Quantitative Assessment of Small Bowel Motility by Nonrigid Registration of Dynamic MR Images Freddy Odille, 1 Alex Menys, 2 Asia Ahmed, 2 Shonit Punwani, 2 Stuart A. Taylor, 2 and David Atkinson 1,2 * Evaluation of small bowel motility from two-dimensional dynamic MRI is of increasing interest for the assessment of segmental diseases such as Crohn’s disease. We propose to use nonrigid image registration for automated motility analy- sis. Registration is particularly difficult when localized inten- sity changes occur (e.g., due to intraluminal flow or through- plane motion). Therefore, displacements were extracted using a joint registration and modeling of intensity changes. Results were analyzed in 10 patients with known or suspected Crohn’s disease. Bowel wall displacements were assessed by tracking bowel cross-sectional markers over time. Automated tracking with the proposed technique was in good agreement with manual tracking (similar bias, improved standard devia- tions). The ability to quantify different grades of bowel motility was investigated by comparing direct motion measures (i.e., changes in bowel cross-sectional diameter or area) and various parametric maps, e.g., based on the Jacobian of the measured displacements with expert grading. Among these measures, the standard deviation over time of the Jacobian was the best at predicting grades from 1 (moving normally) to 4 (not moving at all) in 50 bowel regions with normal and pathologically reduced motility, yielding 93% correct classification with a 1-class tolerance (i.e., same score as the expert 6 1). Magn Reson Med 68:783–793, 2012. V C 2012 Wiley Periodicals, Inc. Key words: Crohn’s disease; Jacobian determinant; image alignment; small intestine; gastrointestinal motility MRI of the small bowel is established in the clinical management of enteric disorders. Evaluation of bowel motility is of increasing interest, both for assessing seg- mental diseases such as Crohn’s disease and potentially for the diagnosis or follow-up of chronic intestinal motil- ity disorders. Examples of pathologies that are known to affect small bowel motility include chronic pain, dys- pepsia, irritable bowel syndrome, chronic intestinal pseudo-obstruction, and bacterial overgrowth (1,2). Long- term monitoring of these disorders is required. Alterna- tive imaging techniques for assessing bowel motility such as X-ray fluoroscopy impart significant doses of ionizing radiation; thus, MRI is a safe option. Small bowel motility assessment from dynamic MRI is challenging because of the complexity of motion pat- terns and the intrinsic limitations of the imaging sys- tem. Motion in the small bowel can be represented by a wave that propagates the chyme through the gastroin- testinal tract, which is referred to as peristalsis, and by radial contractions of smooth muscles at fixed locations along the small bowel, which is referred to as segmen- tal motion (3). Assessment of bowel motility using MRI is limited by the intrinsic constraints of the acquisition system necessitating a trade-off between spatial and temporal resolution. High spatial resolution is required to measure small changes in bowel diameter, whereas temporal resolution needs to be sufficiently high to cap- ture peristaltic activity, typically characterized by approximately 11 waves per minute in normal adults (4). Therefore, acquisition techniques for motility assessment are generally limited to breath-hold two- dimensional (2D) dynamic sequences repeated over multiple slices as three-dimensional sequences are too long and hence impaired by blurring and/or ghosting artifacts. Breath-holding also reduces through-plane motion allowing better visualization of the bowel struc- tures over time. Clinical evaluation of small bowel motility using MRI is generally performed using subjective qualitative assessment by reporting radiologists. In humans, only a limited number of research groups have reported quanti- tative assessment of bowel contraction (4–6). Simple measurement of segmental bowel diameter changes over time provides good characterization of the amplitude and frequency of constriction. However, this requires manual positioning of linear regions of interest across the bowel cross-sectional diameter, a time-consuming process, impractical for clinical use. Recently, a com- plete framework has been proposed for analyzing peri- staltic and segmental motion in rats (7). This framework uses advanced semiautomatic segmentation tools to delineate small bowel segments and medial axis, com- bined with frequency and principal component analyses, and provides detailed information about both peristaltic and segmental motion. In this work, we investigate an alternative approach based on nonrigid image registration to estimate 2D time- varying displacement fields in the sequence of dynamic MR images. The rationale for using image registration is that the whole sequence of images can be processed in a fully automatic manner so as to provide a parametric map that might be used directly for quantitative clinical assessment. The Jacobian associated with these estimated 1 Centre for Medical Image Computing, University College London, London, United Kingdom. 2 Centre for Medical Imaging, Division of Medicine, University College London, London, United Kingdom. Grant sponsor: UK EPSRC; Grant number: EP/D506468/1. *Correspondence to: David Atkinson, PhD, Centre for Medical Imaging, Division of Medicine, University College London, London WC1E 6BT, United Kingdom. E-mail: d.atkinson@ucl.ac.uk Received 26 January 2011; revised 5 October 2011; accepted 21 October 2011. DOI 10.1002/mrm.23298 Published online 10 January 2012 in Wiley Online Library (wileyonlinelibrary. com). Magnetic Resonance in Medicine 68:783–793 (2012) V C 2012 Wiley Periodicals, Inc. 783