Susin_4380_r1.qxd:Layout 1

Although, mechanical heart valves (MHVs) exhibit a durability which is much greater than that of biological substitutes, they are typically prone to thromboembolic events. Consequently, MHV recipients are generally administered with lifelong anticoagulant therapy that must be correctly regulated on the basis of frequent hematological evaluations. The development of thrombi on the valve hinge may also lead to a partial or complete blocking of the leaflets’ movement. Generally, the dynamic behavior of the valve is monitored clinically on an annual basis using echocardiography and, if necessary, also with cinefluoroscopy and/or multidetector computed tomography. In fact, more frequent monitoring could reduce the risk of thromboembolic complications by detecting any signs of valve malfunction before their acute appearance. Classification algorithms have already been developed (1) for implementing an automatic system to monitor bileaflet MHVs on the basis of valve sound evaluation. The asynchronous closure of the leaflets, which produces typical ‘double clicks’ when the prosthesis is correctly functioning, has been investigated (2,3). Moreover, the time-frequency analysis of the closing sound has been applied to identify double and single clicks by using the continuous wavelet transform (1,4). Recently, it has been established that multilayer feedforward networks, coupled with phonocardiographic analysis, can result in a more objective diagnosis of heart valve pathologies (5). Similarly, the closing sounds of biological prostheses have been analyzed and classified (6). The results of previous in-vitro studies with the Sheffield pulse duplicator (PD) have Address for correspondence: Andrea Bagno, Department of Chemical Process Engineering, University of Padova, via Marzolo 9, 35131 Padova, Italy e-mail: [email protected] Background and aim of the study: As with all mechanical prostheses, bileaflet heart valves are prone to thrombus formation, reduced hemodynamic performance, and the occurrence of embolic events. The early detection of thrombotic formations is crucial for correct diagnosis and adequate therapy. The study aim was to analyze the power spectra of the phonocardiographic signals acquired in vitro for various thrombotic deposits reproduced on a bileaflet mechanical valve, in order to monitor and classify their presence. Methods: Data were acquired for the St. Jude Medical Regent valve mounted in the aortic position of a Sheffield Pulse Duplicator. Different pulsatile flow conditions were reproduced, changing the heart rate and stroke volume. Thrombotic deposits of various weights and shapes were placed on the valve leaflet, or on the annular housing. The case of a thrombus completely blocking one leaflet was also investigated. Power spectra were calculated from the phonocardiographic signals and classified by an artificial neural network. Results: The proposed approach resulted in a 95% correct identification of all simulated thrombotic deposits. Interestingly, phonocardiographic analysis is capable of detecting the presence of different types of artificial thrombi and also to classify them, whereas transvalvular pressure values cannot provide such detection. Conclusion: An effective diagnostic tool capable of detecting valvular thrombosis at the early stages of formation may help clinicians to formulate valvular dysfunction diagnoses before the appearance of critical symptoms. The ability to transfer results obtained in vitro to actual clinical situations might represent a significant advance in the follow up of patients.