Manifold Learning for Image-Based Breathing Gating with Application to 4D Ultrasound

Abstract

Abstract. Breathing motion leads to a significant displacement and deformation of organs in the abdominal region. This makes the detection of the breathing phase for numerous applications necessary. We propose a new, purely image-based respiratory gating method for ultrasound. Further, we use this technique to provide a solution for breathing affected 4D ultrasound acquisitions with a wobbler probe. We achieve the gating with Laplacian eigenmaps, a manifold learning technique, to determine the low-dimensional manifold embedded in the high-dimensional image space. Since Laplacian eigenmaps assign each ultrasound frame a coordinate in low-dimensional space by respecting the neighborhood relationship, they are well suited for analyzing the breathing cycle. For the 4D application, we perform the manifold learning for each angle, and consecutively, align all the local curves and perform a curve fitting to achieve a globally consistent breathing signal. We performed the imagebased gating on several 2D and 3D ultrasound datasets over time, and quantified its very good performance by comparing it to measurements from an external gating system.

Hierarchical Manifold Learning

“…For medical imaging applications, these points can represent whole images or image patches. Some common similarity metrics include functions of the Euclidean norm distance (equivalent to sums-of-squared differences between the images) and the correlation coefficient [14] as well as those based on Gabor filter responses [13]. One advantage of LE over other manifold learning techniques is its capacity to additionally handle non-metric similarity measures such as Normalised Mutual Information [15].…”

“…In recent years, the use of manifold learning has become increasingly widespread in medical imaging, being used to uncover underlying structure both within a subject [14] [17], and across populations [2] [9][11] [15]. Manifold learning techniques aim to discover the intrinsic dimensionality of data: a low-dimensional embedding which retains local structure of the data.…”

“…This can also be analysed using manifold learning and previous work has employed this for ultrasound gating [14] and for lung CT volume reconstruction [8]. Two techniques for measuring respiration are through respiratory bellows and free-breathing navigator-gated strategies.…”

Temporal Groupwise Registration for Motion Modeling

“…The visual inspection and also the calculated statistics images show the good performance of our method for aligning the sequence, and therefore its applicability for registering real medical images. We present the (N + 1)D deformation field and Furthermore, we recover the breathing curve for this sequence using the image-based gating technique proposed in [28], see Fig. 7(d).…”

Manifold Learning for Multi-Modal Image Registration

“…The task of dimensionality reduction is to find the underlying structure in a large set of points embedded in a high-dimensional space and to map these points to a lowdimensional space preserving the structure. Manifold learning has recently gained much attention to assist image processing tasks such as segmentation [31], registration [8,20], tracking [9,28], recognition [2,27], and computational anatomy [6]. Common techniques for manifold learning are Isomap [23], local linear embedding [21], and Laplacian eigenmaps [3].…”