Environmental biology and computer recognition of cells from images.

摘要

This PDE contains two parts: the first being the development of a methodology for the machine recognition of microscopic cells, and the second being a book on environmental science targeting the lay public and defining the larger context of our urgent environmental problems, within which this methodology is a contributing solution.; The approach to machine recognition addressed the problems of high volume, high resolution counting and identification of microscopic cells. Manual counts and identification of microscopic cells are tedious and labor intensive, and obtaining consistent results even with the same researcher is difficult. Studies large enough for statistically significance are impractical due to these constraints. There is a pressing need for such studies in human genome, cancer, and environmental studies. Flow Cytometry is very successful at counting the number of particles, but gives very little information as to what exactly is being counted in a diverse sample. This project introduced a methodology for the computer processing of microscopic events, using fluorescence characteristics, shape analysis, pattern recognition, and other digital image processing techniques to sort, measure, count, and identify microscopic cells automatically. Chromatin distribution characteristics are used in human fibroblast nuclei to correlate nuclei to cell cycle stage, apoptosis, and pre-cancerous conditions. The sorting and recognition of human nuclei is a simple subset of the more difficult phytoplankton recognition problem and can be readily expanded to include that capacity.; The difficulty of isolating cells from background material was addressed in both the preparation of slides before and after imaging with processing techniques of thresholding, convolution kernels, and histogram analysis on a Macintosh computer using NIH-Image software and Pascal macros for automation.