6th ESACP Congress, Heidelberg, April 7-11, 1999

A012
QUANTITATIVE MICROSCOPY: PROBLEMS AND SOLUTIONS
Kam Z

Molecular Biology of the Cell, Weizmann Institute of Science, Rehovot, Israel

Light microscopy is very appealing as an automated tool in biology and medicine. Electronic imaging and computerized analysis, together with a wide range of sensitive and specific labeling methods for live and fixed cells and tissues, can generate multiparametric molecular and morphological information at cellular and subcellular resolutions to be used in basic research, and for diagnosis. There are a number of aspects that require attention. First, one must rely on a small sample size. As compared to cell sorters or biochemical essays, microscopic imaging can scan and analyze thousands of cells, as compared to millions. The inherently wide distribution of cellular parameters imposes severe constraint on the statistical quality of the results. In addition, distribution of parameters quantified from images often behave as fractals, meaning that the number of small values increase fast, and merge into noise. This makes average values practically meaningless, and requires application of complex multivariable statistical methods. The instrumental aspects of quantitative light microscopy are not less important, and new technology offers solutions for current and future applications. Microscopes were designed for optimizing a compromise between intensity and homogeneity of the illuminated field. A factor of two fall in intensity at the edges of the field are common, but are not acceptable for quantitative transmission or fluorescence imaging. Spatially and temporally stable illumination enable calibration, but arc lamps hot-spot jumps introduce large changes. Fiber scramblers and intensity integration for each exposure solve this problem. Sensitive and wide dynamic range CCDs with computerized image processing allow acquisition of multicolor and time dependent pictures of fields. But image interpretation software is not an off-the-shelf product. We have developed segmentation algorithms to identify regions in DIC and fluorescence microscopy, followed by morphological and intensity-integration and background-subtraction for multiple colors, in addition to colocalization, energy transfer and ratio imaging that produce quantitative visual data. Aspect of intensity-integration in confocal and wide field microscopy will be discussed with application to cell cycle analysis on single cell basis. Cell suspensions and blood cells are relatively easy specimens for automated analysis. Confluent tissue cultures, and sections are much less so. They require three dimensional imaging, and 3D algorithms for image interpretation are rare, and time consuming. They are essential though diagnosis of biopsies.