437
15 FtIR Imaging of tissues: techniques and methods of Analysis
redirected by plane mirrors, collimated by paraboloidal mirrors, and then rearranged
by another set of plane mirrors to illuminate a sampling area [62]. the 3 × 4 beam
matrix homogenously illuminates an area up to 60 × 40 μm
2
(96 × 96 pixels) with a
pixel projection of 0.54 × 0.54 μm
2
on the sample plane with the use of a 74×, 0.65
NA objective. According to the Rayleigh criterion, for this type of objective, diffraction limited resolution can be achieved with pixel size not larger than 0.59 μm.
the effective geometric pixel area of IRENI meets this criterion [62]. Since, this
pixel size is smaller than the maximum pixel size allowed for correct spatial sampling, diffraction-limited images are possible even at the wavelength of 2.5 μm.
Furthermore, such a design significantly reduces acquisition time from 13 days for
raster scanning of a 280 × 320 μm
2
area using a synchrotron-based IR microscope
at diffraction-limited resolution to ca. 30 min for FPA imaging. the most important
advantage of the IRENI setup is achieving high-contrast images, which clearly resolve subcellular structures of 2–7 μm diameter in a comparable manner to H&E
stained images [57, 62]. Spectral information derived for different substructures is
improved due to pixel localization. In addition, measurements of 1 μm polystyrene
beads have shown that the synchrotron-based FPA system can reach a spectral limit
of detection of 6 ± 1 fmol in a single 0.54 × 0.54 μm
2
pixel area [57]. While there
is certainly an improvement in contrast and spatial resolution the actual signal-tonoise ratio of the resulting spectra are somewhat compromised because in order to
illuminate the FPA evenly the lower Cassegrain condenser must be defocused over
the entire array. defocusing the condenser results in poorer quality spectra and so
this is something that should be addressed in future configurations.
15.4 Spectral/Image Processing Techniques
Prior to image reconstruction the spectra are usually subjected to some form of preprocessing to compensate for scaling differences. Scaling differences arise from
path length effects, scattering effects, source or detector variations, or other general instrumental sensitivity effects. there are two general types of FtIR spectra
pre-processing methods namely filtering and model-based methods. the filtering
methods, include polynomial or linear baseline correction, calculating second derivatives, smoothing to remove noise and normalization for differences in sample
thickness. model-based methods enable the quantification and separating out of the
different types of physical and chemical contributions to the spectra. the information in this case is not lost and statistical estimates of the mathematical parameters
involved in the filtering can be obtained [63]. the type of physical contributions to
the spectra will dictate the pre-processing regime that is required.
15 FtIR Imaging of tissues: techniques and methods of Analysis
redirected by plane mirrors, collimated by paraboloidal mirrors, and then rearranged
by another set of plane mirrors to illuminate a sampling area [62]. the 3 × 4 beam
matrix homogenously illuminates an area up to 60 × 40 μm
2
(96 × 96 pixels) with a
pixel projection of 0.54 × 0.54 μm
2
on the sample plane with the use of a 74×, 0.65
NA objective. According to the Rayleigh criterion, for this type of objective, diffraction limited resolution can be achieved with pixel size not larger than 0.59 μm.
the effective geometric pixel area of IRENI meets this criterion [62]. Since, this
pixel size is smaller than the maximum pixel size allowed for correct spatial sampling, diffraction-limited images are possible even at the wavelength of 2.5 μm.
Furthermore, such a design significantly reduces acquisition time from 13 days for
raster scanning of a 280 × 320 μm
2
area using a synchrotron-based IR microscope
at diffraction-limited resolution to ca. 30 min for FPA imaging. the most important
advantage of the IRENI setup is achieving high-contrast images, which clearly resolve subcellular structures of 2–7 μm diameter in a comparable manner to H&E
stained images [57, 62]. Spectral information derived for different substructures is
improved due to pixel localization. In addition, measurements of 1 μm polystyrene
beads have shown that the synchrotron-based FPA system can reach a spectral limit
of detection of 6 ± 1 fmol in a single 0.54 × 0.54 μm
2
pixel area [57]. While there
is certainly an improvement in contrast and spatial resolution the actual signal-tonoise ratio of the resulting spectra are somewhat compromised because in order to
illuminate the FPA evenly the lower Cassegrain condenser must be defocused over
the entire array. defocusing the condenser results in poorer quality spectra and so
this is something that should be addressed in future configurations.
15.4 Spectral/Image Processing Techniques
Prior to image reconstruction the spectra are usually subjected to some form of preprocessing to compensate for scaling differences. Scaling differences arise from
path length effects, scattering effects, source or detector variations, or other general instrumental sensitivity effects. there are two general types of FtIR spectra
pre-processing methods namely filtering and model-based methods. the filtering
methods, include polynomial or linear baseline correction, calculating second derivatives, smoothing to remove noise and normalization for differences in sample
thickness. model-based methods enable the quantification and separating out of the
different types of physical and chemical contributions to the spectra. the information in this case is not lost and statistical estimates of the mathematical parameters
involved in the filtering can be obtained [63]. the type of physical contributions to
the spectra will dictate the pre-processing regime that is required.