Abstract
Abstract: :
Purpose:To evaluate Dunne’s automated technique for volumetric topographic mapping of intraocular tumors. Methods:Three-dimensional files of intraocular tumors were imaged with a commercial 3D B Scan Ultrasound Tomography System (OTI, Canada). These were used to generate volumetric topographic maps using Dunne’s automated technique recently incorporated into the system (proprietary software program courtesy of Shane Dunne, Ph.D – OTI). Three points on the tumor periphery spaced approximately 120 degrees apart, were chosen to define the tumor base plane. The program computed the central axis of the tumor as the line perpendicular to this plane passing through the center of the three points. The posterior border of the tumor modeled as a hemispherical surface was represented on each para-sagittal slice as a red semicircle. The diameter and precise centering of this red semicircle was adjusted until the largest diameter curve was found which just touched the tumor periphery, but did not go beyond the uninvolved portion of the retina. To find the anterior surface of the tumor, the program then looked at each image in the direction of increasing depth to find the first pixel that exceeded a selected threshold. Once the anterior surface and posterior border were found in each image, the program computed the distances between them at all points. It also located the lateral margins in each image wherever the distance dropped to near zero. These steps were repeated three times to assess reliability of measurements. Results:Topographic mapping with this technique could be achieved easily in 10–15 minutes per 3D file. The algorithm simultaneously estimated tumor height, basal dimensions and volume. Volumetric maps were more reliable when generated from 3D files without movement artifacts. Ideal 3D files were well centered in the imaging screen, acquired without too much pressure applied from the probe, and recorded at medium to slightly low gain. Tumor shape, and presence of overlying retinal detachment or underlying scleral extension, were potentially confounding factors in producing consistent topographic maps. Conclusion:This automated technique is quick, but appeared most reliable only for ideal 3D image files of intraocular tumors. Its use may be limited to dome-shaped tumors without associated retinal detachment or scleral extension. Further enhancements to the algorithm may allow mapping of more complex tumor configurations.
Keywords: 432 imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • 554 retina • 610 tumors