(604c) Spectral Imaging Analysis Methods for Fluorescence Microscopy | AIChE

(604c) Spectral Imaging Analysis Methods for Fluorescence Microscopy

Authors 

Leavesley, S. J. - Presenter, University of South Alabama


Spectral imaging is a technology that was originally developed by the remote sensing community for identifying objects in an image based on their characteristic spectral properties. While traditionally based on the reflectance spectrum of satellite or aerial images, spectral imaging has more recently found a wide array of uses in biomedical applications. Imaging and analysis of the fluorescence excitation or emission spectrum has shown to be especially promising, due to the large number of functional fluorescence labeling methods that have been developed (immunohistochemical labels, fluorescent proteins, functionalized dyes and nanoparticles). Spectral imaging approaches have been applied to both macroscopic (in vivo imaging) and microscopic (fluorescence and confocal microscopy) biomedical imaging platforms. In fact, spectral equipment configurations are now available from most large microscopy companies (Nikon, Olympus, Zeiss), with corresponding analysis software packages. While these systems provide state-of-the-art spectral microscopy capabilities, analysis algorithms marketed with these systems are somewhat streamlined, and generally focus on the use of linear unmixing methods to discriminate between each of the fluorescence signals of interest. One key concern of this streamlined analysis is the accurate identification of pure-component (end-member) spectra. There are many techniques that have been developed in the remote sensing community to identify spectral end-members, but these have generally not been incorporated into current software packages, as they require relatively high levels of user input. Additionally, standard spectral libraries often do not take into account changes in the optical configuration of the microscope that can attenuate the fluorescence signal in a wavelength-dependent manner. These factors can lead to an inappropriate use of controls and end-members when performing spectral analysis, and thus, an inaccurate assumption that the unmixed images actually represent the relative abundances of each end-member in the spectral image set.

We have applied several spectral imaging technologies to investigate regulators of intracellular signaling molecules in highly autofluorescent tissue samples. In specific, we have used spectral fluorescence and confocal microscopy to investigate fluorescent protein labels in the pulmonary vasculature, a tissue that presents high levels of autofluorescence due to its elastin and collagen content. The analysis of this image data is not straight-forward, and has required the construction of custom spectral libraries to model the various autofluorescence components of the lung. We will present the results from these studies, with a specific focus on methods for calibrating spectral image systems and constructing a spectral library for a given animal/tissue model. Additionally, we will present methods for correcting spectral image data for various equipment configurations (alterations to the optical path). These methods are applicable to many spectral imaging scenarios, and will become increasingly important as spectral microscopy systems become available at research institutions.