Projection-based features for reducing false positives in computer- aided detection of colonic polyps in CT colonography Hongbin Zhu 1 , Matthew Barish 1 , Perry Pickhardt 2 , Yi Fan 1 , Erica Posniak 1 , Robert Richards 3 and Zhengrong Liang 1* Departments of Radiology 1 and Medicine 3 , State University of New York, Stony Brook, NY 11794, USA Department of Radiology 2 , University of Wisconsin Medical School, Madison, WI 53792, USA ABSTRACT A large number of false positives (FPs) generated by computer-aided detection schemes is likely to distract radiologists’ attention and decreases their interpretation efficiency. Therefore, it is desirable to reduce FPs as many as possible to increase the detection specificity while maintaining the high detection sensitivity. In this paper, several features are extracted from the projected images of each initial polyp candidate to differentiate FPs from true positives. These features demonstrate the potential to exclude different types of FPs, like haustral folds, rectal tubes and residue stool by an evaluation using a database of 325 patient studies (from two different institutions) which includes 556 scans at supine and/or prone positions with 347 polyps and masses sized from 5 to 60 mm. For comparison purpose, several well- established features are used to generate a baseline reference. At the by-polyp detection sensitivity level of 96% (no loss of detection sensitivity), the number of FPs per scan is 7.8 by the baseline and 3.75 if the new projection features are added, which is a reduction of 51.9% FPs from the baseline. Keywords: Colonic polyps, false positive, projection, feature, computer-aided detection, CT colonography. 1. PURPOSE To reduce false positives (FPs), many researchers applied textural analysis directly in the volume of interest (VOI) to explore the difference of the CT density distributions of the true positives (TPs) and FPs. However, because of the subtle difference in image intensity distribution inside a polyp v.s. its outside surrounding normal tissues, features derived by a voxel-by-voxel fashion inside each VOI have shown limited gain in FP reduction [1]. An implicit operation of projection or line integral through the VOI is expected to enhance the image intensity difference. The usefulness of a single projection image, or so-called electronic biopsy technique, through a suspicious patch has been shown in [2, 3]. For example, a typical polyp larger than 5 mm, including neoplastic and nonneoplastic lesions, would have a uniform concentric ring pattern, with a red core gradually changing to a blue outer ring [3], as shown in Fig. 1(b). According to [3], other colonic objects, e.g., tagged or even untagged stool, impacted diverticula, air bubble artifacts, etc., would have different color patterns. Fig. 1. A 10 mm pedunculated tubular adenoma in the ascending colon of a 62 years old female, showing a typical translucency color signature. (a) The 3D endoluminal view of the polyp. (b) Translucency display applied to the 3D image in (a). (c) The 3D endoluminal view of the same polyp, but viewed at another eye position and direction. (d) Translucency display applied to the 3D image in (c). * Correspondence: Z. Liang; Email: jerome.liang@sunysb.edu ; Phone: (631) 444-7837; Fax: (631) 444-6450. Medical Imaging 2010: Computer-Aided Diagnosis, edited by Nico Karssemeijer, Ronald M. Summers, Proc. of SPIE Vol. 7624, 76240B · © 2010 SPIE · CCC code: 1605-7422/10/$18 · doi: 10.1117/12.844467 Proc. of SPIE Vol. 7624 76240B-1