20 Medical Applications of NIR Spectroscopy
459
area of a light source and the distance between the light source and the detector. Such
an optimally designed finger probe provided a correlation coefficient of 0.869 and a
standard deviation of 0.81 g/dl in predicting total hemoglobin [43].
Another development of a non-invasive hemoglobin analyzer is presented in [44].
The instrument uses eight laser diodes with wavelengths between 600 nm to 1100 nm
for a synchronous recording of photoplethysmographic signals. For a simplification
of the optical assembly, the light sources were modulated with orthogonal square
waves, and together with the design of a corresponding demodulation algorithm,
a beam-splitting system could be avoided. A newly designed algorithm improved
the accuracy of the dynamic spectrum extraction. A population of 220 subjects was
involved in the clinical testing. A machine learning calibration model, regressing the
plethysmographic data, deriving from the arterial pulse cycle, versus the hemoglobin
concentration, was developed. The correlation coefficient and SEP values were
0.8645 and 0.85 g/dl, respectively. The results indicated that the hemoglobin concentration values could be obtained with acceptable precision and accuracy to allow
future clinical translation [44].
A different wavelength regime is again exploited for non-invasive blood ethanol
quantification, which is similar to the development of glucose NIR-sensors based
on the spectral interval of 8000–4000 cm
−1 . The spectral signatures of ethanol are
even more pronounced as those of glucose, but the concentration interval of interest
is quite similar, i.e., in the per mille range. Standard methods for blood alcohol
determinations include the analysis of breath from subjects by using electrochemical
sensors. More accurate are gas chromatographic methods with head-space analysis,
requiring the sampling of blood by a syringe for laboratory analysis. A non-invasive
in vivo spectroscopic method certainly offers a promising alternative to the other
established assays.
Several companies were interested in the development of fast and reagent-free
analyzers, and a few reports on progress have been published nearly a decade ago.
With the last published paper, an impressive device with appropriate spectral data,
as taken from measurements within the NIR combination band region, had been
presented [45]. Using data from Monte Carlo simulations of the photon transport
and experiments, it could be shown that the tissue fiber-optic probe design had an
expected substantial impact on the effective photon random pathlength through the
skin and the signal-to-noise ratio of the spectroscopic measurements. Spectral data
were recorded within 8000–4000 cm
−1 .
The so-called alternate-site phenomenon is known from blood glucose sensors,
and several clinical studies have shown that interstitial finger and forearm glucose
concentrations can exhibit significant concentration differences over time, with
concentrations in fingers much less delayed compared to blood glucose values.
A similar behavior was observed for non-invasive blood ethanol when finger and
forearm tissues were tested. Results of a 26-subject clinical study with controlled
drinking were reported that were designed to evaluate the spectroscopic technique
preferentially at finger measurement sites in comparison to spectroscopic readings of
alternate volar forearm venous blood, and breath measurements. Comparisons with
results from breath, blood and tissue assays demonstrated significant differences in
459
area of a light source and the distance between the light source and the detector. Such
an optimally designed finger probe provided a correlation coefficient of 0.869 and a
standard deviation of 0.81 g/dl in predicting total hemoglobin [43].
Another development of a non-invasive hemoglobin analyzer is presented in [44].
The instrument uses eight laser diodes with wavelengths between 600 nm to 1100 nm
for a synchronous recording of photoplethysmographic signals. For a simplification
of the optical assembly, the light sources were modulated with orthogonal square
waves, and together with the design of a corresponding demodulation algorithm,
a beam-splitting system could be avoided. A newly designed algorithm improved
the accuracy of the dynamic spectrum extraction. A population of 220 subjects was
involved in the clinical testing. A machine learning calibration model, regressing the
plethysmographic data, deriving from the arterial pulse cycle, versus the hemoglobin
concentration, was developed. The correlation coefficient and SEP values were
0.8645 and 0.85 g/dl, respectively. The results indicated that the hemoglobin concentration values could be obtained with acceptable precision and accuracy to allow
future clinical translation [44].
A different wavelength regime is again exploited for non-invasive blood ethanol
quantification, which is similar to the development of glucose NIR-sensors based
on the spectral interval of 8000–4000 cm
−1 . The spectral signatures of ethanol are
even more pronounced as those of glucose, but the concentration interval of interest
is quite similar, i.e., in the per mille range. Standard methods for blood alcohol
determinations include the analysis of breath from subjects by using electrochemical
sensors. More accurate are gas chromatographic methods with head-space analysis,
requiring the sampling of blood by a syringe for laboratory analysis. A non-invasive
in vivo spectroscopic method certainly offers a promising alternative to the other
established assays.
Several companies were interested in the development of fast and reagent-free
analyzers, and a few reports on progress have been published nearly a decade ago.
With the last published paper, an impressive device with appropriate spectral data,
as taken from measurements within the NIR combination band region, had been
presented [45]. Using data from Monte Carlo simulations of the photon transport
and experiments, it could be shown that the tissue fiber-optic probe design had an
expected substantial impact on the effective photon random pathlength through the
skin and the signal-to-noise ratio of the spectroscopic measurements. Spectral data
were recorded within 8000–4000 cm
−1 .
The so-called alternate-site phenomenon is known from blood glucose sensors,
and several clinical studies have shown that interstitial finger and forearm glucose
concentrations can exhibit significant concentration differences over time, with
concentrations in fingers much less delayed compared to blood glucose values.
A similar behavior was observed for non-invasive blood ethanol when finger and
forearm tissues were tested. Results of a 26-subject clinical study with controlled
drinking were reported that were designed to evaluate the spectroscopic technique
preferentially at finger measurement sites in comparison to spectroscopic readings of
alternate volar forearm venous blood, and breath measurements. Comparisons with
results from breath, blood and tissue assays demonstrated significant differences in