Computer aided detection of endobronchial valves Robert A. Ochs * , Jonathan G. Goldin, Fereidoun Abtin, Raffi Ghurabi, Ajay Rao, Shama Ahmad, Irene da Costa, Matthew Brown Department of Radiology, University of California, Los Angeles ABSTRACT The ability to automatically detect and monitor implanted devices may serve an important role in patient care and the evaluation of device and treatment efficacy. The purpose of this research was to develop a system for the automated detection of one-way endobronchial valves implanted as part of a clinical trial for less invasive lung volume reduction. Volumetric thin section CT data was obtained for 275 subjects; 95 subjects implanted with 246 devices were used for system development and 180 subjects implanted with 354 devices were reserved for testing. The detection process consisted of pre-processing, pattern-recognition based detection, and a final device selection. Following the pre-processing, a set of classifiers were trained using AdaBoost to discriminate true devices from false positives (such as calcium deposits). The classifiers in the cascade used simple features (the mean or max attenuation) computed near control points relative to a template model of the valve. Visual confirmation of the system output served as the gold standard. FROC analysis was performed for the evaluation; the system could be set so the mean sensitivity was 96.5% with a mean of 0.18 false positives per subject. These generic device modeling and detection techniques may be applicable to other devices and useful for monitoring the placement and function of implanted devices. Keywords: Computed Aided Detection, Endobronchial Valves, CT, AdaBoost Cascade 1. INTRODUCTION Less invasive emphysema treatments involving the placement of multiple one-way endobronchial valves in patients have been developed, and some are already in clinical use (Maxfield, 2004; Toma et al., 2003). The objective of these alternative therapies is to place one-way valves into the airways of a targeted, diseased lobe of the lung in order to prevent entry of air into the lobe, while still allowing air inside the lobe to flow out. By only allowing air out, the treatment aims to reduce the volume of the lobe so that neighboring, healthier lung tissue will expand, providing a treatment benefit to the patient. The motivation for this research is to improve the care of patients implanted with medical devices. The ability to detect and monitor implanted devices is important for the evaluation of device and treatment efficacy. CT imaging and computer systems offer the potential of a less invasive approach for this task. No methods for detecting implanted devices from volumetric CT images were found during the literature review. Techniques have been developed for detecting fiducial markers on 2D projection images to aid target location for radiation therapy (Harris et al., 2006). An additional template matching technique has been developed to detect paramagnetic markers on 2D MR projection images (van der Weide et al., 2001). Markers are often comprised of solid high attenuation material, allowing them to be detected on projection images; however, endobronchial valves may be comprised of thin wires, more complex geometry, and can be obscured by local tissue or calcifications making them more difficult to detect on both 2D and 3D images, as illustrated by Figure 1. The aim of this work was to develop a system for the automated detection of endobronchial valves (Emphasys Medical, Inc., Redwood City, CA). The underlying detection method utilizes generic techniques, which may allow it to be extended for use with other types of implanted devices. * rochs@mednet.ucla.edu Medical Imaging 2008: Computer-Aided Diagnosis, edited by Maryellen L. Giger, Nico Karssemeijer Proc. of SPIE Vol. 6915, 691519, (2008) · 1605-7422/08/$18 · doi: 10.1117/12.770692 Proc. of SPIE Vol. 6915 691519-1 2008 SPIE Digital Library -- Subscriber Archive Copy