Aims: Aortic Stenosis (AS) is a heart valve disease characterized by the narrowing of the aortic valve opening. Currently AS is diagnosed using echocardiography performed by a trained specialist. The goal of this study was to evaluate the ability of non-invasive microelectromechanical based seismocardiography (SCG) and gyrocardiography (GCG) sensors to detect AS in patients by measuring the cardiac-induced vibrations produced by the mechanical activity of the heart.
Methods: Clinical data was collected from 19 AS patients at the Turku University Hospital using a custom data logger capable of measuring SCG, GCG and single-lead ECG. Additionally, data was collected from 51 healthy volunteers. The signals were retrospectively processed and analyzed. ECG measurements were used in identification of individual cardiac cycles. SCG and GCG each contain three channels of measurement representing a different axis of motion. Signals were segmented based on each cardiac cycle and a continuous wavelet transform was applied to each cycle producing a time-frequency 2-D matrix representation of each channel. The six 2-D matrices were overlaid to create images of cardiac cycles that were supplied to a convolutional neural network. Precision, recall, and accuracy of the classifier were estimated using leave-subject-out cross-validation.
Results: 47928 cardiac cycles from 70 patients were included in this study, 19 patients had been diagnosis with AS (9865 cardiac cycles) while 51 patients (38063 cardiac cycles) had no cardiovascular disease. For AS patients, the classifier provided an average accuracy of 96%, recall of 97%, and precision of 90%. For healthy patients, the classifier provided average accuracy of 97%, recall of 97%, and precision of 99%.
Conclusion: Together SCG and GCG are able to separate AS patients from healthy volunteers. We are currently collecting clinical data from age-matched controls in order to determine whether SCG and GCG has clinical potential to detect AS.