DETECTION of HFrEF and HFpEF USING PPG-DERIVED HRV with MACHINE LEARNING METHODS

Authors : Pınar ÖZEN KAVAS, Mehmet Recep BOZKURT

Pages : 317-329

View : 38 | Download : 9

Publication Date : 2022-12-31

Article Type : Research Paper

Abstract :Regarding heart failure insert ignore into journalissuearticles values(HF);, reducing mortality and prolonging life is one of the main treatment goals. Many clinical studies define HF patients according to Left Ventricular Ejection Fraction insert ignore into journalissuearticles values(LVEF);. Two different subtypes in patients with HF are: HF with preserved ejection fraction insert ignore into journalissuearticles values(HFpEF); and HF with reduced ejection fraction insert ignore into journalissuearticles values(HFrEF);. Echocardiography is generally used to measure LVEF. This is a pricy device that requires an expert and there may be situations where attaining the device is restricted. There may be cases that treatment should be started without echocardiography. Economical and practical measurement and decision support systems are needed to solve such situations. In this study, an algorithm was improved to detect HFrEF and HFpEF by using solely heart rate variability insert ignore into journalissuearticles values(HRV); derived from photoplethysmography insert ignore into journalissuearticles values(PPG);. PPG data were obtained from volunteers for 10s, digital filters were used to clean PPGs, and HRV derivation was made from cleaned PPG. Totally thirty-seven features were obtained. Consequently, features were selected, and classification that was realized with only 3 features extracted from HRV gave significant results. 10-fold cross validation was performed for evaluation. The classification performance parameters were: accuracy %98.33, sensitivity 0.967, specificity 1, AUC 0.983, F-measure 0.981 and Kappa 0.967. This study provided highly reliable non-random results for distinguishing between HFrEF and HFpEF. This system, which works with such high performance with traditional machine learning methods used in real-time systems, makes a significant contribution to the literature in terms of diagnosing HFrEF and HFpEF cases with a single signal.
Keywords : Machine Learning, Artificial Intelligence, PPG derived HRV, HFrEF, HFpEF, Classification

ORIGINAL ARTICLE URL