Content uploaded by Ulrich G. Hofmann
Author content
All content in this area was uploaded by Ulrich G. Hofmann on Oct 23, 2019
Content may be subject to copyright.
Abstract—Intralipid emulsion is often used as optical model
substance to mimic living tissue’s strong scattering properties.
As such it is of considerable importance to utilize realistic
parameters for any type of simulation or calculation in context
of Near Infrared Spectroscopy. We determined optical
characteristics of diluted Intralipid solutions at often used,
realistic volume concentrations ρil and at two wavelengths
(780nm and 850nm) in a simple phantom setup featuring
multiple sensors with different source-detector-separation (SDS)
and penetration depths d. Both, phantom experiments and MC
simulation showed qualitatively similar results and
demonstrated the influence of the three major NIRS factors,
namely the penetrated layer depth (d), the Intralipid
concentration ρil and the source–detector separation (SDS). The
results showed that light reaching the detectors is inversely
related to ρi l and d. It corroborates the need for differential
measurements with at least two SDS to account for superficial
large angle scattering.
I. INTRODUCTION
Functional near–infrared spectroscopy (NIRS) experiences a
revived and growing interest as neuroimaging modality [1],
[2]. It is intrinsically silent, potentially portable and, due to
negligible interferences, not prohibitive to be used
simultaneously with other modalities like EEG [3]–[5].
One of the key factors of a NIRS system’s translational
potential to real world applications is its spatio–temporal
accuracy. In particular, every new NIRS sensor design’s
spatial resolution is of utmost interest for turning
experimental data into useful information, images or even
control signals [BCI cite]. Despite various attempts to
estimate NIRS depth specificity through human skull and thus
* This research was partially funded by the Excellence Cluster BrainLinks-
BrainTools (EXC 1086), and a DAAD stipend to RKA.
R. K. Almajidy and U. G. Hofmann are with the section for
Neuroelectronic Systems, Dept. of Neurosurgery, University of Freiburg
Medical Center, Freiburg, Germany (e-mail:rand.almajidy@klinikum.uni-
freiburg.de; ulrich.hofmann@uniklinikfreiburg.de).
K. Rackebrandt is with Unity AG, Hamburg, Germany (e-mail:
klaas.rackebrandt@unity.de).
H. Gehring is with the Department of Anaesthesiology, University of
Luebeck, Luebeck, Germany( e-mail:hartmut.gehring@uni-luebeck.de).
validate NIRS sensor designs [6],[7] characterizing
experiments still have to be carried out on human
subjects [8]. This clearly hinders engineering throughput and
design cycles, even though numerical simulation has been
used to demonstrate NIR light propagation in different human
body tissues including the human head [9], [10]. However,
reliability and accuracy of light propagation in numerical
simulations depend to a great extent on the actual tissue’s
optical properties. Thus reliable estimates of human tissue
optical properties at NIR wavelengths play an important role
in this regard [11], [12]. They form the foundation for
physical phantoms as calibration data from human subjects is
difficult to acquire and suffers from strong biological
variation [13], [14]. Most reported phantoms of scattering
tissue utilize so called Intralipid, a lecithin–oil emulsion
otherwise used for patients as nourishing diet supplement
[15], even though batch related variations may exist [16].
Still, the expectation is to reliably [17] simulate scattering in
biological tissue by its controlled use [18], [19].
In addition, as optical indices from human samples
considerably depend on the age, gender and race, there does
not one global set of parameters exist to cover them all –
instead ranges and variations have to be expected. The
versatility of our setting may bridge this gap as a tool to
investigate optimum sensor geometry for tissues with
different and easily adjustable optical properties.
Extending the work of Cui and colleagues of 1991 [18] we
aim with the following to provide detailed measurements of
the optical properties of Intralipid at various concentrations
as well as to investigste one NIRS sensor’s depth resolution.
Even more important we demonstrate a simple to reproduce
layered Intralipid emulsion setup, which allows to simulate
light scattering in the NIR wavelength in vitro.
II. METHODS
The experimental setting consists of two laboratory
beakers (SCHOTT DURAN, 1 l and 2 l), the smaller securely
mounted off–center within th bigger by a custom–made
fitting formed with a 3D printer (Makerbot, Replicator2). A
black neoprene sheet is attached to the smaller beaker’s inner
surface to serve as light absorber. Distilled water was mixed
with Intralipid emulsion (Intralipid 20%, Sigma–Aldrich
PCode 1001419820) yielding various Intralipid volume
concentrations (ρil) (0.25, 0.5, 0.75, 1, 1.25, 1.5, 1.75 and 2%,
Rand K. Almajidy, Klaas Rackebrandt, Hartmut Gehring, and Ulrich G. Hofmann
Dual Layered Models of Light Scattering in the Near
Infrared
B: Experimental Results with a Phantom*
respectively). The appropriate mixture is filled into the outer
beaker and serves as the scattering medium (see Fig. 1. (a)).
The two beaker’s glass walls (ca. 2 mm thickness) with the
included Intralipid emulsion are integral part of a three layer
model with two transparent layers. The off–center fixation of
both beakers with an inner light absorbing layer yields with a
single setup and a single Intralipid volume concentration
various light penetration depths, as the distance between the
beaker’s glass walls varies along the outer beaker’s
circumference. To simplify the geometry of light’s paths, we
vertically mount our custom made sensor to the outer beaker
along the height of it.
The sensor features a strip of seven photodiodes (BP 104
FAS OSRAM) aligned at a distance (SDS) of 2, 2.5, 3, 3.5, 4,
4.5 and 5 cm, respectively, from the two NIR–LED sources.
One LED emits at 770 nm (OSA OPTO LIGHT, OIS 330
770) and the other at 855 nm (OSA OPTO LIGHT, OIS–330
IT855) with 6,30 mW/sr radiant intensities, see Fig. 1. (b).
Figure 1. (a) Sketch of the two non–concentric beakers and the multi–
sensor strip positioned for an experimental run, (b) Photo of the
sensor used with highlighting box around the photodiodes.
Measurements are taken at room temperature with the
sensor strip tightly attached vertically to different positions
on the outer beaker’s circumference, thus selecting
penetrated layers depth d between the sensor and the
neoprene light absorber in the inner beaker for each
experimental trial to 0.6, 0.72, 0.8, 1, 1.3, 1.4, 1.5, 1.75, 2,
2.3 and 2.4 cm, respectively (see Fig. 1. (a)). In the above
setting the actual fluid–filled distance between the absorber
and the outer beaker equals d – 4 mm (the thickness of the
two beakers glass walls). The absorbing black neoprene layer
serves to identify the maximum possible NIR photon
penetration depth as the experiment’s purpose was to study
the Intalipid emulsion penetrated depth (d) in relation to the
ρil and the NIRS sensor to SDS distance.
The sensor’s LEDs were controlled by a custom made
electronic described elsewhere [20]. The 770 nm and 855 nm
LEDs were turned on and off sequentially. Photodiode output
at both wavelengths, is amplified by preamplifiers within the
aforementioned electronics and provided voltage values
linearly depending on the backscattered and over the whole
photodiode’s area collected photonic flux (1.7*106 V/A DC
gain). The gain was chosen to guarantee that the minimum
output is greater than zero.
Figure 2. Surface plot of photodiode voltage due to changing SDS at
penetration and Intralipid concentration ρil for penetration depth
d=1cm LED wavelength = 770 nm.
III. RESULTS AND DISCUSSION
A. Multi–sensor Intralipid emulsion setting measurements
Figures 2 and 3 show the NIRS sensors’ measured voltages
for all the Intralipid concentrations and penetration depths.
Each 3D plot visualizes the acquired photovoltages of a
specific penetration layer depth d. They are displayed
depending on the Intralipid concentration and the source-
detector-separation SDS. Not unexpected, the sensor with the
maximal distance to the source (5cm) always shows the
smallest photovoltage. The sensor output decreased when ρil
increased, as well, except for d ≤1cm (see Fig. 3 (a–c)). In
higher concentration Intralipid emulsion layers the
absorption effect is minimal and scattering rules (see Fig. 3
(d–k)). This can also be seen in the in MC simulation results
as well (see Fig. 5 in our companion paper describing the
Optical Measurements and Simulation). Any sensor
measurement is highly dependent on the ρil, especially for d
greater than 1.75 cm and relation for ρil ≥ 1.5%.
Figure 3. Surface plot of PD voltage change due to changing SDS, ρil
and d. LED wavelength = 770 nm.
B. Comparison of MC simulation experimental data
To compare MC simulation (see Fig. 5 in our companion
paper) with the experimental data, we collected MC
simulation data for each SDS over 2.2 mm length resembling
the length of the photodiode sensitive area and thus
integrating over the fluence variations within that area. We
used box plots to illustrate the influence of source detector
separation (SDS) relative to ρil and the detected light fluence
(see Fig. 4.).
For each SDS, the sensor output consists of the
integration of the received light fluence by the photodiode’s
sensitive area. We chose results at minimum (see Fig. 4. (a))
and maximum ρil (see Fig. 4. (b)) at relatively small layer
depths, to show the light fluence detected at each SDS. Both
light fluence and sensor output decreased with SDS
increasing for both ρil. This holds true for all ρil (see Fig. 5.
(a), (b)) were the similarity between the sensor settings and
MC simulation results is further depicted at penetration depth
d = 1.4 cm, a tissue depth we previously investigated using
different experiments with fixed concentrations [14].
Both, numerical or experimental results demonstrate that
sensor output, whether fluence as in the simulation (see Fig.
5 in our companion paper) or voltage as in the experiment,
decreased when ρil or SDS increased. Increasing the source–
detector’s spacing was associated with an increase in the
light penetration depth, nevertheless very large spacing is
known for significant reduction in spatial resolution [21],
[22]. Using multiple LEDs with different light intensities and
a single photodiode to form a multi depth sensor can
compensate for the increased distance effect but the LED
illumination needs to be modulated to prevent cross talk
between the sources [23]. When using a dual LED source and
seven photodiodes, we have the advantage of collecting the
data from all photodiodes at the same instance for each
wavelength as there isn’t cross talk to be avoided. Proper
filtering and gain adjustment are used to compensate for light
intensity changes. Although our phantom setup doesn’t truly
form a human brain phantom, it demonstrates the effect of
scattering medium concentration on NIR penetration depth
and localizes the NIR photon’s travel paths for each SDS. In
Fig. 4. (a),(b) we can see a similar steady decay for both,
sensor’s measured voltage and MC simulation light fluence
with an increasing NIR light source to photodiode distance.
The value of detected fluence depend on concentration ρil and
is governed by the SDS. The larger the SDS is, the stronger
that dependence.
The fluence and sensor measured voltage relation with SDS
suggests that for SDS ≤ 2.5 cm the light detected has mainly
penetrated superficially within depths ≤ 1 cm. For larger
SDS, especially SDS ≥ 3.5 cm the detected light has mainly
undergone multiple scatterings and traveled deeper trough
the phantom..
Figure 4. Both, measured voltages and simulated light fluence values
for all SDS and (a) an Intralipid concentration of 0.25% ρil at d =
0.8 cm and (b) 2% ρil and d = 1 cm. LED wavelength = 770 nm.
The straight lines are weighted least squares linear fits to the data
IV. CONCLUSIONS
Our results (see Fig. 3) show a strong dependency of the
back scattered light on the concentration ρil of the scattering
medium and the permitted penetration depth of the incident
light. This is particularly visible for the long sensor–
detector–distance SDS=5 cm. We hypothesize, that the
reduction of sensor readout with increasing Intralipid
concentration at this SDS is a consequence of the relation
between the scattering media properties and the light
scattering angle.
Figure 5. Box plots of (a) sensor’s measured voltage and (b) MC
simulation light fluence values with SDS for d = 1.4 cm. LED
wavelength = 770 nm for all ρil %.
The light detected by the photodiodes closer to the light
source are mainly photons that have undergone large angle
scattering. The farther the photodiode is separated from the
light source the less dependent the sensor readout is on large
angle scattering. For larger SDS the light detected by the
photodiodes are photons having most likely undergone
multiple scattering [24]. We can see the reduced back
scattering and the increased forward scattering as the ρil
increases. There is a clear increase in large angled light
scattering when the ρil increases while for the multiple
scattered portion forward scattering increased which lead to
decreased back scattered light that reached the farther located
photodiodes. As d increases, the overall behavior is almost
the same, a sign of reduced effect of the absorbent layer on
the detected light.
Our work provides a successful method to characterize the
depth precision of multi detector NIR sensors by employing
both MC simulations as well as Intralipid model substances.
When designing a NIRS sensor to monitor brain activity, the
target areas are usually located at a depth > 1 cm of adults
head surface [25]. The signal detected by those sensors
should be minimally affected by the extracranial layers, so
NIRS sensors design has to incorporate deep penetration
depth and minimum superficial contribution [8]. The results
emphasis the importance of using a dual photodiode sensor to
subtract the superficial contribution to the signal detected by
the further located photodiodes, as even for small d we had
measurements at large SDS, despite the absorber layer effect.
That’s to say a portion of the incident NIR light photons will
always travel superficially and reach the sensor photodiode
even for larger SDS. Increasing the source detector spacing
was associated with an increase in the light penetration depth
and reduction in the extra cranial contamination, nevertheless
very large spacing caused significant reduction in spatial
resolution [21],[22]. Using multiple LEDs, with different
light intensity, and single photodiode to form a multi depth
sensor can compensate for the increased distance effect but
the LED light need to be modulated to prevent cross talk
between the sources [23]. When using a dual LED source and
seven photodiodes, we have the advantage of collecting the
data from all photodiodes at the same instance for each
wavelength as there isn’t cross talk to be avoided.
Our results suggest a photodiode to LED distance to be
around 3.5 cm for brain NIRS system, for the received NIR
light to penetrate deep enough. The results also suggest that
the increase in the photodiode to LED distance within
specific limits will increase that depth. The source detector
separation for the extracranial layer detector should be
around 2 cm.
SDS is a major factor during any NIRS experiments, our
results suggests an SDS < 4 cm to acquire comprehensible
data. We can see in Fig. 3 a low sensor output for most ρil and
d for SDS < 4cm except for thick layer of very low ρil ( Fig.
3. g – k , ρil < 0.75%). We recommended 4 cm > SDS ≥ 3 cm
to monitor a tissue depth greater than 1.4 cm (see Fig. 5. (a),
(b)), where the received sensor voltage and fluence aren’t
minimal for high ρil.
Both our models (Intralipid and Monte Carlo) corroborate
the view, that NIRS readouts are highly sensitive to scatterer’s
layer thickness, concentration and optical properties, next to
geometrical arrangements. These properties do follow quite
sophisticated physical laws and are thus prone to large
changes upon biological variations. We thus conclude, that
real life NIRS applications for bedside imaging have to
incorporate carefully collected human sample values and
patient specific boundary conditions into their reconstruction
of clinically relevant images. Otherwise reconstructed images
may inaccurately relate biological activities from NIRS
imaging to anatomical brain maps – one model might not fit
all patients.
ACKNOWLEDGMENT
The authors are grateful to Olaf Christ for designing the
beaker’s base.
REFERENCES
[1] J. D. Schaeffer, A. S. Yennu, K. C. Gandy, F. Tian, H. Liu, and H.
Park, “An fNIRS investigation of associative recognition in the
prefrontal cortex with a rapid event-related design.,” J. Neurosci.
Methods, vol. 235, pp. 308–15, Sep. 2014.
[2] V. Quaresima, S. Bisconti, and M. Ferrari, “A brief review on the use
of functional near-infrared spectroscopy (fNIRS) for language
imaging studies in human newborns and adults.,” Brain Lang., vol.
121, no. 2, pp. 79–89, May 2012.
[3] R. J. Cooper, N. L. Everdell, L. C. Enfield, a P. Gibson, A. Worley,
and J. C. Hebden, “Design and evaluation of a probe for simultaneous
EEG and near-infrared imaging of cortical activation.,” Phys. Med.
Biol., vol. 54, no. 7, pp. 2093–102, Apr. 2009.
[4] F. Wallois, M. Mahmoudzadeh, a Patil, and R. Grebe, “Usefulness of
simultaneous EEG-NIRS recording in language studies.,” Brain Lang.,
vol. 121, no. 2, pp. 110–23, May 2012.
[5] R. K. Almajidy, Y. Boudria, U. G. Hofmann, W. Besio, and K.
Mankodiya, “A Multimodal 2D Brain Computer Interface,” in 37th
Annual International Conference of the IEEE Engineering in
Medicine and Biology Society, 2015, pp. 1067–1070.
[6] H. Liu, D. A. Boas, Y. Zhang, A. G. Yodh, and B. Chance,
“Determination of optical properties and blood oxygenation in tissue
using continuous NIR light,” Phys. Med. Biol., vol. 40, pp. 1983–
1993, 1995.
[7] T. J. Germon, P. D. Evans, N. J. Barnett, P. Wall, a R. Manara, and R.
J. Nelson, “Cerebral near infrared spectroscopy: emitter-detector
separation must be increased.,” Br. J. Anaesth., vol. 82, no. 6, pp. 831–
7, Jun. 1999.
[8] S. N. Davie and H. P. Grocott, “Impact of Extracranial Contamination
on Regional Cerebral Oxygen Saturation,” Anesthesiology, vol. 116,
no. 4, pp. 834–840, 2012.
[9] V. O. Korhonen, T. S. Myllyla, M. Y. Kirillin, A. P. Popov, A. V
Bykov, A. V Gorshkov, E. A. Sergeeva, M. Kinnunen, and V.
Kiviniemi, “Light Propagation in NIR Spectroscopy of the Human
Brain,” IEEE J. Sel. Top. Quantum Electron., vol. 20, no. 2, pp. 289–
298, Mar. 2014.
[10] G. E. Strangman, Q. Zhang, and Z. Li, “Scalp and skull influence on
near infrared photon propagation in the Colin27 brain template.,”
Neuroimage, vol. 85 Pt 1, pp. 136–49, Jan. 2014.
[11] S. L. Jacques, “Optical properties of biological tissues: a review,”
Phys. Med. Biol., vol. 58, no. 14, pp. R37–R61, Jul. 2013.
[12] W. F. Cheong, S. A. Prahl, and A. J. Welch, “A review of the optical
properties of biological tissues,” IEEE J. Quantum Electron., vol. 26,
no. 12, pp. 2166–2185, 1990.
[13] C. D. Kurth, H. Liu, W. S. Thayer, and B. Chance, “A dynamic
phantom brain model for near-infrared spectroscopy.,” Phys. Med.
Biol., vol. 40, no. 12, pp. 2079–92, Dec. 1995.
[14] R. K. Almajidy, R. D. Kirch, O. Christ, and U. G. Hofmann,
“Estimating the spatial resolution of fNIRS sensors for BCI purposes,”
Proc. SPIE, vol. 8945, p. 894504, Mar. 2014.
[15] H. Bryan, A. Shennan, E. Griffin, and A. Angel, “Intralipid − Its
rational use in parenteral nutrition of the newborn,” Pediatrics, vol.
58, no. 6, pp. 787–790, 1976.
[16] J. T. Allardice, A. M. Abulafi, D. G. Webb, and N. S. Williams,
“Standardization of Intralipid for Light Scattering in Clinical
Photodynamic Therapy Assay system,” Lasers Med. Sci., vol. 7, pp.
461–465, 1992.
[17] P. Di Ninni, F. Martelli, and G. Zaccanti, “Intralipid: towards a
diffusive reference standard for optical tissue phantoms.,” Phys. Med.
Biol., vol. 56, no. 2, pp. N21–N28, Jan. 2011.
[18] W. Cui, C. Kurnar, and B. Chance, “Experimental study of migration
depth for the photons measured at sample surface,” SPIE, vol. 1431,
pp. 180–191, 1991.
[19] L. Spinelli, F. Martelli, A. Farina, A. Pifferi, A. Torricelli, R. Cubeddu,
and G. Zaccanti, “Calibration of scattering and absorption properties
of a liquid diffusive medium at NIR wavelengths. Time-resolved
method,” Opt. Express, vol. 15, no. 11, p. 6589, 2007.
[20] R. K. Almajidy and U. G. Hofmann, “On the design of a multi-channel
NIR system to monitor functional brain activity .,” NIR2013 Proc., pp.
335–338, 2013.
[21] A. Bakker, B. Smith, P. Ainslie, and K. Smith, Near-Infrared
Spectroscopy, Applied Aspects of Ultrasonography in Humans.
InTech, 2012.
[22] A. Liebert, P. Sawosz, D. Milej, M. Kacprzak, W. Weigl, M. Botwicz,
J. Maczewska, K. Fronczewska, E. Mayzner-Zawadzka, L. Królicki,
and R. Maniewski, “Assessment of inflow and washout of
indocyanine green in the adult human brain by monitoring of diffuse
reflectance at large source-detector separation.,” J. Biomed. Opt., vol.
16, no. 4, p. 046011, Apr. 2011.
[23] J. Safaie, R. Grebe, H. A. Moghaddam, and F. Wallois, “Wireless
Distributed Acquisition System for Near Infrared Spectroscopy –
Wda-Nirs,” J. Innov. Opt. Health Sci., vol. 06, no. 03, p. 1350019, Jul.
2013.
[24] S. C. Kanick, V. Krishnaswamy, U. a Gamm, H. J. C. M. Sterenborg,
D. J. Robinson, A. Amelink, and B. W. Pogue, “Scattering phase
function spectrum makes reflectance spectrum measured from
Intralipid phantoms and tissue sensitive to the device detection
geometry.,” Biomed. Opt. Express, vol. 3, no. 5, pp. 1086–1100, May
2012.
[25] M. B. Kim, D. S. Ward, C. R. Cartwright, J. Kolano, S. Chlebowski,
and L. C. Henson, “Estimation of jugular venous O2 saturation from
cerebral oximetry or arterial O2 saturation during isocapnic hypoxia,”
J. Clin. Monit. Comput., vol. 16, no. 3, pp. 191–199, Jan. 2000.
