Automatic correction of motion artifacts in 4D left ventricle model reconstructed from MRI

This paper describes a computer method to correct the shape of three-dimensional (3D) left ventricle (LV) models created from magnetic resonance imaging (MRI) data that is affected by patient motion during scanning. Three-dimensional meshes of the LV endocardial and epicardial surfaces are created from border-delineated MRI data at every time frame of the cardiac cycle to generate a time-series model of the heart. A geometrically-based approach is used to achieve smooth epicardial shapes by iterative in-plane translation of vertices in the LV model. The Principal Curvatures of the LV epicardial surfaces across multiple time frames are used to construct a shape-based optimization objective function to restore the shape of the LV via a dual-resolution semi-rigid deformation process and a free-form geometric deformation process. A limited memory quasi-Newton algorithm, L-BFGS-B, is then used to solve the optimization problem. We tested our algorithm on 9 patient-specific models and it was able to correct motion artifacts without altering the general shape of the heart, such as its asymmetrical shape. The magnitudes of in-plane translations (Δx = 0.972±0.857 mm and Δy = 1.306±1.290 mm in the x- and y-directions, respectively) are also within the range of published experimental findings. The average computational time to correct each 4D model is 6 min 34 s.