| Abstract |
In the field of biomedical infrared spectroscopy there is a drive towards obtaining spectra at high spatial resolution such that biologically relevant information can be obtained at the single cell level [1–
12]. This desire for spatially resolved chemical information applies equally to cells in tissue or isolated cells deposited onto an infrared compatible substrate. The goal is to compare the biochemistry associated
with one cell with that of neighbouring cells in the tissue or culture, usually in an attempt to obtain specific signatures or markers for disease. Unfortunately this is rarely possible since a measured spectrum often contains contributions from absorption, reflection and scattering [13, 14]. Biological cells are typically a few micrometres to a few tens of micrometres in diameter. This is comparable in size to the wavelength of mid-infrared radiation used in a conventional infrared microspectroscopy system. Consequently the cells in the sample scatter the infrared radiation being used to probe them very efficiently. Thus the resulting spectrum contains contributions from both effects. Given that cell size and shape (morphology) is variable within any given sample, obtaining subtle differences in biochemistry alone, from a measured spectrum, is very difficult to achieve. It is therefore imperative that methods are developed to separate the scattering contributions and the pure absorption contribution from the measured spectrum [15–16]. In order to do this one must first understand the predominant cause of the scattering phenomenon. In 2005, Mohlenhoff et al. postulated that Mie scattering from the cell nucleus was the cause of the strong broad oscillations in the spectrum baseline [17]. In 2008, Kohler et al. published the first reliable correction for these Mie type scattering oscillations in the baseline [18], but the sharp reduction in intensity on the high wavenumber side
of the Amide I band, often mistakenly called the "dispersion artefact” frequently remained uncorrected [18]. Recently, Bassan et al. demonstrated that both the broad oscillation in the baseline and the “dispersion artefact” could be described by the phenomenon of Resonant Mie Scattering (RMieS) [16, 19]. It should be noted that the term “dispersion artefact” is a misnomer since the pronounced dip in intensity is only indirectly related to anomalous dispersion and it is not an artefact. It is also not solely a reflection component, as is often mistakenly believed, and although there may well be a pure reflection contribution, this is adequately incorporated in the resonant Mie scattering model (as reflection is a special case of scattering which is explained by Mie theory [20]). Previous papers on the subject of correcting the effects of Mie scattering in IR spectra of scattering samples have utilised the van de Hulst approximation equation [20], for the calculation of the Mie scattering efficiency [18, 19]. This equation is an approximation used instead of full Mie theory [21], which is significantly more complicated and computationally expensive to implement. There are a number of conditions under which the van de Hulst equation should be used: for omogenous non-absorbing spherical particles. Highly scattering samples measured with IR (i.e. single cells and tissues) are inhomogeneous non-spherical absorbing particles rendering the equation of limited applicability. Previous work by Bassan et al. [19] used an adapted form of the van de Hulst equation where the real refractive, n, used was not a constant, but variable. This was a “spectrum” obtained from the
Kramers–Kronig transform of a reference spectrum. A comparison of measured and theoretical data using PMMA microspheres shows that this adaptation of the van de Hulst equation was a reasonable
substitute for the full Mie theory [16]. In this paper we introduce full Mie theory and a method of optimising the reference spectrum iteratively. Full Mie theory is considerably more computationally expensive to implement, and so we have harnessed the power of graphics processing unit (GPU) computing to reduce this time. The full Mie theory (RMieS-EMSC) algorithm is applied to a typical data set consisting of spectra
from single isolated prostate cancer (PC-3) cells which have been a focus of our research recently [6–8, 22, 23]. |
| Referanse |
Bassan, P., Kohler, A., Martens, H., Lee, J., Jackson, E., Lockyer, N., Dumas, P., Brown, M., Clarke, N., Gardner, P. 2010. RMieS-EMSC correction for infrared spectra of biological cells: Extension using full Mie theory and (GPU) computing. Journal of Biophotonics, Vol 3, Issue 8-9, pp 609-620. |
Achim Kohler
Harald Martens
