|
1.IntroductionSpectroscopy is a common and effective method for calculating the absorption and scattering coefficients of biological tissues and other turbid media. The method requires measuring the reflectance of the medium. An inversion algorithm is then used to reconstruct the optical properties of the medium, such as the absorption coefficient and the scattering coefficient. For this purpose, the diffusion equation is employed,1–4 which, however, has several limitations in the application. First, the diffusion theory is applied when the light absorption coefficient is smaller than the scattering coefficient. Second, the diffusion theory becomes invalid when the distance between the detector and the light source increases to several transport mean free paths. Some researchers have published methods for extracting tissue optical properties in low-albedo biological media, or methods for small source–detector separations. Venugopalan et al.5 developed an equation estimating approximate increases in the diffusion of the radiation transport. Kienle and Patterson6 developed a method of fitting the optical properties of a reflectivity measurement of a low exposure medium close to the source by fitting the data to a series of reflection curves produced by a single Monte Carlo simulation. Jia et al.7 established the method of “virtual source” diffuse approximation. These methods, to some extent, improved the accuracy of the diffusion equations. However, researchers are still looking for new methods to alleviate some of their limitations further. For example, using the radiative transfer equation,5 called the approximation model, it is difficult to establish a mathematical expression. However, it can be simplified as a diffusion model ( approximation) and approximation. is considered more accurate than the diffuse approximation and, therefore, has recently become a major direction of research. Boas et al.8 established a frequency-dependent solution for the equation in the media of infinite thickness. Hull and Foster9 obtained the Green’s function of the steady-state radiation transport equation in the equation. They proved that the model had better accuracy than the diffusion theory in the high absorption medium or near the source of the radiation. Dickey et al.10 pointed out that the radiance and the fluence expressions based on the equation were used to optically characterize an intralipid tissue phantom, varying in concentration of the scatterer and absorber. Chai et al.11 established the equation of the spherical harmonics method. Gao et al.12 derived the equation set of the equation and its Green’s function approximate solution. Further, a complete solution of the equation with the extrapolated boundary conditions was also obtained. However, this progression was the continuation of the theory first proposed by Boas et al.,8 and the solutions of the equation had already been derived by Liemert and Kienle13 in a semi-infinite medium illuminated by a collimated beam of light. The solution was evaluated with nearly the same computational effort as needed for solving the standard diffusion equation. For this purpose, the method used a combination of the classic spherical harmonics and rotated reference frames for solving the equations in closed form. In this paper, a similar method as that of the diffusion equation is used. The steady-state solutions of the Green’s function by the equation in the semi-infinite media under extrapolation boundary condition are estimated, the proposed solution is a diffusion-based model. Further, two frequency-domain solutions are estimated and transformed into the time domain by Fourier transform. The spatially resolved reflectance and the time-resolved reflectance of the equation are calculated, and the Monte Carlo simulation is used to verify the equation. The equation and the diffusion equation are compared. 2.Theory2.1.Solution of the Green’s Function of P3 Equation in the Steady StateWe present the solution of the Green’s function of the approximate steady-state equation in the infinite medium (see Fig. 1). We assume that an infinitely thin beam is vertically incident on the medium. All the photons undergo first scatter at depth of one transport mean free path. An isotropic photon source is established at that point, and therefore the beam is scattered isotropically at a depth , where and are the deduced scattering coefficients and the absorption coefficients, , and are the scattering coefficients and anisotropy factor. The origin of the coordinate system is a point where the beam enters the turbid medium, and coordinate has the same direction as the incident beam. The and coordinates lie on the surface of the turbid sample and . If we use the Henyey–Greenstein phase function, all its higher-order moments can be determined.9 We assume Some scholars1,3,14,15 used inverse Fourier transform to solve the diffusion equation, that is, the two solutions () and () of the diffusion equation were obtained. On the basis of the diffuse model, we obtain the four solutions , , and of the equation by the inverse Fourier transform in the same way as the diffusion equation. The four solutions of the equation of Green’s functions are calculated by Eq. (2) as follows: Equation (2) is the inverse Fourier transform. is the result of the Fourier transform that is written in the form of exponentials when . The specific expression is shown by In Eq. (3), and are expressed as where , and are given by In Eq. (5), where the constant items , and in Eq. (2) can be expressed as follows: In Eq. (4), if , Eq. (4) is expressed as and . Thus, and in Eq. (8) can be expressed as shown in If the refractive index of the biological medium is 1.4, . Whereas, if the refractive index is 1.33, . was calculated according to Haskell et al.2 We present four solutions of the equation in the steady-state equation.To calculate the equation, we present the expression of the reflectance of the equation, which is also a continuation of the diffusion equation. The solution of the equation steady-state spatially resolved reflectance is shown by Eq. (10) where the coefficients , and are the constants associated with the refractive index. If the refractive index , then , , , and . If , then , , , and .2.2.Solution of the Green’s Function of P3 Equation in the Time DomainThe basic idea of the time-domain method is to first establish the frequency-domain equation, through the Fourier transform, and then convert it to the time-domain equation. Here, we present two methods in the time domain. The first method, we call the optical parameter method. If we use Eq. (11) instead of Eq. (1), we can get the equation in the frequency domain where is an angular frequency of the incident light and is the speed at which light travels in the medium. Equations (2)–(10) can still be applied to the frequency-domain equation.The second method we call the double-diffusion coefficient method. In Eqs. (1)–(10), only Eq. (6) is modified by Eq. (12), Equation (13) is used in the time-domain and frequency-domain methods of the diffusion equation where is the coefficient of the diffusion equation. To correspond to the diffuse equation, we adopt . Since the new parameter is similar to the diffusion coefficient, we name this approximation method a double-diffusion coefficient method.We estimate different frequencies and then carry out Fourier transform for the solution in the time domain. 2.3.Relationship Between P3 Equation and Related TheoryWe compared the relationship between the model and other related models. Kienle et al.14 used the Fourier approach to solve the diffusion equation for a two-layered turbid medium. They then used this method to publish a series of articles including the study,15 where Eq. (3) was widely used. We also used this method to solve the diffusion equation 3 and establish the equation. The spatially resolved reflectance as utilized in the diffusion equation is given by The refractive index of the biological tissue is given by 1.4 and constants , . From the diffusion equation, we have . Using these values and expression, Eq. (14) takes the form, . This reflectance represents the first two terms of Eq. (10). Hull and Foster9 resolved a solution of the spatially resolved reflectance. However, the expression was more complex and completely different from the one we presented. Gao et al.12 also employed the equation of reflectance given by Hull and Foster9The literature9 presents yet another expression that is different from Eq. (10) in our study. The in the literature9 depends on the numerical aperture of the detector and the refractive-index mismatch. The relationship between Eq. (10) and the literature9 is expressed by where is the coefficient in Eq. (10) and is the coefficient in Ref. 9.In the diffuse equation, is the diffusion coefficient. Equation (6) not only has the same definition but also a further definition, in order to distinguish the diffusion coefficient, we named . Hull and Foster9 presented the specific expression of , which was also used by Gao et al.12 Boas et al.8 established a specific expression of in the frequency domain. In the limit, when the frequency domain is approaching 0, the result8 becomes similar to that of Hull and Foster9 This is expressed as follows: Equations (4) and (5), in the limit, , reduce to Eq. (16). This shows that Eq. (16) has a special form of our model.Equation (9) is called the extrapolated boundary condition. The extrapolated boundary conditions for the equation in Refs. 9 and 12 can be modified to Eq. (17) This is consistent with our Eq. (12). Hull and Foster9 pointed out the drawback of extrapolated boundary conditions. The drawback is that it is does not rigorously satisfy boundary conditions for the higher-order terms. However, it can be demonstrated that this approach yields excellent agreement with experimental data, suggesting that the boundary condition is not especially sensitive to these terms. The solution of the Green’s function by the equation is similar to the diffusion equation. The model is an extension of the diffusion equation. Our model has some similarities to the published equation,9,12 but they have essential differences. The same similarities are the extrapolation boundary condition [Eq. (17)], the interrelated coefficients [Eq. (15)]. The differences are that: (a) the solution of Hull and Foster9 is obtained by polynomial approximation, and our equation is obtained by strict Fourier transform and inverse transformation, (b) the solution of Hull is derived assuming the approximation of a spherically symmetric medium. Our solution is derived assuming the approximation of a nonspherically symmetric medium: (c) the calculated reflectance equation is completely different between our equation and the equation of Hull; (d) using our equations, we can derive the characteristic parameters of Hull and Foster; and (e) our equation is established on the basis of the diffusion equation, which is the continuation of the diffusion equation. 3.Results3.1.Comparison of P3 Equation with Monte Carlo Simulation and Diffusion Equation in the Steady StateMonte Carlo simulation is used as a universal means of verification. The diffusion equation is often used in biomedical optics.3,16 We use the Monte Carlo simulation17 to verify our model and compare the accuracy with the diffuse model. The Monte Carlo simulation uses a pencil photon that was normally incident onto the semi-infinite media, and use the Henyey–Greenstein phase function. The reflectance of Monte Carlo simulations depicts the results of Monte Carlo simulations of isotropic point sources in semi-infinite media. First, we compare the model with the diffusion equation of the semi-infinite medium using the Monte Carlo simulation. We use the diffuse equation method of the paper.3,16 If the detection distance is 0.5 to 30 mm, it is assumed that the scattering coefficient , , and the absorption coefficients are and . The results are shown in Fig. 2(a). The straight part is the equation result, and the square part is the Monte Carlo simulation result. From Fig. 2(a), we can see that the equation result is almost the same as the Monte Carlo simulation. We compare the relative error (unit: %) among the equation, the Monte Carlo simulation, and the diffuse equation. As shown in Figs. 2(b) and 2(c), the relative error is larger except for the front 1 mm. The relative error of the results is very small. In other words, the equation is very consistent with the Monte Carlo simulation results. At the same time, in addition to the near light source, the relative error between the equation and the diffuse approximation is close to zero, that is, the equation has the same result as the diffusion equation. It is assumed that the scattering coefficient , , and the absorption coefficients are and . The results are shown in Fig. 2(d), and the relative errors are shown in Figs. 2(e) and 2(f). These confirm the previous conclusions. If the detection distance is 0.05 to 3 mm, it is assumed that the scattering coefficient , , and the absorption coefficient is . The results are shown in Fig. 3(a). The straight line is the equation result, the dotted line is the Monte Carlo simulation result. From Fig. 3(a), we can see that when the detection distance is greater than a certain value, the result of the equation is almost the same as the Monte Carlo simulation. We compare the relative error (unit: %) among the equation, the Monte Carlo simulation, and the diffuse equation. As shown in Fig. 3(b), when the detection distance is between 0.6 and 1.5 mm, the relative error of is small, and when the detection distance is greater than 1.5 mm, the equation has the same result as the diffuse equation. That is, when the detection distance is greater than 0.6 mm, the equation is more accurate. It is assumed that the scattering coefficient , , and the absorption coefficients are . The results are shown in Fig. 3(c), and the relative error is shown in Fig. 3(d), confirming our conclusions. The diffusion equation is usually used under the condition that the absorption coefficient is small. In the same parameters of Fig. 2(a), we increase the absorption coefficient. The specific parameters are , , and . We compare the equation, diffuse approximation, and Monte Carlo simulation when the distance of the detector is 0.05 to 3 mm. The result is shown in Fig. 4(a). Figure 4(b) shows the relative error between Monte Carlo simulation, equation, and the diffusion approximation. Figure 4(b) shows that when the detection distance is more than 2 mm, has a small error. When the detection distance is more than 2.5 mm, the relative error of is less than 4%, and the error of diffusion is greater than 8%. This shows that is more accurate. 3.2.Comparison of the P3 Equation with Monte Carlo Simulation in Time DomainFirst, the double-diffusion coefficient method of the equation in the time domain is compared with Monte Carlo simulation. Since the double-diffusion coefficient corresponds to the diffusion equation, we add a based on the diffusion equation, so we first verify this method. Assuming that the three semi-infinite dielectrics have a refractive index of 1.4 and their parameters are , , and , , and , we set two detection distances of 5.5 and 8.5 mm, respectively. The results are shown in Figs. 5(a) and 5(b), where the detection distances are 5.5 and 8.5 mm, respectively. It can be seen from Figs. 5(a) and 5(b) that the time-resolved reflectance calculated by the double-diffusion coefficient method is consistent with the Monte Carlo simulation, and it is proved that the time-resolved reflectance calculated by the double-diffusion coefficient method is correct. To further verify the correctness of the double-diffusion coefficient method, we compared the three media with a refractive index of 1.33. The parameters are , , respectively. The absorption coefficients are , , and , respectively. The detection distance is 8.5 mm. The results are shown in Fig. 5(c). It can be seen from Fig. 5(c) that the medium with a refractive index of 1.33 is consistent with the Monte Carlo simulation, further proving that the time-domain method of the double-diffusion coefficient method is correct. We have shown that the time-resolved reflectance calculated by the double-diffusive coefficient method is consistent with the Monte Carlo simulation. We have also verified that the time-resolved reflectance using the optical parametric method is correct. We could have first used the Monte Carlo simulation to verify it, but we found that the results are the same as in Fig. 5. We further compare the two time-domain methods. The parameters in Fig. 6 are the same as the first two parameters in Fig. 5(c), using the same detection distance, we can see that the image is completely coincident, without any difference. This proves that the optical parameter method is accurate. Because the optical parameter method does not have any approximation, it is more accurate than the double-diffusion coefficient method which is the solution corresponding to the diffusion model. In order to compare the differences between the two methods, we calculate the relative error. In time-resolved reflectance, the relative error is usually calculated using the logarithmic relative error. The relative error is given as follows: where log represents the time-resolved reflectance calculated by the double-diffusion coefficient method and log represents the time-resolved reflectance calculated by the optical parametric method. The results are shown in Fig. 7(a). The results show that if the time exceeds 0.2 nm, the error is less than 0.3%.Further, we use the relative error of the steady state as follows: The relative error calculated by Eq. (19), as shown in Fig. 7(b), comes out to be about 4% when the time is greater than 0.2 ns. Therefore, the two time-domain methods are correct. The optical parameter method is better in terms of more accurate solution whereas the double-diffusion coefficient method excels in its simplicity.3.3.Comparison of the P3 Equation with the Diffuse Equation in Time DomainWe have not only verified the equation but also compared the time-domain equation of the equation and diffusion model. Since the two time-domain methods using the model and the time-domain method of the diffusion model are coincident, we simply compare the relative errors between the two models and the diffusion model. In Figs. 8(a) and 8(b), the relative error equation in the steady state is used. The parameters of Fig. 8(a) are refractive index , , , . The parameters of Fig. 8(b) are refractive index , , , . It can be seen from Fig. 8 that when the time exceeds 0.2 ns, the relative error between the optical parameter method and the diffuse equation is about 2%; the error between the double-diffusion coefficient method and the diffuse equation is very small. In fact, the double-diffusive coefficient method is almost identical to that estimated by the diffuse model. Since the optical parameter method is an accurate equation time-domain equation, the result is more accurate than the time-domain equation of the diffusion model. If the logarithmic relative error equation is used, it is also shown that the diffusion equation is relatively consistent with the double-diffusion coefficient method. Therefore, the accuracy of the diffusion equation is not as good as that of approximating the time-domain equation, but the difference is very small. This is shown in Fig. 9. The parameters are the same as in Fig. 8(a). 4.ConclusionThe diffusion equation becomes increasingly inaccurate with the increase of the absorption coefficient value. This justifies the need to establish a model such as the equation, which is more accurate than the diffusion equation. However, due to relative complexity, it is not easy to implement as the mathematical model. In this paper, the solution of the equation in the steady state under extrapolation boundary condition is presented in the form of Fourier transform and inverse transformation, the proposed solution is a diffusion-based model. The solution is consistent with the diffusion model. Monte Carlo simulation is used to demonstrate the accuracy of the model in the steady state. The relative errors of the equation, diffuse equation, and Monte Carlo simulation are calculated, which indicates that is more accurate. In the steady state, the equation is similar to the diffuse equation, and it is shown that the diffusion equation and equation have almost the same value under the condition of satisfying the diffusion equation. Monte Carlo simulation shows that when there is a large absorption coefficient, the relative error of the diffusion equation is quite large, whereas the relative error of the equation is comparatively small. Based on the steady-state equation of the equation, two kinds of frequency-domain equations are established: the optical parameter method and double-diffusion coefficient method, they are transformed into two time-domain equations by Fourier transform in the frequency domain. By utilizing Monte Carlo simulation, we have verified two different time-domain methods with different refractive indices and different parameters. The results show that our two methods are correct. The time-domain method of the diffusion equation is compared. The results show that the numerical and error values calculated by the double-diffusive coefficient method are consistent with the diffusion equation. Since the optical parameter method of the equation does not have any approximation under extrapolation boundary condition, its accuracy is higher than that of the double-diffusion coefficient method. Therefore, the accuracy of the optical parameter method is higher than that of the diffusion equation. However, the double-diffusion coefficient method is simpler, and the logarithmic relative error of the time domain is calculated at about 0.2%, so the error is almost negligible. Our research method can lay the foundation for the solution of the model. We estimate that the model needs to establish the cubic equation of the in Eq. (4) and the reconstruction of Eq. (7). Based on this paper, we can study the approximate multilayer model, which can replace the diffusion equation. 3,16,18 DisclosuresThe authors have no relevant financial interests in this article and no potential conflicts of interest to disclose. ReferencesA. Kienle et al.,
“Investigation of two layered turbid medium with time-resolved reflectance,”
Appl. Opt., 37
(28), 6852
–6861
(1998). http://dx.doi.org/10.1364/AO.37.006852 APOPAI 0003-6935 Google Scholar
R. C. Haskell et al.,
“Boundary conditions for the diffusion equation in radiative transfer,”
J. Opt. Soc. Am. A, 11
(10), 2727
–2741
(1994). http://dx.doi.org/10.1364/JOSAA.11.002727 Google Scholar
X. Wang and S. M. Wang,
“Light transport model in a n-layered mismatched tissue,”
Waves Random Complex Medium, 16
(2), 121
–135
(2006). http://dx.doi.org/10.1080/17455030600683321 Google Scholar
J. B. Fishkin et al.,
“Gigahertz photon density waves in a turbid medium: theory and experiments,”
Phys. Rev. E, 53
(3), 2307
–2319
(1996). http://dx.doi.org/10.1103/PhysRevE.53.2307 Google Scholar
V. Venugopalan, J. S. You and B. J. Tromberg,
“Radiative transport in the diffusion approximation: an extension for highly absorbing media and small source–detector separations,”
Phys. Rev. E, 58
(2), 2395
–2407
(1998). http://dx.doi.org/10.1103/PhysRevE.58.2395 Google Scholar
A. Kienle and M. S. Patterson,
“Determination of the optical properties of turbid media from a single Monte Carlo simulation,”
Phys. Med. Biol., 41 2221
–2227
(1996). http://dx.doi.org/10.1088/0031-9155/41/10/026 PHMBA7 0031-9155 Google Scholar
M. Jia et al.,
“Virtual-source diffusion approximation for enhanced near-field modeling of photon-migration in low-albedo medium,”
Opt. Express, 23
(2), 1337
–1353
(2015). http://dx.doi.org/10.1364/OE.23.001337 OPEXFF 1094-4087 Google Scholar
D.A. Boas et al.,
“Photon migration within the approximation,”
Proc. SPIE, 2389 240
(1995). http://dx.doi.org/10.1117/12.209973 PSISDG 0277-786X Google Scholar
E. L. Hull and T. H. Foster,
“Steady-state reflectance spectroscopy in the approximation,”
J. Opt. Soc. Am. A, 18
(3), 584
–599
(2001). http://dx.doi.org/10.1364/JOSAA.18.000584 JOAOD6 0740-3232 Google Scholar
D. J. Dickey, R. B. Moore and J. Tulip,
“Using radiance predicted by the approximation in a spherical geometry to predict tissue optical properties,”
Proc. SPIE, 4156 181
(2001). http://dx.doi.org/10.1117/12.413701 PSISDG 0277-786X Google Scholar
C. Chai, Q. Luo and Q. Liu,
“Analytical solution of approximation to radiative transfer equation for an infinite homogenous media and its validity,”
J. Mod. Opt., 55
(21), 3611
–3624
(2008). http://dx.doi.org/10.1080/09500340802372801 JMOPEW 0950-0340 Google Scholar
Z. Gao et al.,
“Study of spatially resolved diffuse reflection based on P3 approximation,”
Acta Opt. Sin., 26
(8), 1220
–1225
(2006). GUXUDC 0253-2239 Google Scholar
A. Liemert and A. Kienle,
“Explicit solutions of the radiative transport equation in the approximation,”
Med. Phys., 41
(11), 111916
(2014). http://dx.doi.org/10.1118/1.4898097 MPHYA6 0094-2405 Google Scholar
A. Kienle et al.,
“Noninvasive determination of the optical properties of two-layered turbid media,”
Appl. Opt., 37
(4), 779
–791
(1998). http://dx.doi.org/10.1364/AO.37.000779 APOPAI 0003-6935 Google Scholar
A. Liemert and A. Kienle,
“Light diffusion in a turbid cylinder. I. Homogeneous case,”
Opt. Express, 18
(9), 9456
–9473
(2010). http://dx.doi.org/10.1364/OE.18.009456 Google Scholar
A. Liemert and A. Kienle,
“Light diffusion in N-layered turbid media: steady-state domain,”
J. Biomed. Opt., 15
(2), 025003
(2010). http://dx.doi.org/10.1117/1.3368685 JBOPFO 1083-3668 Google Scholar
L. Wang, S. L. Jaques and L. Zheng,
“MCML—Monte Carlo modeling of light transport in multi-layered tissues,”
Comput. Methods Programs Biomed., 47
(2), 131
–146
(1995). http://dx.doi.org/10.1016/0169-2607(95)01640-F CMPBEK 0169-2607 Google Scholar
A. Liemert and A. Kienle,
“Light diffusion in N-layered turbid media: frequency and time domains,”
J. Biomed. Opt., 15
(2), 025002
(2010). http://dx.doi.org/10.1117/1.3368682 JBOPFO 1083-3668 Google Scholar
Biography |