Automatic material labeling during spectral image data acquisition

Abstract

A system for performing spectral microanalysis delivers analysis results during the course of data collection. As spectra are collected from pixels on a specimen, the system periodically analyzes the spectra to statistically derive underlying spectra representing proposed specimen components, wherein the derived spectra combine in varying proportions to result (at least approximately) in the measured spectra at each pixel. Those pixels having the same dominant proposed component, and/or which contain at least approximately the same proportions of the proposed components, may then have their measured spectra combined (i.e., added or averaged). These spectra may then be cross-referenced via reference libraries to identify the components actually present. During the foregoing analysis, the measured spectra are preferably condensed, as by reducing the number of energy channels/intervals making up the measured spectra and/or by combining the measured spectra of adjacent pixels, to reduce the size of the data cube and expedite analysis results.