Video Stabilization and Region-of-Interest Tracking in Cardiac MRI Domain

Video Stabilization and Region-of-Interest Tracking in Cardiac MRI Domain

Stabilizing Cardiac-MRI (CMRI) sequences, such as the myocardial first pass perfusion, is expected to allow a significant improvement in medical diagnosis. Such stabilization is crucially required due to the diaphragm motion, throughout the respiratory and cardiac cycles. The above challenge is also valid generally in computervision for video-stabilization and region-of-interest (ROI) tracking of non-rigid objects. We suggest a novel algorithm for CMRI tracking and stabilization, which is inspired by cortical mechanisms of the human visual system (HVS), for both edge and region pathways. The algorithm adaptively weights these pathways according to the ROI state. The ROI is tracked through a two-stage pipeline; a coarseengine first extracts a linear approximation of the motion, followed by a fine-engine, which allows edge deformation. The ROI motion is then estimated by common linear-approximation for stabilization. The Video-stabilization is obtained by solving the ROI-tracking problem, while keeping its initial position fixed. The proposed automatic algorithm was tested on several CMRI videos. The stabilization quality was assessed using tools based on Inter-Frame-Similarity (ITF) and Structural Similarity (SSIM) metrics. In addition, the results were clinically rated on a 1-5 scale by two radiologists. Both the engineering and clinical assessments were used in comparing our results with state of the art competitor methods, wherein our results were generally favored over the competitors (7 of 10 cases, 1 case is controversial, i.e. preferred clinically only). Our algorithm manages to stabilize perfusion CMRI slice for long burst of frames, which indicates the potential for allowing a better medical diagnosis.