LSM 710 and LSM 780
LEFT TOOL AREA AND HARDWARE CONTROL TOOLS
Systems
Processing Tab
Carl Zeiss
02/2010 M60-1-0025
e
119
5.3.9
Linear Unmixing
The
Linear Unmixing
processing tool permits to extract the emission of single fluorescence dyes (e.g.
GFP only, YFP only etc.) from strongly overlapping multi-fluorescence data acquired in multi-channel
images and so-called "Lambda stacks" (see section
Imaging in Lambda Mode
).
Linear Unmixing of spectral information from multi-spectral imaging is an established technology known
from remote sensing and satellite imaging. The basics of this increasingly popular image analysis
technique in life science microscopy are described in the following resources (selection from a large
literature, without claiming completeness):
−
Landsford, R., Bearman, G. and Fraser, S.E. / Resolution of multiple green fluorescent protein color variants
and dyes using two photon microscopy.” / Journal of Biomedical Optics (2001); 6, 311-318.
−
Dickinson, M.E., Bearman, G., Tille, S., Landsford and Fraser, S.E. / "Multi-spectral imaging and Linear
Unmixing add a whole new dimension to Laser scanning fluorescence microscopy" / Bio Techniques (2001);
31/6, 1272-1278.
−
Zimmermann T, Rietdorf J, Pepperkok R. / "Spectral imaging and its applications in live cell microscopy." /
FEBS Lett (2003);546:87-92.
−
Timo Zimmermann / "Spectral Imaging and Linear Unmixing in Light Microscopy" / Adv Biochem
Engin/Biotechnol (2005) 95: 245-265
−
or have a look at
http://zeiss-campus.magnet.fsu.edu/
and the respective brochures from Carl Zeiss
MicroImaging GmbH
In brief, with the knowledge of the spectral characteristic of individual components of a multi-component
sample, even heavily overlapping individual spectral characteristics can be mathematically extracted from
experimental multi-channel data. This method is a strictly pixel by pixel image analysis procedure.
Experimentally, fluorescence spectra of mono-labeled samples are acquired and stored in the Spectra
Database (see section
Unmixing View
) as an external reference. Then a multi-channel image or Lambda
stack of the multi-labeled sample is acquired. Finally, the individual components are mathematically
extracted using the information from the reference spectra. Up to ten different reference signals can be
fed into the least-square-fit based algorithm to produce a 10-channel multi-fluorescence stack without
any partial overlap between the channels.
If no mono-labeled samples are available, the references can be obtained by the following methods:
a)
Interactively by user-selection of regions in the image where only one fluorescence dye is present
(only available in the
Unmixing View
tab (see section
Unmixing View
)
b)
Automatically by software analyses of what the individual spectral signatures are. This processing
function is called "Multi Channel Unmixing" or "Automatic Component Extraction" (ACE). Note:
in some cases, spectrally acquired images are not appropriate for ACE or "Multi Channel
Unmixing" and linear unmixing can lead to wrong results.