Bileaflet Mechanical Heart Valves Thrombosis: In vitro Detection by Artificial Neural Networks

Endothelial pannus and thrombus formation which usually deposit on bileaflet mechanical heart valves progressively lead to incomplete opening of the valve. At the final state, they block the hinge-leaflet mechanism and cause embolic events. As a consequence, hemodynamic performance of bileaflet valvular prostheses can be severely reduced. The early detection of thrombotic formations is then crucial for correct diagnosis. The present study analyzes the power spectra calculated from the phonocardiographic signals corresponding to prostheses’ sounds as acquired in vitro, in order to check for the presence of differently shaped thrombotic deposits and to differentiate deposits in classes. Data were acquired during simulations in the aortic position with the Sheffield Pulse Duplicator. Different hydrodynamic working conditions were investigated, changing the pulse rate and the stroke volume. Thrombotic deposits of different weight and shape were placed on the valve leaflet and onto the annular housing, including the case of a thrombus completely blocking one leaflet. Power spectra were classified by an artificial neural network, specifically designed for this purpose. Thrombotic event classification was applied to five commercially available mechanical prostheses. The results obtained allow implementation of a diagnostic tool for the early detection of thrombotic deposit formation: it will result in an appropriate calibration of the anticoagulant therapy, preventing mechanical heart valve dysfunctions and thromboembolic complications.