Noninvasive in vivo imaging is a rapidly growing field with applications in basic biology, drug discovery and clinical medicine. Because of the high cost of magnetic resonance (MR)- and computed tomography (CT)-based systems, a great deal of effort has gone into developing optical imaging methods, which offer, in some modalities, the promise of high spatial resolution and the ability to detect multiple markers simultaneously However, the ability to image and quantitate fluorescently labeled tumors and other fluorescently labeled markers in vivo has generally been limited by the autofluorescence of the tissue, which reduces the sensitivity of detection and accuracy of quantitation of the labeled target. Multispectral imaging methodology, which spectrally characterizes and computationally eliminates autofluorescence, enhances signal-to-background dramatically, revealing otherwise invisible labeled targets. Signal-to-noise considerations can guide the choice of appropriate sensors for fluorescence-based imaging, which generally does not benefit from the use of highly cooled (and expensive) cameras. Effective use of spectral tools to remove autofluorescence signal requires accurate spectra of the individual components. Using manual and automated algorithms to generate these spectra, it is possible to detect as many as three fluorescent protein-labeled tumors and two separate autofluorescent signals in a single subject.