Multimodal Image Fusion for Cardiac Resynchronization Therapy Planning

Cardiac resynchronization therapy (CRT) has shown its efficiency to treat patients with left-sided heart failure, however with 30% of them not responding to the therapy. One way to optimize CRT is to pre-operatively plan the implantation of the CRT device and especially the positioning of the stimulation lead pacing the left ventricle (LV), which is implanted through the coronary veins. Indeed, it has been shown that this lead should target LV sites with a late mechanical activation and without fibrosis. Additional imaging modalities should therefore be part of CRT’s planning, in order to describe the anatomy, mechanical activation, and tissue characteristics of the LV. We developed a full workflow to process, register, and fuse CT images, ultrasound (US) images, and MRI, including cine-MRI and late gadolinium enhancement (LGE) MRI. It results in a 3D patient-specific model, describing the anatomy of the LV and of the coronary veins, the electro-mechanical delays, and the presence of fibrosis. The process includes a semi-automatic segmentation of CT images to extract the LV cavity and the veins. 2D US images are processed using speckle tracking echography (STE) to estimate the mechanical strains. LGE-MRI is segmented to extract macroscopic fibrosis. All these images are registered using CT as the anatomical reference. Registration methods have thus been developed to register STE to CT, LGE to cine-MRI, and cine-MRI to CT. This whole process furnishes to the physician, before the CRT implantation, a patient-specific 3D model representing all the information needed to select the most appropriate LV pacing sites. Results obtained on patients undergoing CRT are presented.