28 August 2013 Diffuse reflectance relations based on diffusion dipole theory for large absorption and reduced scattering
Author Affiliations +
J. of Biomedical Optics, 18(8), 087007 (2013). doi:10.1117/1.JBO.18.8.087007
Diffuse reflectance spectra are used to determine the optical properties of biological samples. In medicine and forensic science, the turbid objects under study often possess large absorption and/or scattering properties. However, data analysis is frequently based on the diffusion approximation to the radiative transfer equation, implying that it is limited to tissues where the reduced scattering coefficient dominates over the absorption coefficient. Nevertheless, up to absorption coefficients of 20  mm −1 at reduced scattering coefficients of 1 and 11.5  mm −1 , we observed excellent agreement (r 2 =0.994 ) between reflectance measurements of phantoms and the diffuse reflectance equation proposed by Zonios et al. [Appl. Opt.38, 6628–6637 (1999)], derived as an approximation to one of the diffusion dipole equations of Farrell et al. [Med. Phys.19, 879–888 (1992)]. However, two parameters were fitted to all phantom experiments, including strongly absorbing samples, implying that the reflectance equation differs from diffusion theory. Yet, the exact diffusion dipole approximation at high reduced scattering and absorption also showed agreement with the phantom measurements. The mathematical structure of the diffuse reflectance relation used, derived by Zonios et al. [Appl. Opt.38, 6628–6637 (1999)], explains this observation. In conclusion, diffuse reflectance relations derived as an approximation to the diffusion dipole theory of Farrell et al. can analyze reflectance ratios accurately, even for much larger absorption than reduced scattering coefficients. This allows calibration of fiber-probe set-ups so that the object’s diffuse reflectance can be related to its absorption even when large. These findings will greatly expand the application of diffuse reflection spectroscopy. In medicine, it may allow the use of blue/green wavelengths and measurements on whole blood, and in forensic science, it may allow inclusion of objects such as blood stains and cloth at crime scenes.
Bremmer, van Gemert, Faber, van Leeuwen, and Aalders: Diffuse reflectance relations based on diffusion dipole theory for large absorption and reduced scattering



Interpretation of diffuse optical reflectance spectra requires translating the remitted detected light flux from the turbid material into an absorption coefficient by applying photon transport models.1,2 Although radiative transport theory is the gold standard here,3 it cannot be used because it lacks analytical solutions that allow general analysis of diffuse reflectances. The diffusion approximation of radiative transport theory2,4 or empirical relations between reflectance and tissue optical properties5 is often used as a next best approach. Unlike empirical relations, diffusion theory requires the reduced scattering coefficient, μs, to be much larger than the absorption coefficient, μa.

Diffuse reflection spectroscopy is applied in several disciplines, e.g., in medicine, in forensic science, in the clothing industry, and in the film animation industry. In this article, we restrict ourselves to medicine and forensic science, where diffuse reflection spectroscopy is used to assess the state of health or disease of patients,1,2,6 and for example, the age of a blood stain found at crime scenes7 or the age of bruises when abuse is suspected.8 In medicine, tissues have μs values in the visible part of the spectrum that vary between 0.5 and 10mm1, whereas μa of whole blood ranges from 0.1 to 30mm1. Consequently, diffusion theory can only be used for infrared wavelengths and/or tissues with small (<1%) blood volume fractions, thus excluding the study of well-perfused organs, or bruises,8 in addition to whole blood itself.5,7 In forensic science, large absorption coefficients occur, e.g., in blood stains at crime scenes or colored clothing but also large (reduced) scattering coefficients, e.g., in wool clothing.

An interesting approach to relate the diffuse reflectance to the optical properties is the approximation to diffusion dipole theory2 by Zonios et al.1 This model matches our experimental setup and shows applicability for larger values of μa, up to μa5mm1, at μs1.73.2mm1, hence ratios of μa/μs up to 3. Although these findings may already expand the application of diffuse reflection spectroscopy, its use on parenchymatous tissues at visible wavelengths or on strongly absorbing and scattering objects found at crime scenes is still difficult if not impossible.

In this article, therefore, we will further explore the application range of Zonios’ model for larger values of the absorption and reduced scattering coefficients. Our approach is to use phantoms with well-defined values for μa and μs. We will increase the μa from 5 (Refs. 1 and 9) to 20 and the μs from 1.7–3.2 to 11.5mm1. Next, we will compare Zonios’ approximate equation with exact diffusion dipole theory,2 which surprisingly allows rationalizing that diffusion theory may remain applicable when absorption is not small compared to reduced scattering. Overall, this approach permits calibration of the experimental set-up of fiber-based probes so that the material optical properties, including small and large μa, μs, can be related to diffuse reflectance values.




Diffuse Reflectance Spectroscopy

Reflectance spectra, R(λ;μa,μs), where λ denotes wavelength, were recorded with a noncontact reflectance spectroscopy setup, containing a spectrograph (USB4000; Ocean Optics, Dunedin, Florida), a tungsten-halogen light source (HL-2000; Ocean Optics), and a probe (QR400-7-UV/BX; Ocean Optics), containing six 400-μm core diameter delivery fibers, circularly placed around an identical central collecting fiber. The probe was tilted 13 deg off-normal and fixed at a height of 17 mm above the phantom, implying a measured illumination spot radius of ri=3.1mm and a collection radius of the remitted light of rc=2.7mm.5 Reflectance spectra were recorded over the wavelength range of 400 to 900 nm and were smoothed by averaging the data points into bins of 10 pixels, which allowed calculation of a standard deviation that represents noise within the signal. Data analysis was at 611 nm, chosen because the absorbing dye of the phantom has its maximum there.

Phantoms consisted of a mixture of Intralipid 20% (Fresenius, Kabi AG, Bad Homburg, Germany), phosphate buffered saline, and Evans Blue (Sigma Aldrich, St. Louis, MO) as absorber.5 The refractive index was assumed to be 1.35. The μa of each phantom was controlled by varying the concentration of Evans Blue, and the μs was controlled by varying the amount of Intralipid 20%. For high Intralipid concentrations (>8%), μs was corrected for dependent scattering effects.10 Evans Blue has an absorption of 18(g/L)1mm1 at 611 nm. Intralipid 20% has μs=18mm1 at 611 nm.11 Phantoms were constructed with μa=0.1, 0.2, 0.4, 1, 2, 3, 5, 10, 20mm1, at μs=1 and 11.5mm1, and a separate set was constructed with μs=0.2, 0.5, 1, 2, 3.8, 5.5, 8.9, 11.48, 12.75, 13.6, 15.3mm1, at μa=1 and 10mm1. Additional phantoms were prepared at each μs without Evans Blue added, and utilized as baseline measurements, R(μa=0,μs). Finally, the reflectance ratio R(μa,μs)/R(0,μs) was calculated.



Our starting point is Eq. (18) of the diffusion dipole theory approach of Farrell et al.2 who calculated the diffuse reflectance R(μa,μs,r) as a function of radial distance, r, in response to a pencil beam at r=0 as





Parameter A depends on the refractive index mismatch of the air–tissue boundary and is given by Ref. 2 as


where Θc is the refractive angle and using n as the tissue–air refractive index mismatch, R0 is given by



This means that A=3 for a phantom refractive index of 1.35. The reflectance ratio in response to an irradiation with radius ri as captured by the collection fiber requires integrating R(μa,μs) and R(0,μs) over the collection spot size with radius rc, with radial position ρ between 0ρrc.1





Because Eq. (5) cannot be evaluated analytically, a simple but approximate analytical expression for R(μa,μs) was derived by Zonios et al.1 They replaced the irradiation area with radius ri by a point source at r=0 and integrated the backscattered light over a circular collection spot with radius rc as



Using Eqs. (1) and (2) for R(μa,μs,ρ), the result was





We added factor 0.5 in Eq. (8) for completeness, despite canceling out here. This factor was missing in the original Eq. (3) for R(μa,μs) and may have caused the need for the empirical intensity factor 1.66.1 Further, we did not empirically change z0 in Eq. (2) into z0=1/(0.425μa+0.9575μs) for μs<10μa and z0=1/μs otherwise, as was done in Ref. 1.



Exact diffusion dipole theory, Eq. (5), are compared with our phantom measurements by numerical integration of R(μa,μs,r) and R(0,μs,r). Approximate diffusion dipole Eq. (4) is evaluated in two ways: first, with the exact parameters A=3 and rc=2.7mm and second, we fitted parameters A, rc to all phantom reflectance measurements, using a Levenberg–Marquandt fitting algorithm,12 with error margins represented by 95% confidence intervals. Table 1 summarizes the various parameters for the three situations.

Table 1

Overview of parameters in Eqs. (3) and (4).

SymbolDefinitionExact valueFitted value
Exact theory, Eq. (3)ρRadial position of collectionfrom 0 to ri
riIllumination spot radius3.1 mm
rcCollection spot radius2.7 mm
ARefractive index parameter3
Zonios’ model, Eq. (4)rcCollection spot radius2.7 mm6.7±0.3  mm
ARefractive index parameter32.0±0.4

We also evaluate the asymptotic behavior of the reflectance ratio, Eq. (8), for small and large μa, μs. We use these results to discuss the method’s applicability at larger optical properties.



The four limits of Eq. (8), for μa,μs0,, as well as for the case that μaμs are large, are summarized in Table 2. Except for μs0, these limits give the correct outcomes.

Table 2

Asymptotic behavior of Eq. (8) for μa, μs′→0, ∞.

R(μs′,μa)R(μs′,0)≈0.5·μs′μa+μs′·{exp(−3μaμa+μs′)+exp[−(1+4A3)·3μaμa+μs′]}, μa≈μs′large(2.5)

The measured reflectance ratios of the phantoms, as well as the models of Farrell, Eq. (3), and Zonios, Eq. (8), (parameters given in Table 1), are shown in Fig. 1 for varying μa at μs are 1 and 11.5mm1, and in Fig. 1(b) for varying μs at μa are 1 and 10mm1. Exact diffusion dipole theory, Eq. (3), has r2=0.918, and Zonios’ Eq. (4) with exact A, rc values (3 and 2.7 mm, respectively) has r2=0.922, differing from the experimental results, especially for low values of μs. However, Eq. (4) with fitted parameters (A=2.0±0.4 and rc=6.7±0.3mm) matches much better the phantom measurements, with r2=0.994.

Fig. 1

(a) Reflectance ratio as a function of the absorption coefficient, μa, for high (squares) and low (dots) reduced scattering coefficients. Green dashed line: exact diffusion theory, Eq. (3), with r2=0.918. Red dotted line: Eq. (8) with exact parameters A=3, rc=2.7mm, with r2=0.922. Blue dot-dash line: Eq. (8) with fitted parameters A=2, rc=6.7mm, r2=0.994. (b) Reflectance ratio as a function of the reduced scattering coefficient, μs, for high (crosses) and low (dots) absorption.


Finally, the reflectance ratio for large μaμs, Eq. (2.5) in Table 2, gives that R(μa,μs)/R(0,μs)0.074. The experimental results [Figs. 1(a) and 1(b)] are 0.087 for μa=μs=1mm1 and 0.075 for μs=11.5mm1 and μa=9.8mm1.



We have shown that Zonios’ approximate diffusion dipole model, Eq. (8), describes the reflectance ratio of phantoms with an r2 as large as 0.994, even up to μa/μs=20 at μs=1mm1. This, however, requires fitting parameters A,rc to all experimental results (Table 1), including those where μaμs. Zonios et al.1 already reported this for μa/μs1.63 but used a different fitting approach (see last paragraph of Sec. 2.2). Venugopalan et al.9 showed validity for μa/μs=3, without a fitting procedure. For exact A,rc parameters and large μs, Eqs. (5) and (8) describe the reflectance ratio with an r2 of around 0.92, even when μaμs. This suggests that diffusion theory may remain applicable beyond its accepted condition of validity to analyze diffuse reflectance spectra, contrary to, e.g., fluence rates (see Fig. 6.2 of Star4).

The explanation is based on the following observations. First, Eq. (8) qualitatively describes the morphology of R(μa,μs)/R(0,μs) versus μa because the limits for μa=0 and large μa,μs are correct. Second, in the numerator of Eq. (8), the last two terms are negligible compared to the first two terms. So, this numerator has two exponentials that include μa and one that also includes A. The denominator includes A,rc. Thus, compared by using exact A,rc values, fitting the A,rc to all experiments obviously upgrades Eq. (8) in describing the experiments. In fact, the fitted A is smaller than the exact A (2 versus 3, Table 1). This enhances the importance of the second exponent (e.g., for μs=10mm1 and μa=1mm1, respectively, 25% versus 12.4% of the first term). Additionally, diffusion theory may remain applicable for higher μa because Eq. (8) depends less and less on the actual values of A,rc if μa,μs become larger because the numerator’s second term, which includes A, or becomes small relative to the first. It helps that rc is also relatively large. Hence, Eq. (8), with exact or with fitted parameters, approaches the same outcome when μs and/or μa become large.

We acknowledge that Eq. (8) becomes an empirical relation when A,rc are fitted to all phantom reflectances, as in Zonios’ approach.1 In contrast, Venugopalan’s method is based on exact diffusion theory and does not require a fitting procedure,9 but they limited their experiments to μa0.0909mm1 and μs0.12mm1, so the applicability for larger absorptions and reduced scattering remains unknown. Nevertheless, compared with empirical relations not based on any underlying theory, where applicability may be limited to a chosen range of μa, μs values, Eq. (8) gives correct asymptotes for very small as well as very large μa and also for very large μs, but not for μs0.

In conclusion, using Eq. (8) with A,rc parameters fitted to phantom reflectance measurements of widely varying μa,μs, allows calibration of fiber-probe set-ups so that the diffuse reflectance values can be related to absorption of the materials under study, even when μa/μs1 and large μs. These findings will greatly expand the application of diffuse reflection spectroscopy. In medicine, it may include blue/green wavelengths in tissues and even whole blood, and in forensic science, it may allow inclusion of strongly absorbing and scattering objects such as blood stains and colored cloth.



G. Zonioset al., “Diffuse reflectance spectroscopy of human adenomatous colon polyps in vivo,” Appl. Opt. 38, 6628–6637 (1999).APOPAI0003-6935http://dx.doi.org/10.1364/AO.38.006628Google Scholar


T. J. FarrellM. S. PattersonB. Wilson, “A diffusion theory model of spatially resolved steady-state diffuse reflectance for the noninvasive determination of tissue optical properties in vivo,” Med. Phys. 19, 879–888 (1992).MPHYA60094-2405http://dx.doi.org/10.1118/1.596777Google Scholar


A. J. WelchM. J. C. van GemertW. M. Star, Diffusion Theory of Light Transport in Optical-Thermal Response of Laser-Irradiated Tissue, 2nd ed., A. J. WelchM. J. C. van Gemert, Eds., pp. 145–202, Plenum Press, New York, NY (2011).Google Scholar


W. M. Star, Diffusion theory of light transport, Chapter 6, Fig. 6.2 in Optical-Thermal Response of Laser-Irradiated Tissue, 2nd ed., A. J. WelchM. J. C. van Gemert, Eds., Plenum Press, New York, NY (2011).Google Scholar


R. H. Bremmeret al., “Non-contact spectroscopic determination of large blood volume fractions in turbid media,” Biomed. Opt. Exp. 2(2), 396–407 (2011).BOEICL2156-7085http://dx.doi.org/10.1364/BOE.2.000396Google Scholar


N. Bosschaartet al., “Limitations and opportunities of transcutaneous bilirubin measurements,” Paediatrics 129(4), 689–694 (2012).http://dx.doi.org/10.1542/peds.2011-2586Google Scholar


G. EdelmanT. G. van LeeuwenM. C. Aalders, “Hyperspectral imaging for the age estimation of blood stains at the crime scene,” Forensic Sci. Int. 223(1–3), 72–77 (2012).FSINDR0379-0738http://dx.doi.org/10.1016/j.forsciint.2012.08.003Google Scholar


B. Stamet al., “Can color inhomogeneity of bruises be used to establish their age?,” J. Biophoton. 4(10), 759–767 (2011).JBOIBX1864-063Xhttp://dx.doi.org/10.1002/jbio.v4.10Google Scholar


V. VenugopalanJ. S. YouB. J. Tromberg, “Radiative transport in the diffusion approximation: an extension for highly absorbing media and small source-detector separations,” Phys. Rev. 58(2), 2395–2407 (1998).PHRVAO0031-899Xhttp://dx.doi.org/10.1103/PhysRevB.58.2395Google Scholar


G. ZaccantiS. Del BiancoE. Martelli, “Measurements of optical properties of high density media,” Appl. Opt. 42(19), 4023–4030 (2003).APOPAI0003-6935http://dx.doi.org/10.1364/AO.42.004023Google Scholar


H. J. van Staverenet al., “Light scattering of intralipid-10% in the wavelength range of 400 - 1100 nm,” Appl. Opt. 30(31), 4507–4514 (1991).APOPAI0003-6935http://dx.doi.org/10.1364/AO.30.004507Google Scholar


K. Levenberg, “A method for the solution of certain problems in least squares,” Q. Appl. Math. 2, 164–168 (1944).QAMAAY0033-569XGoogle Scholar

Rolf H. Bremmer, Martin J. C. van Gemert, Dirk J. Faber, Ton G. van Leeuwen, Maurice C. Aalders, "Diffuse reflectance relations based on diffusion dipole theory for large absorption and reduced scattering," Journal of Biomedical Optics 18(8), 087007 (28 August 2013). http://dx.doi.org/10.1117/1.JBO.18.8.087007
Submission: Received ; Accepted



Diffuse reflectance spectroscopy



Reflectance spectroscopy


Back to Top