PURPOSE: To evaluate two key processing steps for detection of colon polyps with spiral computed tomographic (CT) colography with perspective volume rendering (PVR): image reconstruction and opacity assignment of the attenuation data. MATERIALS AND METHODS: Spiral CT was performed in 10 patients with known polyps confirmed at colonoscopy, and detailed quantitative analyses were performed of data obtained in four. First, anatomic fidelity of three-dimensional (3D) images generated from two-dimensional (2D) source images with equal voxel dimensions (87%-90% overlap) was compared with 3D images generated from 2D source images with unequal voxel dimensions (0%-80% overlap). Next, the relative dimensions of colorectal polyps to adjacent structures were evaluated for various opacity threshold settings. Then, step and sigmoidal opacity functions were compared with respect to image smoothness and edge sharpness. RESULTS: PVR images generated after interpolation of image data reconstructed with at least 60% overlap were equivalent in image quality to PVR images generated from source images with equal voxel dimensions. Relative polyp-to-haustral fold dimensions demonstrated substantial distortions with opacity thresholds below -700 HU. The 3D PVR images generated with the sigmoidal opacity function were significantly smoother than those generated with the step opacity function (paired t test, P < .02), with small differences noted in edge sharpness. CONCLUSION: Use of highly overlapping source images (87%-90%) was not necessary to generate 3D PVR images of colorectal polyps. Image artifacts were suppressed with use of an appropriate opacity threshold and a sigmoidal opacity function without substantial loss in edge sharpness.
PURPOSE: To evaluate two key processing steps for detection of colon polyps with spiral computed tomographic (CT) colography with perspective volume rendering (PVR): image reconstruction and opacity assignment of the attenuation data. MATERIALS AND METHODS: Spiral CT was performed in 10 patients with known polyps confirmed at colonoscopy, and detailed quantitative analyses were performed of data obtained in four. First, anatomic fidelity of three-dimensional (3D) images generated from two-dimensional (2D) source images with equal voxel dimensions (87%-90% overlap) was compared with 3D images generated from 2D source images with unequal voxel dimensions (0%-80% overlap). Next, the relative dimensions of colorectal polyps to adjacent structures were evaluated for various opacity threshold settings. Then, step and sigmoidal opacity functions were compared with respect to image smoothness and edge sharpness. RESULTS: PVR images generated after interpolation of image data reconstructed with at least 60% overlap were equivalent in image quality to PVR images generated from source images with equal voxel dimensions. Relative polyp-to-haustral fold dimensions demonstrated substantial distortions with opacity thresholds below -700 HU. The 3D PVR images generated with the sigmoidal opacity function were significantly smoother than those generated with the step opacity function (paired t test, P < .02), with small differences noted in edge sharpness. CONCLUSION: Use of highly overlapping source images (87%-90%) was not necessary to generate 3D PVR images of colorectal polyps. Image artifacts were suppressed with use of an appropriate opacity threshold and a sigmoidal opacity function without substantial loss in edge sharpness.
Authors: Valeria A Fisichella; Magnus Båth; Ase Allansdotter Johnsson; Fredrik Jäderling; Tommy Bergsten; Ulf Persson; Kristin Mellingen; Mikael Hellström Journal: Eur Radiol Date: 2009-09-16 Impact factor: 5.315