Biomedical Engineering Reference
In-Depth Information
fibrils. Water immersion objectives with 0.6-1.0 numerical aperture and 10-40× magnification
are a good choice. Since fibrotic processes are usually heterogeneous, it is worth mapping larger
sample areas by using a motorized stage to scan the tissue in the focal plane. Subsequent stitching
of the images provides centimeter-size images with micrometer resolution of a full human biopsy
or animal small organ.
• Quantitative imaging requires using always the same experimental conditions for the sake of
reproducibility. In practice, it pertains to the sample preparation and to the excitation power.
Quantitative imaging may be performed indifferently in fresh, fixed, or frozen samples; however,
since different sample preparations result in different tissue properties, all the samples must be
prepared the same way to allow for reliable comparison. Images must then be recorded using the
same setup with the same excitation wavelength, the same filters, the same detector settings, and
the same excitation power. In these conditions, there is no need for any reference. However, slight
misalignments of the setup may induce variations of the laser pulse profile or of the focal volume
and affect the SHG response. A reference may then be used for long-term experiments or for com-
parison between similar setups in different laboratories. The most convenient reference sample is
a fixed biopsy of a collagenous tissue (for instance, unstained paraffin-embedded skin dermis).
• Simultaneous recording of the 2PEF image is strongly recommended to visualize the tissue mor-
phology by use of endogenous fluorescence signals. Segmentation of relevant regions of interest
may then be implemented to selectively quantify the fibrosis in these regions. For instance, kidney
fibrosis has to be scored in the cortex where interstitial fibrosis takes place. We therefore devel-
oped a semi-automated segmentation algorithm based on the 2PEF image to extract the cortical
region from our murine coronal slices (see Figure 15.3). Briefly, we manually outlined the capsule,
the papilla, and the series of arcuate arteries and fitted these curves as ellipses. The ellipses were
then dilated or contracted to eliminate the frontiers and delimitate the cortical region without
any contribution from the arcuate arteries, the medulla region, and the papilla region (Strupler
et al. 2008).
15.4.3 image Processing to Quantify Fibrosis extent
Algorithms have been developed to automate SHG image processing and optimize the accuracy of fibro-
sis quantization. Figure 15.9 displays the algorithm developed by Strupler et al. (2007) for renal fibrosis
and highlights the most important steps for reliable fibrosis quantization. Note that it can be imple-
mented using ImageJ software (custom-written macros) or using MATLAB
®
for large stitched images.
15.4.3.1 correction for nonhomogeneous illumination of the Sample
The laser excitation is usually slightly more important in the center of the field of view of the micro-
scope than at the edges. Strictly speaking, correction for this vignetting effect is not necessary since it
is the same for any sample. However, it is preferred for a better outcome when stitching several images
or reconstructing 3D volumes. For that purpose, the multiphoton excitation profile is obtained from
parabolic 2D fitting of the 2PEF image of a fluorescent plastic slide, and it is used to divide the SHG and
2PEF images. Note that the laser excitation also exhibits depth attenuation due to absorption, scattering,
and aberrations in the tissue. We hypothesize that this effect is similar for control and fibrotic sample
and do not impede quantitative comparisons. It is usually not corrected for because the signal-to-noise
ratio and the resolution also deteriorate when depth increases and cannot be satisfactorily restored by
image processing.
15.4.3.2 Background correction
Background in the SHG images is quite small once the laser excitation is properly filtered out using laser
blocking emission filters and the SHG is properly selected using around 10 nm-wide bandpass filters.
The small background is then mainly due to the blue-edge component of the fluorescence partly passing
