Qiang Li PhD, PRESENTER: Nothing to Disclose

Abstract: HTML Purpose: To develop a high performance computer-aided diagnostic (CAD) scheme for lung nodule detection in thin-slice computed tomography (TSCT), which will provide output as a second opinion to assist radiologists in improving their detection accuracy for lung nodules. Methods and Materials: Our database consisted of 112 thin slice CT scans with 153 nodules obtained in Shinshu University, Japan (80 scans, 91 nodules, including 41 cancers and 50 benign nodules) and the University of Chicago (32 scans, 62 nodules). The database included nodules with a large variation in size (5-30 mm, mean 11 mm), shape, and contrast (solid and ground glass opacity). Our CAD scheme consisted of a lung segmentation, an initial nodule detection, and a feature extraction and analysis techniques. Rule-based classification technique was first used to remove many false positives, and linear discriminant analysis was employed to further eliminate false positives based on a leave-one-case-out testing method. To achieve a high performance, two new techniques, i.e., a three-dimensional (3D) selective, multi-scale nodule enhancement filter and a 3D accurate segmentation technique, were developed. Unlike conventional enhancement filters, the selective nodule enhancement filter consisted of a magnitude and a likelihood components, which were related to nodule contrast and shape, respectively. Therefore, only nodule-like circular objects with a relatively large contrast were substantially enhanced, whereas other normal anatomies such as vessels were significantly suppressed. For the 3D accurate segmentation, reliable edge candidates were first detected by use of an object-based contour line method that is significantly more robust to noise than pixel-based edge detectors such as Sobel operator. A spiral scanning technique was developed to sequentially sample the edge candidates in 3D image space, in order to facilitate dynamic programming technique to connect optimal edge candidates in terms of a cost function. Results: In this preliminary study, our CAD scheme detected 138 of 153 nodules (90% sensitivity) with a false positive rate of 5.9 per scan. Conclusion: The CAD scheme achieved a low false positive rate and a high detection rate for nodules with a large variation in size, shape, and contrast, and would be useful in improving radiologists' performance in detecting early lung cancer in TSCT scans.     (K.D. is a shareholder in R2 Technologies Inc., Los Altos, CA, and in Deus Technologies Inc., Rockville MD.)

Li PhD, Q, A High Performance Computer-aided Diagnostic Scheme for Lung Nodule Detection in Thin-slice CT Based on 3D Nodule Enhancement Filter and 3D Spiral Segmentation Technique.  Radiological Society of North America 2003 Scientific Assembly and Annual Meeting, November 30 - December 5, 2003 ,Chicago IL. http://archive.rsna.org/2003/3104835.html