University of Missouri, Department of Biological Engineering, Columbia, MO 65211, USA

University of Georgia, Faculty of Engineering, Athens, GA 30602, USA

Abstract

Background

Confocal microscopy has become an important option for examining tissues

Methods

We built a fully automated confocal scattered-light scanner to examine how light scatters in Intralipid, a common tissue phantom, and three-dimensional collagen gels. Intralipid with 0.5%, 1.0%, 1.5%, and 2.0% concentration was filled between precisely spaced glass coverslips. Collagen gels at collagen concentrations from 0.30 mg/mL to 3.30 mg/mL were prepared, and all samples underwent A-mode scanning with multiple averaged scans. In Intralipid samples, light reflected from the upper fluid-glass interface was measured. In collagen gels, average scattering intensity inside the actual gel was measured. In both cases, intensity was correlated with concentration.

Results

By measuring light attenuation at interface reflections of various thicknesses using our device, we were able to determine that the scattering coefficient at 660 nm of Intralipid at increasing concentrations in water to be 39 cm^{-1 }for each percent increase of Intralipid. We were also able to measure the amount of scattering of various concentrations of collagen in gels directly using backscattered light. The results show a highly linear relationship with an increase of 8.2 arbitrary units in backscattering intensity for every 1 mg increase of collagen within a 1 mL gel volume.

Conclusion

The confocal scattered-light scanner allows to accurately quantify scattering in Intralipid and collagen gels. Furthermore, a linear relationship between collagen concentration and intensity was found. Confocal scattered-light scanning therefore promises to allow imaging of collagen content in soft tissue layers.

Background

Confocal microscopy is most widely used to image fluorescence - either intrinsic or extrinsic - of an object, such as biological tissue. Different information can be obtained from the tissue under examination when the light scattering properties are examined. The hypothesis of this study is that light scattering quantitatively depends on collagen content and that scattered-light confocal microscopy can be used to determine collagen content in tissues.

Being able to analyze collagen content in tissue is important for several fields of research, including tissue physiology and tissue engineering. Collagen plays a key role in soft tissue repair

Collagen is the primary source of light scattering contrast in the visible range of light, and some studies have examined epithelial scattering coefficient with confocal microscopy

Methods

Our scattered-light confocal scanner was based on to the principle of single-pinhole confocal microscopy and shown in Figure

Schematic of the confocal scanner

**Schematic of the confocal scanner**. The inset shows the FC/APC (fiberoptic coupling with angle-polished coating) connection used to minimize reflection at the fiber-air interface.

A single axial scan (A-mode scan) was obtained through the raising of the vertical translation stage towards the sample, changing the focal plane of the objective. Backscattered light from the sample was detected by the PMT after traveling through the fiberoptic coupler. A 10-bit analog to digital converter processed the PMT signal which was displayed as intensity as a function of linear distance. A B-mode scan (multiple A-mode scans along a linear path) could also be obtained using the X-Y positioning capabilities of the instrument.

Sufficient signal-to-background ratio (S/B) is necessary for identification of a source of contrast, such as a reflective interface. S/B is limited as light is focused deep within scattering media since light attenuation is governed by Beer's law. In this first part of the study, the primary source of contrast is refractive index change, caused by reflection between two interfaces. In order to ensure that our signal would not be overly attenuated at the imaging depths relevant to this study, we first computed the maximum depth that we would be able to resolve a refractive index change of Δn = 0.22 (water - glass). The computation was based on Beer's Law under the assumption that attenuation by absorption _{a }is negligible against attenuation by scattering _{s}. The absorption coefficient _{a }in Intralipid at 660 nm is approximately 0.002 cm^{-1 }_{s }varies between 50 and 92 cm^{-1 }

where ** I **is the intensity of light after distance

in which _{ref }is the intensity of the reflected light and _{1 }and _{2 }are the refractive indices of the two media at the interface. If we consider a four-layer system consisting of air, the lower glass plate of the container holding the media, the scattering media itself, and the top glass layer with variable refractive index, we can predict the maximum imaging depth _{PD }as a function of scattering coefficient and refractive indices of each medium from manipulation of Equations 1 and 2. This yields the following equation

where the limits of detection are given as ratio _{d}/_{p }(ratio of reflected light from the upper media-to-glass interface relative to the reflected light from the lower glass-to-media interface), and the refractive indices are given as _{t }for the scattering media, _{g }for the lower glass plate of the container, _{i }and for the detecting interface, that is, the upper glass plate. These values are referred to in Figure _{i }= _{g}, Equation 3 simplifies to Equation 4:

A-mode scan of Intralipid

**A-mode scan of Intralipid**. The intensity _{d }of the peak resulting from the refractive index change from Intralipid to Glass is recorded and evaluated relative to the first peak _{p }that indicates the glass-to-Intralipid interface. Actual thickness

In our calculations, we define maximum imaging depth as the point where _{d }is 2% of _{p}. The value of 2% is an estimate, chosen based on our experience with the confocal device so that the intensity would be sufficiently above baseline and noise to be recognized as a peak. Under this assumption, we computed the maximum imaging depth as a function of scattering coefficient of the media at different values for refractive index change Δn (_{t }to _{i}), with the results shown in Figure

Theoretical maximum imaging depth due to scattering coefficient at an assumed minimum ratio of 2% for I_{d}/I_{p }(Equation 3)

**Theoretical maximum imaging depth due to scattering coefficient at an assumed minimum ratio of 2% for I _{d}/I_{p }(Equation 3)**. The maximum depth to detect refractive index changes of 0.05, 0.10, 0.15, and 0.20 are shown.

For the first part of this study, 20% Intralipid (Sigma Aldrich) was diluted to concentrations of 0.5%, 1.0%, 1.5%, and 2.0% in ultrapure water. Small volumes of each Intralipid concentration were placed between two glass slides separated by shim spacers of known thicknesses of 127, 191, 254, 318, 381, and 508

For the second part of the study, we used collagen gels in order to determine the effect of collagen scattering directly, rather than through refractive index changes. Gels were fabricated using high concentration Type 1 (8.00 mg/mL) rat tail collagen (BD Biosciences, Bedford, MA). Twenty-one concentrations were made ranging from 0.30 mg/mL to 3.30 mg/mL at 0.15 mg/mL intervals. Volumes for each desired concentration were mixed into test tubes (Fisher Scientific) which were kept on ice, where volume of collagen was determined as the ratio of the product of desired original volume of solution and final desired concentration to the concentration of the stock volume. 10% of the desired original volume consisted of 10× phosphate buffer solution. 1 N NaOH (2.3

Results

A typical A-mode scan can be seen in Figure _{p }and _{d }as defined in Equation 3. In this sample scan, _{d }is 9.2% of _{p}, which is above the detection limit. In between the two peaks, scattered light causes a measurable signal. A comparison of this signal with a scan of a non-scattering medium (background scan) yields a signal-to-background ratio of more than 24 dB. This ratio suggests that the assumed minimum ratio of 2% for _{d}/_{p }is a conservative assumption.

Taking the natural logarithm on both sides of Beer's Law, Equation 1, leads to Equation 5,

which is the equation of a straight line of the logarithmic intensity ln _{s}. For this reason, the natural logarithms of the intensity of the peak measurements for the Intralipid-glass interface were recorded for each concentration and plotted against thickness of the tissue phantom as shown in Figure ^{-1}, 7.19 mm^{-1}, 11.3 mm^{-1 }and 15.0 mm^{-1 }for the concentrations of 0.5%, 1.0%, 1.5%, and 2.0%, respectively. R^{2 }goodness of fit values were all greater than 0.99 for each concentration. All linear regressions had a significant non-zero slope with P < 0.005. Furthermore, the extrapolation of the lines to _{0}, and the low variation between the data sets further confirmed the validity of the measurement.

Natural logarithm of the intensity signal measured from a 0.22 refractive index change (Intralipid to glass) at various thicknesses at four concentrations

**Natural logarithm of the intensity signal measured from a 0.22 refractive index change (Intralipid to glass) at various thicknesses at four concentrations**. Each concentration was fit to a linear regression to obtain the decay constants.

The slopes obtained from the regression were used to determine the scattering coefficient of the various concentrations of Intralipid. Considering the round-trip attenuation, the slope values need to be divided by two to obtain the Intralipid scattering coefficient, and we obtained _{s }as 17.0 cm^{-1 }for 0.5% Intralipid, 35.9 cm^{-1 }for 1.0%, 56.7 cm^{-1 }for 1.5%, and 75.1 cm^{-1 }for 2.0%. The values for _{s }were plotted against Intralipid concentration in Figure _{s }with Intralipid concentration. Linear regression of the data yielded a slope of 39 ± 0.59 cm^{-1 }per percent. R^{2 }goodness of fit value was approximately 1.0 with P-value 0.0002, showing a significant non-zero slope. Referring back to Equation 3, we were now able to model the maximum imaging depth as a function of refractive index change at the fixed scattering coefficients obtained from Figure

Extracted scattering coefficient from the decay constants in Figure 4

**Extracted scattering coefficient from the decay constants in Figure 4**. Error bars denote 95% confidence of the decay constants. This was fit to a linear regression in order to extrapolate additional concentrations.

Theoretical penetration depth assuming minimum ratio of 2% for I_{d}/I_{p }in order to identify a specific refractive index change Δn at the four scattering coefficients shown in Figure 5

**Theoretical penetration depth assuming minimum ratio of 2% for I _{d}/I_{p }in order to identify a specific refractive index change Δn at the four scattering coefficients shown in Figure 5**.

In the case of collagen gel scattering, the averaged scattering signal from each of the 42 samples (three of which are shown in Figure ^{2 }goodness of fit value was 0.95 with P-value of less than 0.0001, showing a significant non-zero slope. Analysis of residuals show a random normal distribution of mean approximately zero and standard deviation of 0.0141. Pearson Product Moment correlation is 2.183 × 10^{-9 }with p-value 1.0, therefore showing no correlation of residuals with the data and indicating our linear model was appropriate in defining collagen scattering.

Selected B-mode scans of collagen gels (top) and their averages (bottom)

**Selected B-mode scans of collagen gels (top) and their averages (bottom)**. The rectangle indicates the region that was averaged again to determine the single-quantity scattering signal.

Quantified backscattering intensity from increasing concentrations of collagen gels

**Quantified backscattering intensity from increasing concentrations of collagen gels**.

Discussion

In this study, we used two different types of samples to examine the ability of the scattering-confocal microscope to quantitatively recover the scattering properties of the samples. First, we attempted to estimate the scattering coefficient of various concentrations of Intralipid by measuring reflection at various depths from refractive index mismatches, and second, we related the absolute scattering signal from collagen gels to collagen concentration.

In strongly scattering media, such as Intralipid, depth imaging ability is important. While a reflection signal can still be attained at high imaging depths, the resolution decreases heavily at higher thicknesses due to specimen-induced spherical aberration ^{-1}. The detection limit strongly depends on the signal-to background ratio of the system which we determined to be in the order of 24 dB. This value is predominantly determined by digitization noise, and an analog-to-digital converter with higher resolution could markedly improve the signal-to-background ratio. In practice, however, we found better sensitivity than the theoretical considerations allowed for. From Figure ^{-1 }for the scattering coefficient, the round-trip transmission through the Intralipid solution is 0.5%, a value that is in excellent agreement with the measured signal-to-background ratio of 24 dB or 1:250. We conclude that the estimate presented in Figure

From Figure _{s }for the Intralipid dilutions. Figure _{s }and Intralipid concentrations are linearly dependent for the concentration range examined. For our case, an extrapolation of Figure ^{-1 }at 660 nm. This value falls between the values measured by Flock ^{-1 }_{0 }needs to be determined, for example by scanning the glass plates at a location outside of the tissue sample, and _{s }obtained from the average reflected intensity <_{s }with the known intensities <_{0 }and the thickness _{s }(x, y) over the area of the tissue under examination. In applications of tissue engineering, such a scan would provide information about average scattering, homogeneity of scattering, and possible areas of unexpectedly low scattering in the tissue sample. Since collagen is the primary source of scattering in tissue

The main disadvantage of the above method is the invasive nature of the measurement: the tissue needs to be placed between two plates of glass to allow imaging of the scattering map _{s }(x, y). In addition, there is an error of up to 5% in the scaling of the

Residual analysis of Figure 8

**Residual analysis of Figure 8**. Concentration of 1.8 mg/mL is circled with values denoted as asterisks on histogram.

When the ability to determine the scattering coefficient of a low absorbing medium is combined with the ability to measure backscattering directly, it is possible to generate a calibration curve to relate intensity directly to backscattering for diagnostic imaging purposes. This would provide the possibility to not only determine a relative concentration of collagen in a given area, but to actually quantify the amount of collagen itself. This would particularly be useful in the field of tissue engineering, where collagen scaffolds are necessary to provide tensile stiffness and mechanical strength to the graft

Conclusion

In conclusion, we showed that our confocal scanner can be used to investigate the scattering properties of turbid media. The use of Intralipid demonstrates our ability to measure reflection within low scattering medium in order to extract optical properties when absorption coefficient is negligible. The optical sectioning capability of the single-mode fiber provides the necessary penetration to accomplish this task. The recognition of increased scattering from higher density collagen gels provides supportive evidence that this device can be used to analyze collagen density in terms of contrast for

Competing interests

The authors declare that they have no competing interests.

Authors' contributions

JTL and MH jointly designed and built the confocal scattered-light scanner. JTL performed the experiments described in this paper, and MH supervised the study. JTL wrote major parts of the manuscript, and MH contributed most of the introduction and discussion.

Acknowledgements

The authors would like to acknowledge financial support from the Jürgen Manchot Foundation (Düsseldorf, Germany) and from the National Institutes of Health, grant R21 HL081308. JTL is a University of Missouri Life Sciences Predoctoral Fellow.

Pre-publication history

The pre-publication history for this paper can be accessed here: