Answers are grounded in this paper and cite their sources.
We are developing novel deconvolution algorithms for 3D light microscopy that include the widefield fluorescence, transmitted light brightfield and confocal fluorescence modalities. The main novelties that we have introduced are (1), the usage of maximum likelihood estimation as a mathematical optimization criterion and (2), incorporation of blind deconvolution such that the optical point spread function does not need to be measured (see references in 1,2). Our recent research has included evaluations of performance. This includes validating the images, using artificial test objects of (reasonably) “known” geometry and evaluating the accuracy of morphometries that are aided by deconvolution. Test objects include microspheres and specially drawn, fluorescent filled micropipette tips. The performance measures are comparisons of various morphometries against chosen benchmarks Rat hippocampal neurons filled with lucifer yellow were used in evaluating performance with a biological sample. Morphometries obtained from nondeblurred and deblurred data sets, by way of automatic segmentation, were compared against benchmark measurements. The benchmarks, in these cases, were measurements obtained by way of the Stereocon software. The Stereocon software uses operator judgement to segment the object, and, in our case, uses the assumption that the dendrites have circular cross sections. This assumption provides a realistic benchmark measurement because it ensures that the thickness of the dendrites, on average, is isotropic. Some of these benchmark comparisons, with confocal fluorescence, are summarized in Table 1. Fig. 1 shows projectional images of the data sets from which some of these measurements were obtained. These data imply that incorporation of deconvolution methods will improve the accuracy of absolute morphometric measurements whenever automated segmentation schemes are embedded within the measurement process.
No takes yet. Share an insight, caveat, or question.
Holmes et al. (1996) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: