performance of FF with GK-MARS structure in satisfying AAMI (SP10) standard in systolic and diastolic BP estimation of all stages.