Linear Unmixing in Direct Processing

If you are using Linear Unmixing in Direct Processing, Automatic Component Extraction (ACE) is not available as the image is not yet created when Direct Processing is set up. It offers only the functionality to import and use reference spectra. You can import reference spectra (Import from) and use the functionality to Calculate Residuals.

With this function you can 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 or Lambda stacks (only available in LSM imaging mode). Note that the functions needs (at least) two fluorescence channels in the input image.

With the knowledge of the spectral characteristic of individual dyes within a sample with multiple dyes, even heavily overlapping individual dye spectra can be mathematically extracted. This method is a pixel-by-pixel image analysis procedure. Ideally, fluorescence spectra of samples labeled with one dye only are acquired and stored in the spectra database as an external reference. This can be done either by employing the spectral detector of a LSM system or by setting up a multichannel experiment on filter based multichannel systems. Then a multi-channel image or Lambda stack from the multi-labeled sample is acquired. The individual dye spectra are then mathematically extracted using the information from the reference spectra. Up to ten different reference signals can be used in the least-square-fit based algorithm to produce a 10-channel output image without any partial overlap between the channels.

Avoid detector saturation of fluorescence signal in the data set to be unmixed. Saturation generates a high signal in the residual channel and will have a negative impact on the unmixing result.

If samples are not available labeled with individual dyes only, the references can be obtained by the following methods:

  • Interactively by user-selection of regions in the image where only one fluorescence dye is present (only available in the Unmixing view).
  • Automatically by Automatic Component Extraction (ACE). Here the software tries to identify pixels in the acquired multichannel image whose intensity results from an individual dye only.

Note that ACE does not work in all cases and linear unmixing can then lead to wrong results. This is especially the case when unmixing widefield multichannel fluorescence images, where there might not be areas which have sufficiently pure single dye contribution. Here it is especially important to acquire single-dye reference spectra first.

Parameter

Description

Import Reference Spectra

For the unmixing process previously generated emission spectra of ideally pure dyes can be loaded and used for unmixing. This function is mutually exclusive to the Automatic Component Extraction function.

–

Import from

Allows you to select and import reference spectra by clicking on .

–

Spectra List

Displays the list of imported spectra with an ID, the File Name and the Channel.

Automatic Component Extraction

Use this function if no reference spectra are available. Indicates the number of components the system should be looking for in the image. The number of components cannot be higher than the number of channels. It will only work if each of the emission signals is present in an area of the image without overlap of another emission signal. Otherwise, ACE cannot produce a reliable result.

Components

Adjust the number of spectrally distinguishable fluorescent components within the imaged sample. The number of extractable components cannot be higher than the number of acquired channels. The maximum possible value is 10 components.

Weighted Unmixing

If activated, spectral channels with high noise contribute less to the unmixing result. This option includes a statistical analysis of the signal-related (Poisson-) noise and weighs the respective contribution for the fitting with the combination of reference spectra to the experimental data.
Note: This option involves a more sophisticated unmixing algorithm and therefore takes longer than the basic unmixing analysis. Weighted unmixing generates improved unmixing results when acquisition channels are not so well balanced but still have a good signal-to-noise ratio.

Autoscale

Activated: Balances the intensity of the unmixed channels to equal levels.

Calculate Residuals

Activated: Generates an additional channel in which the intensity values represent the difference between the acquired spectral data and the fitted linear combination of the reference spectra. In essence, the residual value is the biggest remaining "residual" from the least square fit routine. The residuals are a general measure for how good the fit of the algorithm has performed. The higher the intensity in this additional channel, the worse is the fit of the spectra to the data set.

Widefield Crosstalk Removal

Not visible for LSM images. This option allows Unmixing to be performed on multichannel fluorescence images created without a spectral or confocal detector. Typically, this would be multichannel images acquired with a filter based multichannel microscope system. In this case, the function will automatically create the same number and type of channels present in the input image for the output image of the Unmixing function.

Activated: Removes the crosstalk of widefield channels and ensures that the channel information (emission and excitation wavelength) and metadata of the input image are copied to the output image.
Note that the number of spectra must be equal to the number of channels in the image. The function tries to automatically identify which channel matches which spectrum. If this fails, the function assumes that the order of spectra matches the sequence of the channels.