2 Microscopy and Imaging Systems
63
Registration errors (also called pixel shift) due to the emission filters will
always be a potential problem, particularly in multi fluorochrome applications. If the pixel shift is reproducible, the software should automatically
correct registration errors after having once calibrated the respective shifts
for each filter cube. In applications with similar information in each individual color channel, e.g. comparative genomic hybridization, automatic image registration can be performed by cross-correlation of the color channels.
Raw images usually show some fluorescence background. Manual or
interactive thresholding followed by automatic contrast normalization
will result in enhanced images that display the relevant information at
maximum contrast.
Particularly for quantitative analyses like CGH (metaphase as well as
array-based), mFISH, multi color banding (mBAND), or telomere quantification a non-uniform background will add a variable bias to the result.
An effective, locally adaptive background correction algorithm is absolutely necessary to obtain reliable data. For qualitative FISH analysis
and for image documentation, local or "region of interest" operations
for thresholding, contrast stretching, etc. are helpfuL
Data analysis of FISH images can be divided into two classes: 1. Measurement of image parameters and 2. Classification of objects. Calculation of
CGH ratio profiles and quantitation of telomere size by measuring telomere signal intensity are applications of the first class. Classification
means extracting appropriate data, called features, from the objects of interest and using these features to recognize or identify objects. In mFISH
and mBAND analyses for example, spectral features are analyzed to identify chromosomes or fragments of chromosomes. In scanning applications
morphometric features, often combined with spectral information, are
used to find metaphases or to pin-point rare fluorescent labeled cells
in a huge number of unlabeled normal cells. The analysis strategies employed are highly application-dependent; their detailed discussion is beyond the scope of this chapter.
Practical examples
Figure 7 shows an interphase nucleus with several small hybridization signals (BAC clones) directly labeled with Texas Red. Only the Texas Red
information is displayed in black and white, the DAPI counter stain is
not relevant for the following considerations and has been omitted.
Image
pre-processing
Quantitative image
analysis
63
Registration errors (also called pixel shift) due to the emission filters will
always be a potential problem, particularly in multi fluorochrome applications. If the pixel shift is reproducible, the software should automatically
correct registration errors after having once calibrated the respective shifts
for each filter cube. In applications with similar information in each individual color channel, e.g. comparative genomic hybridization, automatic image registration can be performed by cross-correlation of the color channels.
Raw images usually show some fluorescence background. Manual or
interactive thresholding followed by automatic contrast normalization
will result in enhanced images that display the relevant information at
maximum contrast.
Particularly for quantitative analyses like CGH (metaphase as well as
array-based), mFISH, multi color banding (mBAND), or telomere quantification a non-uniform background will add a variable bias to the result.
An effective, locally adaptive background correction algorithm is absolutely necessary to obtain reliable data. For qualitative FISH analysis
and for image documentation, local or "region of interest" operations
for thresholding, contrast stretching, etc. are helpfuL
Data analysis of FISH images can be divided into two classes: 1. Measurement of image parameters and 2. Classification of objects. Calculation of
CGH ratio profiles and quantitation of telomere size by measuring telomere signal intensity are applications of the first class. Classification
means extracting appropriate data, called features, from the objects of interest and using these features to recognize or identify objects. In mFISH
and mBAND analyses for example, spectral features are analyzed to identify chromosomes or fragments of chromosomes. In scanning applications
morphometric features, often combined with spectral information, are
used to find metaphases or to pin-point rare fluorescent labeled cells
in a huge number of unlabeled normal cells. The analysis strategies employed are highly application-dependent; their detailed discussion is beyond the scope of this chapter.
Practical examples
Figure 7 shows an interphase nucleus with several small hybridization signals (BAC clones) directly labeled with Texas Red. Only the Texas Red
information is displayed in black and white, the DAPI counter stain is
not relevant for the following considerations and has been omitted.
Image
pre-processing
Quantitative image
analysis