|کد مقاله||کد نشریه||سال انتشار||مقاله انگلیسی||ترجمه فارسی||نسخه تمام متن|
|1180882||1491561||2012||10 صفحه PDF||سفارش دهید||دانلود رایگان|
Multivariate curve resolution (MCR) is a useful and important analysis tool for extracting quantitative information from hyperspectral image data. However, in the case of hyperspectral fluorescence microscope images acquired with CCD-type technologies, cosmic spikes and the presence of detector artifacts in the spectral data can make the extraction of the pure-component spectra and their relative concentrations challenging when applying MCR to the images. In this paper, we present new generalized and automated approaches for preprocessing spectral image data to improve the robustness of the MCR analysis of spectral images. These novel preprocessing steps remove cosmic spikes, correct for the presence of detector offsets and structured noise as well as select spectral and spatial regions to reduce the detrimental effects of detector noise. These preprocessing and MCR analysis techniques incorporate the use of an optical filter to prevent light from impinging on a small number of spectral pixels in the CCD detector. This dark spectral region can be incorporated into any spectral imaging system to enhance modeling of detector offset and structured noise components as well as the automated selection of spatial regions to restrict the analysis to only those regions containing viable spectral information. The success of these automated preprocessing methods combined with new MCR modeling approaches are demonstrated with realistically simulated data derived from spectral images of macrophage cells with green fluorescence protein (GFP). Further, we demonstrate using spectral images from the green alga, Chlorella, approaches for the analyses when fluorescent species with widely different relative spectral intensities are present in the image. We believe that the preprocessing and MCR approaches introduced in this paper can be generalized to several other hyperspectral image technologies and can improve the success of automated MCR analyses with little or no a priori information required about the spectral components present in the samples.
Journal: Chemometrics and Intelligent Laboratory Systems - Volume 117, 1 August 2012, Pages 149–158