Evaluation of a semi-automatic algorithm for tracking tricuspid valve annulus on magnetic resonance images

Despite cardiac magnetic resonance (CMR) is the reference imaging technique for the evaluation of ventricular function, its use for the assessment of tricuspid valve is still minimal. Tricuspid annulus (TA) evaluation in 3D is feasible from CMR when performed in rotational long-axis. However, this process is based on manual identification of TA reference points. Our aim was to test the accuracy of semi-automated algorithm, based on normalized cross-correlation, in order to reduce the cumbersome manual identification frame-by-frame. Analysis was performed on 10 healthy volunteers, and an expert reader visually inspected and classified as correct or not the TA points position. The proposed algorithm, showed accuracy equal to 87%, demonstrating its applicability to obtain the dynamic 3D morphology of TA, with potential benefits in growing the understanding of tricuspid valve, and therefore in patient.