458
H. M. Heise
spectral quality. As already discussed, best discrimination between both monosaccharides can be obtained by using spectral data from the combination band region,
whereas in higher wavenumber, regions absorptivities are nearly indistinguishable.
For in vivo diffuse reflection measurements, the photon penetration depth in tissue
is wavenumber dependent, much different from in - vitro measurements in cuvettes,
which must be taken into account. A classical least squares (CLS) approach had been
chosen by Maruo and Yamada [41] by evaluating diffuse reflection spectra of human
forearm skin between 1350 and 1850 nm [7400 – 5400 cm
−1 ]. It is based on a modified Beer’s law, assuming that absorbance difference spectra, as measured versus
a first series spectrum, can be modeled by a linear combination of water, protein,
glucose and fat spectra, as well as a baseline for scattering equivalent absorption.
Another approach is SBC calibration modeling [9], which requires data input such
as quantitative analyte absorptivity, as well as estimates of the instrumental noise.
Based on the analyte spectral signals and of those tissue constituents showing crosssensitivities in combination with radiation penetration depth, valuable analytical
method parameters such as limit of detection and method selectivity can be derived
also for diffuse reflection measurement scenarios. Due to known penetration depths,
wavenumber-dependent scaling of the component spectral signatures can thus be
realized.
20.3.3 Non-invasive Technology for Hemoglobin and Blood
Ethanol Monitoring
The importance of hemoglobin (Hb) measurements has already been highlighted
when presenting in vitro assays. However, there are scenarios, where it is desirable to
monitor hemoglobin continuously as during transfusion, for patients under intensive
care or with a postoperative follow-up. In principle, the time-resolved signals from
photoplethysmography are evaluated with the observed alternating current (AC) and
direct current (DC) signals measured. The selected radiation is found within the
visible and SW-NIR spectral range with wavelengths between 600 and 1000 nm,
realized, e.g., by a light-emitting diode (LED) array with center wavelengths of 569,
660, 805, 940 and 975 nm (selected wavelengths represent two isosbestic points
and three for compensation of tissue scattering). For example, different physical
models were considered, and with a “finger model,” the ratio of AC and DC signals
at different wavelengths (at least two) is taken into account. A more sophisticated
model considers also the scattering of the arterial blood arising from the red blood
cells. In principle, one parameter can be established and linked with a wavelengthdependent variable thickness due to scattering. Multiple linear regression analysis
was applied for the prediction of total hemoglobin concentration of 129 different
patients. The relative percentage error and standard deviation of the prediction set
were 8.5% and 1.14 g/dl, respectively [42]. Further improvements were achieved by
designing a special finger probe with optimizations of the detector area, the emission
H. M. Heise
spectral quality. As already discussed, best discrimination between both monosaccharides can be obtained by using spectral data from the combination band region,
whereas in higher wavenumber, regions absorptivities are nearly indistinguishable.
For in vivo diffuse reflection measurements, the photon penetration depth in tissue
is wavenumber dependent, much different from in - vitro measurements in cuvettes,
which must be taken into account. A classical least squares (CLS) approach had been
chosen by Maruo and Yamada [41] by evaluating diffuse reflection spectra of human
forearm skin between 1350 and 1850 nm [7400 – 5400 cm
−1 ]. It is based on a modified Beer’s law, assuming that absorbance difference spectra, as measured versus
a first series spectrum, can be modeled by a linear combination of water, protein,
glucose and fat spectra, as well as a baseline for scattering equivalent absorption.
Another approach is SBC calibration modeling [9], which requires data input such
as quantitative analyte absorptivity, as well as estimates of the instrumental noise.
Based on the analyte spectral signals and of those tissue constituents showing crosssensitivities in combination with radiation penetration depth, valuable analytical
method parameters such as limit of detection and method selectivity can be derived
also for diffuse reflection measurement scenarios. Due to known penetration depths,
wavenumber-dependent scaling of the component spectral signatures can thus be
realized.
20.3.3 Non-invasive Technology for Hemoglobin and Blood
Ethanol Monitoring
The importance of hemoglobin (Hb) measurements has already been highlighted
when presenting in vitro assays. However, there are scenarios, where it is desirable to
monitor hemoglobin continuously as during transfusion, for patients under intensive
care or with a postoperative follow-up. In principle, the time-resolved signals from
photoplethysmography are evaluated with the observed alternating current (AC) and
direct current (DC) signals measured. The selected radiation is found within the
visible and SW-NIR spectral range with wavelengths between 600 and 1000 nm,
realized, e.g., by a light-emitting diode (LED) array with center wavelengths of 569,
660, 805, 940 and 975 nm (selected wavelengths represent two isosbestic points
and three for compensation of tissue scattering). For example, different physical
models were considered, and with a “finger model,” the ratio of AC and DC signals
at different wavelengths (at least two) is taken into account. A more sophisticated
model considers also the scattering of the arterial blood arising from the red blood
cells. In principle, one parameter can be established and linked with a wavelengthdependent variable thickness due to scattering. Multiple linear regression analysis
was applied for the prediction of total hemoglobin concentration of 129 different
patients. The relative percentage error and standard deviation of the prediction set
were 8.5% and 1.14 g/dl, respectively [42]. Further improvements were achieved by
designing a special finger probe with optimizations of the detector area, the emission