Robust and automatic diagnosis of the intraventricular mechanical dyssynchrony for the left ventricle in cardiac magnetic resonance images.

PURPOSE: Robust and automatic diagnosis of the mechanical dyssynchrony for the left ventricle is essential for the cardiac resynchronization therapy. However, no existing method could meet the desired accuracy yet. In this paper, a new approach is proposed for auto-diagnosis of the intraventricular mechanical dyssynchrony of the left ventricle based on a series of image processing and signal processing techniques.

METHODS: Firstly, the boundary of the left ventricle is identified automatically by the segmentation method. Secondly, the correspondence trajectories are computed based on the tangent field to make sure that they are perpendicular to all the intercepted boundaries. Thirdly, the intercepted points are smoothed by the proposed Fourier shape filter to eliminate noise and increase the diagnosis accuracy. Fourthly, the mechanical dyssynchrony is defined as the lag times between the periodic change of the sampled boundary points and the periodic change of the area of the left ventricle. It is calculated by cross-correlation.

RESULTS: The segmentation method is trained with 10 cases, the proposed approach is evaluated with other 40 tested cases (20 normal cases and 20 patient cases), and the diagnosis accuracy is 100%.

CONCLUSIONS: Experimental results showed that the proposed approach can accurately diagnose whether the left ventricle has mechanical dyssynchrony or not from the computed lag times. The proposed approach is robust in auto-diagnosis of the mechanical dyssynchrony for the left ventricle in cardiac magnetic resonance images.