Multimodal Diffuse Optical Imaging - Biomedical Optical Imaging Technologies: Design and Applications - page 355

resolution, dense optode measurements are necessary. Whereas to distinguish
efficiently different biological chromophores, acquisition of dense multispectral
data is crucial. The implementation of dense spatial and spectral data sets comes
at an increased cost in terms of both acquisition time and instrument complexity.
The acquisition of high-resolution spatial information is essential for improved
resolution and quantitative accuracy [ 42 ]. However, due to time constraints, con-
fined space in the imaging chambers, and/or hardware memory restrictions for
reconstructions, the spatial measurement density that can be acquired/employed is
realistically limited. The number of measurements becomes then the limiting factor
for quantitative high-resolution imaging. Within these restrictions, it then becomes
essential to identify the best possible spatial sampling strategy. To guide the design
of the instrument and/or develop practical imaging protocols, optimization strategies
can be used a priori to determine the optimal acquisition parameters.
One objective criterion to determine the optimal spatial sampling, consist in
simulating different acquisition scenarios, is to compute the associated forward
models and perform a singular value analysis (SVA) of the Jacobians [ 43 , 44 ].
The distribution of the eigenvalues obtained through singular value decomposition
(SVD) provides an insight on the number of usable eigenmodes in the measurement
space after truncation. The optimal set of parameters is the one providing the
maximum usable measurement eigenmodes. The approach can be used to estimate
the optimal number of sources, number of detectors, number of projections, number
of frequencies, etc. However, it is important to note that this technique provides the
general optimal set of parameters to design instrument and imaging protocols, but
specimen-specific optimization strategies may still be required.
Similarly, the choice of the optimal sets of wavelength for functional imaging
is paramount in designing the most effective imaging platform. The choice of
wavelengths to perform diffuse optical imaging has been historically dictated by the
limited offer of sources in the NIR spectral range. However, with recent advances in
laser diode technology, a much greater choice of wavelengths is available nowadays.
Therefore, objective strategies identifying the optimal set of wavelengths for NIR
imaging should be employed a priori to design the instrument. For functional
imaging studies, the main molecules of interest are oxy-, deoxyhemoglobin, water,
and lipids. The absorption spectra of these chromophores do not have preeminent
features in the NIR spectral window. Thus, when employing a limited number of
spectral information, quantitative estimation of the concentration suffers from cross
talk between the absorbers investigated.
Corlu et al. [ 45 ] proposed an objective strategy to select the optimal set of
wavelengths, given a fixed number of wavelengths available and given the a priori
knowledge of the absorption spectra of the chromophores. Based on simulations,
they identified the best set of 4 wavelengths that provided minimal cross talk
between oxy-, deoxyhemoglobin, and water by maximizing the uniqueness with
which the chromophores can be distinguished. Recently, Brendel et al. [ 46 ]have
extended the formulation of Corlu in the case of uncertainty in the extinction
coefficient of the chromophores. Available absorption spectra from the literature
show significant deviations, and thus this uncertainty should be taken into account