文摘
This paper presents a novel, semiautomatic method formicroscopic identification of multicomponent samples,which allows the identification, location, and percentagequantity of each component to be determined. Themethod involves applying discriminant analysis to asequence of multichannel fluorescence microscopy images via a supervised learning approach; by selectinggroups of pixels that are representative for each component type in a "known" sample, a computer is"taught"how to recognize the behavior (i.e., fluorescenceemission)of the various components when illuminated under different spectral conditions. The identity, quantity,andlocation of these components in "unknown" samples(i.e.,samples with the same component types but in differentratios or distributions) can then be investigated.Thetechnique therefore enables semiautomatic quantitativefluorescence microscopy and has potential as a qualitycontrol tool. This work demonstrates the applicationofthe technique to artificial and natural samples and critically discusses its quality, potential, and limitations.