Contour based respiratory motion analysis for free breathing CT Justus Adamson a,n , Tingliang Zhuang b , Fang-Fang Yin a a Department of Radiation Oncology, Duke University Medical Center, Box 3295, Durham NC 27710, USA b Department of Radiation Oncology, Cleveland Clinic, Mail Code T28, 9500 Euclid Avenue, Cleveland OH 44195, USA article info Article history: Received 22 March 2011 Accepted 1 August 2011 Keywords: Organ motion FBCT Imaging artifacts Motion management Traveling salesman abstract We propose a method to quantify superior–inferior (SI) motion of a rigid target using the 3D contour from free-breathing CT (FBCT). The technique utilizes similarity between 2D contours (Jaccard Index) and a population based density function for probability of motion amplitude, and is applicable both when the static target shape is and is not known beforehand. Simulations and phantom measurements showed that motion reconstruction is often feasible, with decreasing accuracy as discrepancy is introduced between assumed and actual static shape. When no static shape is used the analysis is most robust for slow scanning speeds relative to the motion period. & 2011 Elsevier Ltd. All rights reserved. 1. Introduction Radiation therapy often requires accurate knowledge of the spatial uncertainties introduced by organ motion. The most common management strategy is adding a margin around the target volume. Four-dimensional computed tomography (4DCT) is the clinical standard for quantifying respiratory motion in radia- tion therapy treatment planning and results in a series of volu- metric images, each at a different phase of the respiratory breathing cycle [1,2]. The respiratory motion from 4DCT can then be summarized in a Maximum Intensity Projection (MIP) volu- metric image [3]. However, estimation of respiratory motion with 4DCT requires greater imaging duration and dose than free breathing CT (FBCT), and it is not as readily available as FBCT due to the dedicated hardware and software required for increased computational power and file storage. Furthermore 4DCT can also be subject to imaging artifacts when breathing is irregular, and use of 4DCT in radiation oncology treatment planning often leads to redundant imaging since it is often accompanied by a FBCT [1]. Another method that is useful for quantifying respiratory motion is fluoroscopic imaging. Fluoroscopy provides motion data at a high temporal resolution but is only useful when the region of interest is visible via fluoroscopy (lung tumor) or when fiducial markers are implanted [4]. Surrogates such as external markers or external contour tracking have also been used for respiratory motion analysis and have the advantage of not requiring imaging with ionizing radiation, but have the disadvantage of limited correlation with internal target motion [5,6]. Furthermore they also require that the relationship between surrogate and internal target motion be derived initially using 4D imaging (fluoroscopy, 4DCT). Respiratory motion causes identifiable and quantifiable arti- facts in FBCT imaging, which can in turn affect treatment plan quality. For example, studies have shown respiratory motion to cause nonphysical features in FBCT such as jaggedness [7], as well as considerable deformation, with spherical objects being shor- tened or elongated by as much as half the amplitude [8]. One recent study showed that splitting artifacts can be eliminated for scanning speeds (a.k.a. velocity of couch translation) above the maximum organ velocity, that slow scanning speeds are useful for obtaining accurate internal target volumes (ITVs), which is the minimum volume that encompasses the entire target throughout all internal organ motion, and that fast scanning speeds are useful for obtaining accurate organ shapes [9]. Most of these studies have focused on reporting what artifacts are expected and possible management strategies. However in theory, some information about the respiratory motion may be extracted directly from the motion induced artifacts. It has been suggested that a motion trajectory could potentially be estimated using prior knowledge of the static organ shape and acquisition parameters; however this has not been investigated in detail [8]. A reliable analysis of respiratory motion using FBCT could have a number of uses. It could be used to quantify whether respiratory motion is substantial enough to justify a 4DCT acquisition. Also, it could serve as a verification of a 4DCT motion analysis, as previous studies have shown that variations in the respiratory pattern can occur over the course of treatment [10,11]. It would also be useful in conjunction with studies in which a series of Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/cbm Computers in Biology and Medicine 0010-4825/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.compbiomed.2011.08.002 n Corresponding author. Tel.: þ919 613 6722; fax: þ919 681 7183. E-mail address: justus.adamson@duke.edu (J. Adamson). Computers in Biology and Medicine 41 (2011) 908–915