28 May 2002 Evaluation of least squares designed contrast-enhancing FIR filters for automatic segmentation of 3D confocal images
Author Affiliations +
Proceedings Volume 4622, Optical Diagnostics of Living Cells V; (2002) https://doi.org/10.1117/12.468352
Event: International Symposium on Biomedical Optics, 2002, San Jose, CA, United States
With recent advances in high-speed confocal imaging, data storage, and computational power, practical high-speed 3D cytometry instrumentation is on the horizon. For 3D cytometry to become practical for example from the perspective of a pathologist, speed attained in part by walk-away automation is fundamentally important. This level of automation can only be obtained with fully automated segmentation of image objects from background. Accuracy of this first image analysis task is crucial since it determines the results of all subsequent quantitative analyses. Confocal cell images often have low contrast due to both inherently low signal-to-noise ratios and high cell- cell contrast ratios that can occupy much of the available imaging dynamic range. A contrast-enhancing technique previously developed for 2D images of fluorescent cell nuclei was extended for 3D confocal images (stacks of 2D image slices). Edge sharpening and contrast-enhancement necessary for automatic thresholding are achieved by filtering with a finite impulse response (FIR) filter. These optimal FIR filters range in size from 3 X3X3X to 13X13X13 and were designed by utilizing the perceptron criterion and nonlinear least squares on confocal training datasets derived from fluorescent microspheres. By utilizing fluorescent beads of known shapes and sizes, the ideal (or standard) segmented image is known a priori. The contrast-enhancing performance of these filters on 3D confocal images of DAPI stained cell nuclei demonstrates that they should lead to accurate, fully automated 3D image segmentation.
© (2002) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Lam K. Nguyen, Lam K. Nguyen, Jeffrey H. Price, Jeffrey H. Price, } "Evaluation of least squares designed contrast-enhancing FIR filters for automatic segmentation of 3D confocal images", Proc. SPIE 4622, Optical Diagnostics of Living Cells V, (28 May 2002); doi: 10.1117/12.468352; https://doi.org/10.1117/12.468352

Back to Top