Multiscale Registration of Real-Time and Prior MRI Data for Image-Guided Cardiac Interventions

Abstract

Recently, there is a growing interest in using magnetic resonance imaging (MRI) to guide interventional procedures due to its excellent soft tissue contrast and lack of ionizing radiation compared to traditional radiographic guidance. One of these applications is the use of MRI to guide radio frequency ablation of anatomic substrates, within the left ventricle, responsible for ventricular tachycardia. However, different MRI acquisition schemes have significant tradeoffs between image quality and acquisition time. Guidance using high-quality preoperative 3-D MR images is limited in the case of cardiac interventions because the heart moves dynamically during the procedure. On the other hand, 2-D real-time MR images acquired during the intervention sacrifice image quality for shorter image acquisition time, leading to real-time positional updates of cardiac anatomy. Ideally, we wish to combine the advantages of live feedback from real-time images and accurate visualization of anatomical structures from preoperative images. Therefore, to improve the MRI guidance capabilities for cardiac interventions, we describe a novel multiscale rigid registration framework to correct for respiratory motion between the prior and real-time datasets. In the proposed approach, we use a weighted total variation flow algorithm to extract coarse-to-fine features from the input images and subsequently register the corresponding scales in a hierarchical manner. Registration experiments were performed with in vivo human imaging data, and the target registration error achieved was 1.51 mm. Thus, the feasibility of motion correction in an interventional setting has been demonstrated, which may lead to significant improvements in the guidance of cardiac interventions.

Slice-to-volume medical image registration: A survey

“…Po Su et al (2013) applied SSD on both image intensities and gradients, combining them in a unique similarity measure. In (Xu et al, 2014b), the matching criterion was defined using the normalized gradient field of the images, while CC of both intensity and gradient magnitude was adopted in (Xu et al, 2008). In (Lin et al, 2013), a merit function called Weighted Edge-Matching Score (WEMS) was used to evaluate the similarity of a 2D real-time image and the corresponding reformated slice to search for the 3D pose of a bone model.…”

“…The standard way to estimate 6-DOF rigid transformations, consists in minimizing an energy functional (based on an iconic or geometric matching criterion) often with a continuous optimization algorithm (see section 4.1) where the search space is part of the Euclidean group SE(3) of rigid transformations. Sensors as described in section 2.3 can be used as well for rigid alignment as proposed in a number of papers (Chandler et al, 2008;Elen et al, 2010;Fei et al, 2002Fei et al, , 2003aFrühwald et al, 2009;Fuerst et al, 2014;Gholipour et al, 2010;Huang et al, 2009;Jiang et al, 2007a;Kainz et al, 2015;Kim et al, 1999Kim et al, , 2010aNir and Tannenbaum, 2011;Penney et al, 2004a;Rousseau et al, 2006;Smolíková Wachowiak et al, 2005;Xu et al, 2014b;Yeo et al, 2004;Yu et al, 2011;Chen et al, 2016).…”

Real‐time MRI guidance of cardiac interventions

“…For this, a weighted total variation flow algorithm was used to extract coarse-to-fine features from the input images and subsequently register the corresponding scale images in a hierarchical manner. 87 The registration method yielded a corresponding model that was used to characterize the respiratory motion during an MRI-guided ablation study. 88 Further, accelerating the registration algorithm with a GPU-based implementation facilitated updating the respiratory motion model every heart beat to account for potential intraprocedural changes.…”

Self‐prior image‐guided MRI reconstruction with dictionary learning

“…also demonstrated that existing MRI scans can serve as prior images to improve the quality of CS reconstruction. It has been shown that prior images can promote both the contrast and signal‐to‐noise ratio (SNR) of the reconstruction results . In addition, Chen et al .…”

“…It has been shown that prior images can promote both the contrast and signal-to-noise ratio (SNR) of the reconstruction results. 41 In addition, Chen et al [42][43][44] developed the prior image constrained compressed sensing (PICCS) method for dynamic CT imaging with few views of projections. The subtraction of the target image with the prior image is sparser than the target image, so PICCS utilizes the sparsity of the subtracted image to improve the reconstruction quality.…”