Field of the invention

The invention relates to a method for determining the absorption coefficient of turbid media.

The invention further relates to a system for determining the absorption coefficient of turbid media.

The invention still further relates to a computer program product comprising instructions for causing a processor to determine the absorption coefficient of turbid media. Background of the invention

Non invasive measurement of the concentrations of different absorbing substances in optically turbid media, such as living tissue, is challenging. A commonly used method for this employs the measurement of the reflected light from such an optically turbid medium. An essential part of the light reflected from a turbid medium has travelled through the medium and was directed out of the medium by scattering. The main problem of such measurements is that the optical path length of the detected photons is strongly dependent on the optical properties, such as the absorption coefficient, the scattering coefficient and the angular distribution of scattering, also referred to as scattering phase function. As a result, the path length of detected photons is dependent on the measurement geometry and optical properties, and varies with wavelength. Absolute measurements of

concentrations based on absorption spectroscopy in turbid media may be compromised by the dependence of the path length on the properties of a medium under consideration.

Classical reflectance spectroscopy devices known from the prior art often utilized multiple optical fibers to deliver and collect light during measurement. However, the potential advantages of reflectance probes with a single optical fiber to deliver/collect light are numerous. Advantages of the single fiber design include small probe size and simple device design, making it more-suitable than multi-fiber probes for clinical applications, such as optical biopsy of potential malignancies via endoscopy or biopsy needles. However, there exists no empirical or analytical description of light transport in the regime associated with overlapping source-detector areas, such as when using a single fiber.

An embodiment of a reflectance spectroscopy system using overlapping illumination-detection areas for determining the absorption coefficient in a turbid medium is known from Kanick et al. Phys. Biol. 54, 6991-7008 (2009). In the known embodiment a method is disclosed wherein a single fiber is used and positioned at a surface of the tissue under

investigation. The fiber is used for illuminating the tissue as well as for collecting the reflected light.

It is a disadvantage of the known method that the dependence of the effective path length of photons on scattering phase function as well as the reduced scattering coefficients were guessed, which might lead to inaccurate determination of the absorption coefficient.

Brief description of the invention

It is an object of the invention to provide a method for determining the absorption coefficient of a turbid medium without knowledge of the scattering coefficient and scattering phase function when overlapping illumination and detection areas are used.

To this end the method according to the invention comprises the steps of:

retrieving a calibration spectrum (CA) from a reference measurement using a reference sample;

carrying out a measurement on an actual sample for

determining the absolute reflection spectrum (R _a bs) using a raw spectrum measured on the sample (Smedmm) and the calibration spectrum (CA);

using the absolute reflection spectrum (R _a bs) for determining the wavelength dependent absorption coefficient by minimizing the difference between the measured absolute reflection spectrum (R _a bs) and a model function (R _a b _s ^model ), wherein the model function (R _a b _s ^model ) is being modelled using a predetermined equation based on prior knowledge of the

combination of

i. a dependence of the effective photon path length (LPF) on scattering phase function (PF);

ii. a dependence of the absolute reflectance in the absence of absorption (R _a bs°) on scattering phase function (PF).

It will be appreciated that a plurality of different per se known embodiments may be used for providing such overlapping illumination- detection geometry. In a preferred embodiment a single optical fiber is used, for example having dimensions between 10 pm and 3 mm.

In accordance with the invention a calibrated assessment of reflectance in the absence of absorption is used to appropriately estimate the combined effect that the reduced scattering coefficient and the scattering phase function have on the effective photon path length. Thereby estimation of the absorption coefficient is substantially improved.

Application of this methodology to measurement of living tissue provides aspects of vascular physiology which may be useful in

characterization of tissue health status. For example, blood volume fraction, average vessel diameter and haemoglobin oxygen saturation, as well as concentrations of other light absorbing substances, including billirubin, beta- carotene, melanin, glucose, fat and water may be determined. In addition, the method can be used to measure concentrations of exogenous substances in tissue, such as drugs, optical contrast agents, dyes, pollutants, as long as hey have appropriate absorption properties in the wavelength region used.

The invention is based on the following insights. White-light reflectance measurements provide information about absorption and scattering properties of an optically sampled turbid medium such as tissue. Specifically the absorption coefficient μ ₉ relates to aspects of the tissue physiology. It is found that quantitative estimation of μ ₉ from a reflectance spectrum requires mathematical correction for the effects that μ ₉ , reduced scattering coefficient s and scattering phase function PF have on the effective photon path length LSF. An example of a mathematical representation of this relationship is given by equation (1):

_T model

^SF _ ^PFPI

d fiber (Vs ^d fiber Y" (Pi + (i fiber ) ^¾ ) wherein

CPF describes the dependence of LSF on PF;

dfiber is a diameter of the fiber which is used for measurements.

It will be appreciated that the values 1.54, 0.18 and 0.64 correspond to the empirically established constants p i, p2, p3, respectively. These coefficients were established for the single fiber embodiment as reported in Phys. Biol. 54,

6991-7008 (2009), and may have different values for different conditions and or embodiments.

In the methods known from the prior art, practical application of equation (1) to analyze spectra measured in tissue in vivo has required an assumption about the tissue PF to estimate CPF, as well as an assumption about the value of p _s ' at least at one wavelength. This approach is found to be not accurate.

It is further found that the single fiber reflectance intensity in the absence of absorption R _a bs° showed a PF-specific dependence on dimensionless scattering, defined as the product of p _s ' and fiber diameter dfiber. However, it will be appreciated that this finding may be generalized to any overlapping illumination-detection geometry. An example of a mathematical representation of the relation between R _a bs° and dimensionless scattering is given in equation (2): wherein,

q _c is the asymptotic value, i.e. the diffuse limit to the single fiber collection efficiency, which is proportional to the NA of the fiber and is about 2.7% for a single fiber with NA=0.22. P ₄ , P5, ΡΘ are PF-specific parameters. It is found that for the single fiber embodiment, in equation (2) P5 usually falls within the range of 4.3 - 9.2; Ρβ usually falls within the range of 0.81 - 1.14 and P ₄ usually falls in the range of 1.07 - 2.16. It will be appreciated that, although, not specified in equations (1) and (2) explicitly, CPF, p _s , Rabs are variables which depend on the scattering phase function (PF). Therefore, the effective photon path length (LPF) and the absolute reflectance in the absence of absorption (Rabs °) are dependent on the scattering phase function (PF). In accordance with an aspect of the invention, the model function (R _a bs ^model ) is modeled using inter alia a prior knowledge on the dependence of LPF and R _a bs ⁰ on the scattering phase function (PF).

Accordingly, in accordance with the insight of the invention, first a reference calibration measurement is carried out. A sample having high scattering coefficient (such that p _s 'dfiber>10) can be selected, because for very high scattering coefficients the collected reflectance becomes independent of the (often unknown) phase function of the calibration sample and approaches the diffuse limit q _c . Alternatively, a sample with a smaller scattering coefficient may be used if its phase function is known. The measurement may be performed with the fiber in contact with the calibration sample. However, other calibration geometries may be used. The absolute device calibration spectrum in case a high scattering reference sample is used can be calculated from the calibration measurement as follows: ) = β » (3)

reference \ )

wherein

C(A) stands for the cahbration spectrum of the measurement device using the calibration sample;

Sreference( ) stands for the raw, unprocessed spectrum measured on the calibration sample,

q _c (NA(A)) stands for the maximum reflection of a scattering sample. It is further found that q _c (NA(A)) may depend slightly on the wavelength if the scattering coefficient of the calibration sample is wavelength dependent.

When the results of the calibration measurements are processed they are further used in the method of the invention in the following way.

The absolute reflection spectrum of a sample under investigation (tissue) may be obtained using the calibration data as follows:

(4) wherein

R _a bs( ) stands for the absolute reflection of the medium

Smedium(A) stands for the raw, unprocessed spectrum of the actual sample (tissue).

At the next step in accordance with the invention, the optical properties are extracted from the measured spectrum R _a b _s ( ). It is appreciated that a general problem in analysing such spectra is that three unknown parameters (reduced scattering coefficient p _s ', scattering phase function PF and absorption coefficient μ ₉ ) for each measurement point have to be

calculated. As a result, the equations do not converge to a single solution.

In accordance with the invention the wavelength dependent absorption coefficient μ ₉ (λ) is calculated from the measured reflectance R _a bs(A) by minimizing the difference between the measured absolute reflection spectrum R _a bs( ) and the model function R _a b _s ^model (^), wherein the model function R _a b _s ^model (^) is modelled using a pre-determined equation based on prior knowledge of the combination of the dependence of the effective photon path length LsFmodei( ) on the phase function PF (e.g. equation 1) and the dependence of the absolute reflectance in the absence of absorption R _a b _s °( ) on the phase function PF (e.g. equation 2). The may be modelled using the Lambert-Beer equation, according to R _a bs ^model G =R _a bs°(A)e(- ^Da ( ^A > ^L sFmodei ^(A)) .

Accordingly, in accordance with the invention, in Equation (2) _s ' is estimated from R _a bs° such that the effect of a potential mis-estimation of _s ' in Equation (1) is compensated by a corresponding mis-estimation of CPF. In this way the effective path-length is close to its true value (within 7.5% for biological tissues), even when CPF and _s ' are incorrectly specified. Preferably, the values for CPF, P4, P5 and Ρβ are chosen to be 0.944, 1.55, 6.82, and 0.969, respectively; this choice of parameters minimizes the error in estimated path length LsFmodei(A).

It will be appreciated, however, that the compensative effect occurs for other combined values of CPF, P4, P5 and Ρβ as well. Furthermore, different mathematical expressions than shown in Eqs. (1) and (2) may also be used to describe the combined effect of phase function on photon path length LsFmodei( ) and on absolute reflectance R _a bs°( ). Moreover, lookup tables that directly link R _a bs°( ) to a combined CPF-U _s ' set can be used as well. This named

compensative effect in the mis-estimation of the core parameters in the equations is found to be surprising, however enabling to solve a single equation having three unknowns. More details on the named compensative effect will be given with reference to Figure 2.

In an embodiment of the method according to the invention the method further comprises the step of using a single fiber for delivering the light beam towards the sample and for collecting the reflected beam from the sample.

It is found that such solution may be practical for clinical purposes as both the impinging and the reflected beams may be delivered by the same fiber, allowing for small fiber-probe profiles and facilitating measurements through thin needles such as Fine Needle Aspiration needles.

In a further embodiment of the method according to the invention the light used for the cahbration and sample measurements is generated by a plurality of monochromatic sources. However, it will be appreciated that a source having a continuous spectrum of wavelengths may also be used.

The system according to the invention for determining the wavelength dependent absorption coefficient of a diffuse medium for a light beam comprises:

a light source adapted to generate the light beam;

a processor adapted for:

a. retrieving a calibration spectrum (CA) from a reference measurement using a reference sample;

b. retrieving results of a further measurement on an actual sample for determining the absolute reflection spectrum (Rabs) using a raw spectrum measured on the sample (Smedium) and the calibration spectrum (CA);

c. using the absolute reflection spectrum (R _a bs) for

determining the wavelength dependent absorption coefficient by minimizing the difference between the measured absolute reflection spectrum (R _a bs) and a model function (R _a b _s ^model ), wherein the model function (R _a b _s ^model ) is modelled using a pre-determined equation based on prior knowledge of the combination of:

i. a dependence of the effective photon path length (LSF) on scattering phase function (PF);

ii. a dependence of the absolute reflectance in the absence of absorption (R _a bs°) on scattering phase function (PF).

Advantageous embodiments of the system according to the invention are given in the dependent claims.

The computer program according to the invention comprises instructions for causing a processor to carry out the following steps:

a. retrieving a calibration spectrum (CA) from a reference measurement using a reference sample;

b. retrieving data of a measurement on an actual sample for

determining the absolute reflection spectrum (R _a bs) using a raw spectrum measured on the sample (Smedmm) and the calibration spectrum (CA);

c. using the absolute reflection spectrum (R _a bs) for determining the wavelength dependent absorption coefficient by minimizing the difference between the measured absolute reflection spectrum (R _a bs) and a model function (R _a b _s ^model ), wherein d. the model function (R _a b _s ^model ) is modelled using a predetermined equation based on prior knowledge of the combination of

i. a dependence of the effective photon path length (LSF) on scattering phase function (PF);

ii. a dependence of the absolute reflectance in the absence of absorption (R _a bs°) on scattering phase function (PF).

These and other aspects of the invention will be discussed in more detail with reference to figures wherein like reference numerals refer to like elements. It will be appreciated that the figures are presented for illustrative purposes and may not be used for limiting the scope of the appended claims.

Brief description of the figures

Figure 1 presents in a schematic way an embodiment of a system which may be used for carrying out a calibration measurement.

Figure 2 presents a number of characteristic curves.

Detailed description

Figure 1 presents in a schematic way an embodiment of a system which may be used for carrying out a calibration measurement on a sample 8. For this purpose the system 10 comprises a probe 2, a bifurcated optical cable 4, one end of which is connected to a light source 6 and the other end of which is connected to a suitable spectrometer 8. Accordingly, the optical fiber 2 is used for delivering a light beam from the light source 6 to the sample and for collecting reflected light from the sample 8.

It is further found to be advantageous to polish the probe 2 at an angle larger than arcsin(NA/n _sam pie) with respect to a vertical line for

minimizing specular reflections, where NA is the numerical aperture of the fiber and nmedium IS the refractive index of the sample.

The reflectance of the sample is in case of a high scattering sample given by

Rsample =q _c (NA(A)).

When the absorption coefficient of a turbid medium (tissue) is to be determined, the equation 1 has to be used in a Lambert-Beer equation, according to In a general way, the equation 1 can be written as: As has been indicated earlier, in equation (5) PF and p _s ' of tissue are not known, which implies that CPF is not known and that specification of p _s ' from reflectance R _a bs° also requires knowledge of PF for specifying the correct constants P ₄ , P5, and Ρβ in equation 2.

In accordance with the invention, _s ' is estimated from reflectance

Rabs ⁰ such that a potential mis-estimation of p _s ' is compensated by a

corresponding mis-estimation of CPF.

It is found that the ratio of CPF/( μ _δ ' ) ^p2 is approximately equal to its true value (within 7.5% for biological tissues), provided the CPF is properly linked to the phase function used to estimate _s ' from R _a bs° (i.e. CPF is linked to the values of P ₄ , P5, and Ρβ in equation 2).

It is found that as high angle scattering events become more likely, Rabs ⁰ increases because incident photons are more likely to be collected and the photon path length LSF decreases as those collected photons are hkely to travel a shorter path.

In Figure 2, a number of characteristic curves is presented, wherein it is found that the curve for CPF=0.944, P ₄ =1.55, P ₅ =6.82 and P ₆ =0.969 is an optimal curve for practicing the invention. Figure 2 shows the relation between Rabs ⁰ and p _s 'dfiber for 3 exemplary embodiments of known samples having different known phase functions PF with different backscattering components. The CPF values and P ₄ , P5, and Ρβ values for these phase functions PF are also indicated for each of these selected PF's. The graphs presented in Figure 2 are calculated using equation (2) discussed with reference to the foregoing.

For an unknown sample, such as tissue, utilization of equations (1) and (2) to calculate the photon path length LSF requires an assumption about the phase function PF, which is also unknown.

It is found that it is particularly suitable to assume that the phase function PF is characterized by Cp _F =0.944, P ₄ =1.55, P ₅ =6.82 and P ₆ =0.969 (see solid line, curve 1 in figure 2). However, it is also possible to implement the equation using the other curves given in Figure 2, or any other alternative curves which may be produced using equation 2 or using Monte Carlo simulations applied to a known sample. However, it is found advantageous to select a curve whose reflectance properties are close to the reflectance properties which may be expected from a sample under investigation, such as tissue.

Utilization of the P ₄ , P5 and Ρβ, discussed with reference to the foregoing regarding an assumed phase function PF in equation (2)

corresponding to curve number 1 in figure 2, would yield an estimated value of s'(Est) for an unknown sample based on a measured reflectance R _a bs°(Real).

The following effect has been found when analyzing equations (1),

(2) and the graphs given in Figure 2. Should the true sample phase function PF of an unknown sample be characterized by a higher backscattering component than the assumed PF (e.g. the true combination of parameters corresponding to true curve 2 in figure 2 is CPF=0.86, P ₄ =1.24, Ρδ=4.47 and then the p _s ' would be overestimated from R _a b _s °(Real) since p _s '(Est) > s'(Real), see corresponding notations in Figure 2.

However, the initially assumed CPF (0.944), corresponding to the assumed sample curve 1 is larger than the true" CPF (0.86), corresponding to the "true" sample curve 2. Accordingly, an over-estimation of CPF compensates for the effect of over-estimation of p _s ' on LSF in Eq. (1).

Next, if the true phase function PF has in fact a smaller

backscattering component than the value assumed for the phase function PF (e.g. the true sample PF corresponds to curve 3 in figure 2, instead of the assumed curve 2), then the resulting _s '(Est) obtained by using the assumed Rabs ⁰ curve 2 of figure 2 would be underestimated. Accordingly, also for this situation an effect of compensation takes place - i.e. an under-estimation of CPF in Eq. (1), since in this case the assumed CPF (0.944) was smaller than the true" CPF (1.0).

Preferably, for the assumed scattering phase function PF assumed _a gamma value between γ = 1.6 and 1.8 is used, where gamma is related to the first and second moments (gi and g2, respectively) of the scattering phase function according to γ = (l-g2) / (1-gi).

The inter-related, compensating effects of mis-estimation of CPF and s' through assumption of an estimated phase function PF can be further analysed by evaluation of the ratios CpF ^est /CpF ^real and (u _s '(Est)/p _s '(Real)) ^{0 18} .

It is found that these two metrics either both are smaller than unity or both are greater than unity, indicating a compensating effect on estimates of LSF. Moreover, the magnitudes of these effects are also very similar:

CpF ^est /CpF ^real ranges from 0.9 to 1.12 in biological tissues, while

( s'(Est)/ _s '(Real))° ¹⁸ ranges from 0.85 to 1.25 in case equation (2) is used to calculate p _s '(Est) from R _a bs°.

The inset (I) in figure 2 shows a histogram plot of the ratio of these 2 metrics (defined as the compensation factor). It will be appreciated that perfect compensation of the effect of mis-estimation of CPF and _s ' on path length would yield a compensation factor of 1.0. The histogram clearly shows a narrow distribution centred around 1.0, with 76% of the data within 5% of this value, and 99% of the data within 10% of this value.

It will be appreciated that while specific embodiments of the invention have been described above, the invention may be practiced otherwise than as described. For example, for specific turbid media different constants in the equations may be used. However, the method for determining the appropriate constants will lie within the ordinary skill of the person skilled in the art, when reducing the invention into practice.