Shoji Kido MD, PhD, PRESENTER: Nothing to Disclose

Abstract: HTML Purpose: To eliminate rib opacities on chest radiographs for quantitative evaluation of diffuse lung opacities, we developed an automated elimination algorithm of rib opacities on chest radiographs. We evaluated the usefulness of this algorithm for a computer-aided diagnosis (CAD) algorithm of diffuse lung opacities. Methods and Materials: The computerized scheme of automated elimination of rib opacities is as follows: In the first step, a region of interest (ROI) was selected manually in chest radiographs displayed on a CRT. In the second step, a background density correction was performed, and a Laplacian-Gaussian filter was used for enhancement of rib opacities. Next, the image was binarized. A morphological filter was used for rough segmentation of rib opacities for this binary image, and then the polynomial approximation algorithm was used for estimation of the smooth contour of fine rib opacities. Finally, the rib opacities on the original ROI were eliminated by use of these estimated rib opacities as templates. In the CAD algorithm of diffuse lung opacities, linear opacities corresponding to diffuse lung opacities were extracted. This algorithm used a multiple threshold method. For quantitative analysis, we employed a radiographic index, which was the percent area of extracted opacities in the selected ROI. To evaluate the performance of the automated elimination algorithm of rib opacities, we selected 21 patients with diffuse lung abnormalities and 18 with normal lungs. For each patient, radiographic indices were obtained from two processed ROI images (Rib-: automated elimination algorithm of rib opacities was applied, Rib+: not applied). Results: Automated elimination of rib opacities was successful in 23 of 39 cases (10 with abnormal and 13 with normal lungs). In these 23 processed ROIs, there were significant differences between abnormal and normal lungs (Rib+: 20.18+/-1.99 vs. 16.57+/-2.55, P = 0.0014; Rib-: 10.47+/-2.51 vs. 5.48+/-2.55, P < 0.0001). By use of an automated elimination algorithm of rib opacities, the performance of the ratio of radiographic indices obtained from abnormal and normal (abnormal / normal) lungs improved (Rib+: 1.41+/-0.78 vs. Rib-: 2.03+/-1.02, P < 0.05). Conclusion: Our automated elimination algorithm of rib opacities on chest radiographs can improve the performance of the CAD algorithm of diffuse lung opacities.       Questions about this event email: kido@ai.csse.yamaguchi-u.ac.jp

Kido MD, PhD, S, Automated Elimination of Rib Opacities on Chest Radiographs for Computer-aided Diagnosis of Diffuse Lung Opacities.  Radiological Society of North America 2003 Scientific Assembly and Annual Meeting, November 30 - December 5, 2003 ,Chicago IL. http://archive.rsna.org/2003/3103964.html