MRF-Based Deformable Registration and Ventilation Estimation of Lung CT

Abstract

Deformable image registration is an important tool in medical image analysis. In the case of lung computed tomography (CT) registration there are three major challenges: large motion of small features, sliding motions between organs, and changing image contrast due to compression. Recently, Markov random field (MRF)-based discrete optimization strategies have been proposed to overcome problems involved with continuous optimization for registration, in particular its susceptibility to local minima. However, to date the simplifications made to obtain tractable computational complexity reduced the registration accuracy. We address these challenges and preserve the potentially higher quality of discrete approaches with three novel contributions. First, we use an image-derived minimum spanning tree as a simplified graph structure, which copes well with the complex sliding motion and allows us to find the global optimum very efficiently. Second, a stochastic sampling approach for the similarity cost between images is introduced within a symmetric, diffeomorphic B-spline transformation model with diffusion regularization. The complexity is reduced by orders of magnitude and enables the minimization of much larger label spaces. In addition to the geometric transform labels, hyper-labels are introduced, which represent local intensity variations in this task, and allow for the direct estimation of lung ventilation. We validate the improvements in accuracy and performance on exhale-inhale CT volume pairs using a large number of expert landmarks.

Bilateral Weighted Adaptive Local Similarity Measure for Registration in Neurosurgery

“…However, [4] reformulated non-rigid registration as a discrete Markov Random Field (MRF) optimization problem, for which the similarity measure gradient is not needed. We employ the proposed measure in a related discrete optimization scheme of [6]. A grid P of B-spline transformation control points p ∈ P with positions c p is overlaid onto the reference image.…”

Comparison of breathing gated CT images generated using a 5DCT technique and a commercial clinical protocol in a porcine model

“…Registration was done using the "deeds" algorithm, which employed a multilevel B-spline transform model, a similarity metric based on local selfsimilarity, and a discrete optimization framework. 10,11 The deformable registration provided the positions of the tissues at each voxel in the reference image in each of the other 29 images. Deformation vectors to the reference scan geometry calculated by the deeds algorithm along with the corresponding breathing amplitude and rate measurements collected during each scan were used to independently calculate the per-voxel parameters for the 5D breathing motion model.…”

A hybrid optimization strategy for registering images with large local deformations and intensity variations

“…We encountered such a challenge upon registering three-dimensional pre-to post-contrast abdominal images acquired by magnetic resonance imaging (MRI) for quantifying Crohn's disease activity [1]. This registration is far from trivial as peristalsis of the bowel (or organ motion) may cause large local deformations, which makes that the registration algorithm gets trapped in a local minimum as discussed in [2]. Furthermore, the diseased regions of interest are composed of relatively thin bowel structures, which are easily mismatched.…”

“…A coarse to fine registration method can be combined with discrete optimization approaches, such as graph cuts [4] and linear programming [5]. Discrete optimization is typically less sensitive to the initial conditions [5] and suitable to deal with large deformations [2]. However, discrete opti-mization goes at the expense of the registration precision because of quantization effects.…”