Spectral Imaging
Emission Fingerprinting
Emission Fingerprinting is a method for the com-
plete separation (unmixing) of overlapping emission
spectra. It is used with specimens labeled with more
than one fluorescent dye, exhibiting excitation and
emission crosstalk.
The typical raw data for Emission Fingerprinting
are lambda stacks. The previous chapter described
how they are recorded by means of the META
Detector. The second step is to define reference
spectra for all spectral components contained in
the specimen. As a rule, these are dyes interna-
tionally used for labeling the specimen. Other pos-
sible components are autofluorescent and highly
reflecting structures. Autofluorescences, in partic-
ular, often have rather broad emission spectra that
overlap with the fluorescent markers; this makes
them an added source of “impurities” degrading
the signals in conventional laser scanning micros-
copy.
With Emission Fingerprinting, autofluorescences
are simply included in the unmixing process. The
user can subsequently decide between switching
the autofluorescence channel off and using it to
obtain structural information possibly contained in
the specimen.
The reference spectra can either be loaded from a
spectra database, or directly extracted from the
lambda stack. For the latter version, the user has
two options. One is to define spectra via ROIs.
The other uses a statistical method, Automatic
Component Extraction (ACE), to find the refer-
ence spectra. In either case, the images of the
lambda stack must contain structures marked with
a single fluorochrome only.
The third step of Emission Fingerprinting is Linear
Unmixing, which converts the lambda stack into a
multichannel image. Each spectral components of
the specimen is then displayed in one channel
only. The accuracy of the technique allows the
complete unmixing even of such dyes whose
spectra have almost identical emission maxima.
16
Linear Unmixing
Linear mathematical algorithm for spectral unmixing.
If we regard a pixel of a lambda stack that represents a locus in the specimen
where three fluorescent dyes A, B and C with their spectra S(
λ
)
dye A, B and C
overlap,
the cumulative spectrum
Σ
S(
λ
) measured can be expressed as
By means of known reference spectra S(
λ
)
dye A, B and C
, the equation can be
solved for the intensities of the dyes A, B and C, which yields the intensity
shares of the three dyes for this pixel. If this calculation is made for each pixel,
a quantitatively correct 3-channel image results, in which each channel
represents a single dye.
Σ
S(
λ
) = [intensity· S(
λ
)]
dye A
+ [intensity· S(
λ
)]
dye B
+ [intensity· S(
λ
)]
dye C
