## Quantitative comparison of analysis methods for spectroscopic optical coherence tomography |

Biomedical Optics Express, Vol. 4, Issue 11, pp. 2570-2584 (2013)

http://dx.doi.org/10.1364/BOE.4.002570

Acrobat PDF (3064 KB)

### Abstract

Spectroscopic optical coherence tomography (sOCT) enables the mapping of chromophore concentrations and image contrast enhancement in tissue. Acquisition of depth resolved spectra by sOCT requires analysis methods with optimal spectral/spatial resolution and spectral recovery. In this article, we quantitatively compare the available methods, i.e. the short time Fourier transform (STFT), wavelet transforms, the Wigner-Ville distribution and the dual window method through simulations in tissue-like media. We conclude that all methods suffer from the trade-off in spectral/spatial resolution, and that the STFT is the optimal method for the specific application of the localized quantification of hemoglobin concentration and oxygen saturation.

© 2013 Optical Society of America

## 1. Introduction

4. C. Xu, J. Ye, D. L. Marks, and S. A. Boppart, “Near-infrared dyes as contrast-enhancing agents for spectroscopic optical coherence tomography,” Opt. Lett. **29**(14), 1647–1649 (2004). [CrossRef] [PubMed]

6. N. Bosschaart, M. C. G. Aalders, D. J. Faber, J. J. A. Weda, M. J. C. van Gemert, and T. G. van Leeuwen, “Quantitative measurements of absorption spectra in scattering media by low-coherence spectroscopy,” Opt. Lett. **34**(23), 3746–3748 (2009). [CrossRef] [PubMed]

10. C. P. Fleming, J. Eckert, E. F. Halpern, J. A. Gardecki, and G. J. Tearney, “Depth resolved detection of lipid using spectroscopic optical coherence tomography,” Biomed. Opt. Express **4**(8), 1269–1284 (2013). [CrossRef] [PubMed]

3. U. Morgner, W. Drexler, F. X. Kärtner, X. D. Li, C. Pitris, E. P. Ippen, and J. G. Fujimoto, “Spectroscopic optical coherence tomography,” Opt. Lett. **25**(2), 111–113 (2000). [CrossRef] [PubMed]

11. R. N. Graf and A. Wax, “Temporal coherence and time-frequency distributions in spectroscopic optical coherence tomography,” J. Opt. Soc. Am. A **24**(8), 2186–2195 (2007). [CrossRef] [PubMed]

5. F. E. Robles, C. Wilson, G. Grant, and A. Wax, “Molecular imaging true-colour spectroscopic optical coherence tomography,” Nat. Photonics **5**(12), 744–747 (2011). [CrossRef] [PubMed]

12. F. Robles, R. N. Graf, and A. Wax, “Dual window method for processing spectroscopic optical coherence tomography signals with simultaneously high spectral and temporal resolution,” Opt. Express **17**(8), 6799–6812 (2009). [CrossRef] [PubMed]

13. C. Xu, F. Kamalabadi, and S. A. Boppart, “Comparative performance analysis of time-frequency distributions for spectroscopic optical coherence tomography,” Appl. Opt. **44**(10), 1813–1822 (2005). [CrossRef] [PubMed]

8. N. Bosschaart, D. J. Faber, T. G. van Leeuwen, and M. C. G. Aalders, “Measurements of wavelength dependent scattering and backscattering coefficients by low-coherence spectroscopy,” J. Biomed. Opt. **16**(3), 030503 (2011). [CrossRef] [PubMed]

9. J. Yi and V. Backman, “Imaging a full set of optical scattering properties of biological tissue by inverse spectroscopic optical coherence tomography,” Opt. Lett. **37**(21), 4443–4445 (2012). [CrossRef] [PubMed]

^{2}) between the input and recovered absorption coefficient spectra, and we evaluate the quantification of the total hemoglobin concentration and oxygen saturation.

## 2. Theory

*λ*– and

*z*–domain simultaneously, with high resolution in both

*λ*and

*z*. Due to wavelength-dependent scattering and absorption by the different structures in tissue, the spectral content of the LCI signal changes with depth. If we would directly apply the Fourier transform on the LCI signal (Eq. (2), either the depth, or wavelength varying information will be lost – depending on time, or spectral domain signal acquisition, respectively. Instead, depth-resolved spectroscopic information can be obtained using time-frequency (TF) analysis methods (or, for the present context: spectral/spatial analysis). The four accepted methods for spectral-spatial analysis in sOCT and LCS will be described next. We point out that these methods can be performed equivalently on OCT data that is acquired in the spatial domain (described here) or in the spectral domain [14

14. R. Leitgeb, M. Wojtkowski, A. Kowalczyk, C. K. Hitzenberger, M. Sticker, and A. F. Fercher, “Spectral measurement of absorption by spectroscopic frequency-domain optical coherence tomography,” Opt. Lett. **25**(11), 820–822 (2000). [CrossRef] [PubMed]

15. A. Wax, C. Yang, and J. A. Izatt, “Fourier-domain low-coherence interferometry for light-scattering spectroscopy,” Opt. Lett. **28**(14), 1230–1232 (2003). [CrossRef] [PubMed]

### 2.1 Short time Fourier transform (STFT)

*w(z,*Δ

*z)*is an analysis window confined in space around

*z*with spatial width

*Δz*, for example a Gaussian function. The multiplication with a relatively short window effectively suppresses the signal outside the analysis point

*z*± ½Δ

*z*. Physically, the STFT can be considered as the result of passing a signal through an array of band-pass filters with linearly increasing center frequency and constant bandwidth which is inversely proportional to Δ

*z*. Thus, there is an inherent trade-off between spectral and spatial resolution. A window with short spatial width Δ

*z*will localize the signal well in space but will have reduced

*k*-resolution; conversely a signal with long spatial width will be less well localized in space with the benefit of increased spectral resolution. For a Gaussian window, a spatial domain width Δ

*z*will yield a spectral (wavenumber) resolution of Δ

*k*= 1/(2Δ

*z*), or equivalent wavelength resolution Δ

*λ*:In practice, starting with a real-valued time-domain A-scan, a spatial window w

_{z}with center at z and width Δz is chosen. This window is multiplied with the A-scan and subsequently Fourier transformed with z→k (Eq. (3). Repeating for every z results in a complex valued STFT(k,z; w

_{z}) spectrogram with z-axis equal to that of the time-domain A-scan, and k-axis from k = 0 to k = π/2δz where δz is the sampling increment of the A-scan. The resolution in the spatial domain is given by de window width Δz; the resolution in the spectral domain is Δk = 1/2Δz assuming a Gaussian window (Eq. (4). When operating on real-valued Fourier domain interference spectra, a k-domain window w

_{k}centered at k and width Δk is chosen, and multiplied with the interference spectrum. Fourier transform k→z at each k then yields the complex spectrogram STFT(k,z; w

_{k}). The spectral range equals that of the interference spectrum; the spatial range runs from z = 0 to z = π/2δk where δk is the spectral domain sampling interval. For further processing the amplitude of the complex spectrogram is taken.

### 2.2 Wavelet transforms

3. U. Morgner, W. Drexler, F. X. Kärtner, X. D. Li, C. Pitris, E. P. Ippen, and J. G. Fujimoto, “Spectroscopic optical coherence tomography,” Opt. Lett. **25**(2), 111–113 (2000). [CrossRef] [PubMed]

### 2.3 Wigner-Ville distribution (WV)

14. R. Leitgeb, M. Wojtkowski, A. Kowalczyk, C. K. Hitzenberger, M. Sticker, and A. F. Fercher, “Spectral measurement of absorption by spectroscopic frequency-domain optical coherence tomography,” Opt. Lett. **25**(11), 820–822 (2000). [CrossRef] [PubMed]

*i*. Whereas in conventional autocorrelation computations the result is a function of lag (

_{D}(z)*z*–

*z’*) only, here the dependence on

*z*is maintained. In concordance with the Wiener-Khinchine theorem, the straightforward physical interpretation of the WV is as a localized power spectral density of the detector signal. The result of Eq. (5) is complex; for further processing the absolute value is taken.

_{x}(

*k,z*) + STFT

_{y}(

*k,z*), the WV will contain interference terms i.e. WV

_{x+y}(

*k,z*) = WV

_{x}(

*k,z*) + WV

_{y}(

*k,z*) + 2Re[WV

_{x,y}(

*k,z*)]. Even though the interference term contains information on the separation of X and Y (i.e. in space), when it overlaps with the signal terms, interpretation of the WV becomes challenging. In practice, the signal is analyzed using a window

*w*, similar to the STFT. This Pseudo-WV effectively smoothes the spectral/spatial distribution, suppressing the interference terms. However, a short spatial smoothing window will be wide in frequency, leading to a good spatial resolution but bad spectral resolution and vice versa. It is possible to add a degree of freedom by considering a separable smoothing function Δ(

*k,z*) =

*w*(

_{k}*k,*Δ

*k*)

*w*(

_{z}*z,*Δ

*z*) that allows independent control in both depth and frequency of the smoothing applied to the WV; Δ

*k*and Δ

*z*denote the width of the window functions in the spectral and spatial domain, respectively. The STFT compromise between spatial and spectral resolution is now replaced by a compromise between the joint spectral/spatial resolution and the level of suppression of the interference terms.

### 2.4 Dual window method (DW)

*et al.*showed the pseudo WV can also be obtained by STFT analysis using two k-domain window sizes Δk

_{1}>>Δk

_{2}[12

12. F. Robles, R. N. Graf, and A. Wax, “Dual window method for processing spectroscopic optical coherence tomography signals with simultaneously high spectral and temporal resolution,” Opt. Express **17**(8), 6799–6812 (2009). [CrossRef] [PubMed]

_{1}<<Δz

_{2}). The result STFT

_{1}(k,z) × STFT

_{2}(k,z) is mathematically equivalent to a pseudo WV with window widths of

*w*(

_{k}*k*,Δ

*k*= Δ

*k*/2) in the spectral domain and window

_{2}*w*(

_{z}*z*,Δ

*z*= 1/(2Δ

*k*)) in the spatial domain. As for the WV, any remaining interference terms may carry information, for example on average spatial scatterer separation. Robles

_{1}*et al*state that the DW method is superior to the STFT in terms of the trade-off in spectral/spatial resolution – i.e. the spatial resolution equals the width of the narrow spatial window, and the spectral resolution equals the width of the narrow spectral window. When applied in practice, interference terms between nearby reflectors (both located within one long spatial window) appear as frequency modulations on the recovered spectra. These modulations can be removed by low-pass frequency domain filtering [12

12. F. Robles, R. N. Graf, and A. Wax, “Dual window method for processing spectroscopic optical coherence tomography signals with simultaneously high spectral and temporal resolution,” Opt. Express **17**(8), 6799–6812 (2009). [CrossRef] [PubMed]

## 3. Methods

### 3.1 Simulations

_{max}of 375 µm. In the A-scan, we introduce two point reflectors with a spacing of D = 25 µm: R1 at 220 µm and R2 at 245 µm (Fig. 1(a)), representing the walls of a blood vessel. The spectral reflection S

_{R1}(k) of R1 (Fig. 1(b): actual) is defined by a Gaussian ‘source’ spectrum, centered around k

_{0}= 11.5 µm

^{−1}(corresponding to wavelength λ

_{0}= 2π/k

_{0}= 546nm) on a wavenumber range k = 8-15 µm

^{−1}(λ = 419-785nm) with a full width half max of k

_{FWHM}= 2.35 µm

^{−1}(λ

_{FWHM}= 113nm). The spacing between R1 and R2 is simulated as an absorbing blood volume or vessel, with a total hemoglobin concentration of [tHb] = 150 g/L, an oxygen saturation of SO

_{2}= 85% and zero scattering. We can therefore define the spectral reflection S

_{R2}(k) of R2 (Fig. 1(c): actual) as the source spectrum S

_{R1}(k) from R1, multiplied with the influence of absorption from the blood volume:where the absorption coefficient spectrum:is obtained from the absorption spectra of deoxygenized (Hb) and oxygenized (HbO

_{2}) hemoglobin in g/L [17

17. Data tabulated from various sources compiled by S. Prahl, http://omlc.ogi.edu/spectra.

_{2}as a percentage. Subsequently, we convolute the point reflectors with an OCT interferogram that has a coherence length corresponding to the spectral content at R1 and R2 (l

_{c}≈1.3 µm) and calculate the actual A-scan envelope (Fig. 1(a): actual).

_{M}with Δz

_{M}= 22 µm in the spatial domain – as we used in our previous work [6

6. N. Bosschaart, M. C. G. Aalders, D. J. Faber, J. J. A. Weda, M. J. C. van Gemert, and T. G. van Leeuwen, “Quantitative measurements of absorption spectra in scattering media by low-coherence spectroscopy,” Opt. Lett. **34**(23), 3746–3748 (2009). [CrossRef] [PubMed]

8. N. Bosschaart, D. J. Faber, T. G. van Leeuwen, and M. C. G. Aalders, “Measurements of wavelength dependent scattering and backscattering coefficients by low-coherence spectroscopy,” J. Biomed. Opt. **16**(3), 030503 (2011). [CrossRef] [PubMed]

_{S}) and a long (w

_{L}) window for the DW method – comparable to those used by Robles

*et al*[5

5. F. E. Robles, C. Wilson, G. Grant, and A. Wax, “Molecular imaging true-colour spectroscopic optical coherence tomography,” Nat. Photonics **5**(12), 744–747 (2011). [CrossRef] [PubMed]

_{0}) spectrally, whereas the resolutions for the DW method are 5 µm in depth and 2 nm (at λ

_{0}) spectrally.

_{L}and w

_{S}.

_{L}of the DW method probes both R1 and R2 simultaneously, a frequency modulation is visible in the DW spectra due to the interference between both reflectors. When taking the Fourier transform of the DW and STFT spectra in Fig. 1(c), the modulation frequency appears exactly at the reflector spacing of 25 µm for the DW (Fig. 1(d)). Hence, this modulation can be removed efficiently by applying an (ideal) low-pass filter with a cut-off frequency of 7 µm on the Fourier transformed spectra. Note that this cut-off frequency is higher than the previously reported cut-off frequency of 3 µm for the DW method [12

**17**(8), 6799–6812 (2009). [CrossRef] [PubMed]

### 3.2 Simulation cases

_{a}:

- Case 1: Simulation of the two reflectors R1 and R2, as described in Section 3.1. The aim of this case is to recover µ
_{a}in an ‘ideal’ situation, i.e. without the influence of other reflectors. - Case 2: Simulation of three reflectors, with a third reflector R3 introduced at depths larger than R2. R3 is simulated as a strong reflector without any absorption between R2 and R3 (magnitude S
_{R3}= 3.5 × magnitude S_{R2}). The distance between R3 and R2 is varied from 0 – 100 µm. In this case, we also investigate the influence of noise on the recovery of µ_{a}for 20 µm and 100 µm separation distance. - Case 3: Identical to case 2, but with varying spectral powers for R3 (S
_{R3}= α*S_{R2}, with α = 0.25–3.5). - Case 4: Identical to case 2, but with moderate power for R3 (S
_{R3}= S_{R2}) and varying size of the long DW window w_{L}(Δz_{L}= 5 – 75 µm).

### 3.3 Quantification of performance

_{a}of the simulated blood volume:

- 1) The normalized chi-squared χ
^{2}value of µ_{a}recovery. The χ^{2}value of µ_{a}is obtained by:where µ_{a,act}is the actual (input) µ_{a}and µ_{a,sim}is the recovered µ_{a}by either the DW, or the STFT method. All χ^{2}values are obtained between 475 and 650 nm, hence λ_{1}= 475 nm and λ_{N}= 650 nm. Although the wavelength resolutions differ between the DW and STFT method (Table 1), N is equal for both methods in the 475-650 nm range, due to zeropadding in the calculation of the spectral/spatial distributions. Equation (6) shows that the lower χ^{2}, the better the recovery of µ_{a}. - 2) The effective spatial resolution. It is expected that χ
^{2}will increase when a third reflector R3 (at a distance >R2) approaches R2 towards a distance that is similar to the spatial resolution of both methods, since S_{R2}(k) cannot be recovered independently of S_{R3}(k) in that case. We define the*effective*spatial resolution as the distance between R2 and R3 where χ^{2}deviates more than 15% from the χ^{2}without the presence of R3. (The allowed deviation can be chosen dependent on application). - 3) The correct recovery of the actual total hemoglobin concentration [tHb] and oxygen saturation (SO
_{2}) from µ_{a}. The [tHb] and SO_{2}are recovered from the simulated µ_{a}spectra through a non-linear least squares chromophore fit of the model of Eq. (7) to the simulated µ_{a}with fit parameters [tHb] and SO_{2}.

## 4. Results

### 4.1 Case 1: two reflectors

_{2}= 85%). The black line in Fig. 1(a) corresponds to the actual A-scan envelope; the black lines in Figs. 1(b), 1(c), 1(e) and1(f) to the actual reflection spectrum (i.e. the simulation inputs). The recovered µ

_{a}spectrum from the blood volume is shown in Fig. 2 for both methods, using the filtered spectra of R1 and R2. Compared to the STFT, the spatial recovery is much better for the DW method (Fig. 1(a)). However, the recovered spectra by the DW method deviate stronger from the actual spectrum than the spectra recovered by the STFT (Figs. 1(e), 1(f) and Fig. 2), which is reflected in the values for χ

^{2}. Table 2 summarizes the performance of the STFT and DW method for the spectral recovery of R1 and R2 and the µ

_{a}in between. Low-pass filtering of the spectra enhances the χ

^{2}values for the DW method, but it remains higher with 1-2 orders of magnitude than the χ

^{2}values for the STFT method.

^{2}values for the unfiltered and filtered spectra of R1 and R2 within the presented significance (Table 2). The slight increase in the χ

^{2}value of µ

_{a}after filtering is attributed to the minor loss of frequencies containing hemoglobin-specific absorption features above 7 µm (Fig. 1(d)).

_{2}of 85% is severely underestimated by both the STFT (with 18%) and the DW method (with 40%). This underestimation results from the flattening of the HbO

_{2}absorption peaks due to spectral smoothing by the finite window sizes. Figure 3 shows the dependency of the fitted SO

_{2}(Fig. 3(a)), fitted [tHb] (Fig. 3(b)), and χ

^{2}of µ

_{a}(Fig. 3(c)) on the actual (input) value of the SO

_{2}. Since the HbO

_{2}peaks become sharper for increasing SO

_{2}values, the amount of peak flattening increases with SO

_{2}. As a consequence, the underestimation of the fitted SO

_{2}and [tHb] increases, as well as the χ

^{2}of µ

_{a}. Due to the consistent linear behavior of the fitted SO

_{2}and [tHb] on the actual SO

_{2}, the fitted values in practice can be recalculated to the actual values if this relation is known – e.g. from a simulation. For the STFT method, this correction can also be applied directly by smoothing of the µ

_{a,HbO2}and µ

_{a,Hb}spectra in the chromophore fit with the spectral resolution of the STFT [18

18. N. Bosschaart, M. C. G. Aalders, T. G. van Leeuwen, and D. J. Faber, “Spectral domain detection in low-coherence spectroscopy,” Biomed. Opt. Express **3**(9), 2263–2272 (2012). [CrossRef] [PubMed]

### 4.2 Case 2: three reflectors, varying separation

_{R3}= 3.5 × S

_{R2}) on the determination of the µ

_{a}of the blood volume between R1 and R2. It is expected that R3 will influence this determination if R3 approaches R2 within the spatial resolution of the analysis methods (within 22 µm for the STFT, and 5 µm for the DW). Figure 4 shows a situation where R3 is located at a distance of 20 µm from R2 (Fig. 4(a)). Although this is well outside the spatial resolution of the DW method, the recovered µ

_{a}between R1 and R2 by the DW method is highly influenced by the presence of R3 (Fig. 4(b)).

_{a}by the STFT remains practically unaffected and agrees well with the actual (input) spectrum of µ

_{a}. Also for this case, the A-scan recovery of the DW method is superior to the STFT. Nevertheless, the spectral performance of the DW method is expectedly compromised by poorer spectral recovery when the large reflector R3 is probed simultaneously with R2 by the long spatial DW window w

_{L}.

_{a}between R1 and R2. Figure 5 shows the influence of R3 on the determination of the µ

_{a}of the blood volume between R1 and R2 for several more separations of R2 and R3 (345 to 245 µm).

^{2}of µ

_{a}, Fig. 5(b) on the fitted [tHb] and Fig. 5(c) on the fitted SO

_{2}. For all parameters, a deviation from the baseline value is observed at separations smaller than ~16 µm for the STFT, and smaller than ~30 µm for the DW method. Although the χ

^{2}of µ

_{a}for the DW method first decreases for separations <~30 µm (indicating better spectral recovery), the fitted values of the [tHb] and SO2 are also affected at these separations – indicating worse spectral recovery. This is caused by the introduction of frequency modulations on the recovered µ

_{a}(see Fig. 4, Media 1) for separations <~30 µm, which ‘deceptively’ results in a better χ

^{2}of µ

_{a}. For this case, the

*effective*spatial resolution (defined as a deviation in χ

^{2}of >15% from the χ

^{2}without the presence of R3) of both methods is therefore 16 µm for the STFT, and 30 µm for the DW method – compared to the theoretical resolutions of 22 µm and 5 µm, respectively.

^{2}of μ

_{a}recovery, we added uniform white noise to the simulated signals as illustrated in Fig. 4(c) and Fig. 4(d). We varied noise powers, represented as a signal-to-noise ratio, here defined as:

_{L}. In this case, the influence of R3 is negligible. As can be expected, χ

^{2}increases with decreasing SNR for both methods. Performance of the DW method based on the metric χ

^{2}is better than STFT under low SNR conditions. When R3 is positioned within the large DW window (separation 20 μm; Δz

_{S}< 20 μm < Δz

_{L}), the performance of the STFT is expectedly better for low noise conditions (Fig. 6(b)). For low SNR, the STFT and the DW method appear to perform equally in this situation. In the practice of sOCT and LCS, the influence of noise on the measurement of optical property spectra is reduced by temporal and/or spatial averaging of the acquired spectra [7

7. N. Bosschaart, D. J. Faber, T. G. van Leeuwen, and M. C. G. Aalders, “In vivo low-coherence spectroscopic measurements of local hemoglobin absorption spectra in human skin,” J. Biomed. Opt. **16**(10), 100504 (2011). [CrossRef] [PubMed]

### 4.3 Case 3: three reflectors, varying separation and power

_{R3}= α × S

_{R2}, with α = 0.25–3.5). Figure 7 shows the resulting effective spatial resolutions as a function of the power ratio α between R3 and R2. As expected, the effective spatial resolution becomes smaller (i.e. better) for lower power reflected from R3. When the power of R3 is very small compared to the power of R2 (α = 0.25), the effective spatial resolution of the DW method (8 µm) is close to the theoretical spatial resolution of 5 µm. Although the effective spatial resolution of the DW is better than that of the STFT for this power ratio α, the spectral recovery for the STFT method remains better with a χ

^{2}of 0.15, compared to a χ

^{2}of 3.7 for the DW.

### 4.4 Case 4: three reflectors, varying separation and window sizes

_{S}), we expect that the long spatial window (Δz

_{L}) influences the effective spatial resolution. To test this hypothesis, we repeated Case 2 with moderate power for R3 (S

_{R3}= S

_{R2}) and varying sizes of Δz

_{L}. Figure 8(a) shows that the effective spatial resolution of the DW method decreases as a function of the size of Δz

_{L}and only comes close to the theoretical spatial resolution of 5 µm when Δz

_{L}≈Δz

_{S}. For window sizes Δz

_{L}smaller than approximately 45 µm (~2*Δz

_{M}), the effective resolution of the DW method is better than that of the STFT. However, the χ

^{2}of µ

_{a}at this effective resolution is 3.5 for the DW method (i.e. at a separation between R2 and R3 of ~45 µm), whereas the χ

^{2}of µ

_{a}for the STFT is only 0.12 for the same separation between R2 and R3 – indicating better spectral recovery for the STFT.

^{2}of µ

_{a}as a function of the separation between R2 and R3 for the STFT method and for a selection of long spatial window sizes (Δz

_{L}= 5, 7.5, 15 and 75 µm) of the DW method. The χ

^{2}of the DW method is better (i.e. smaller) than that of the STFT method when the separation between R2 and R3 is smaller than ~7 µm, for all window sizes Δz

_{L}. However, the spectral recovery in this case is poor for both methods with a χ

^{2}of µ

_{a}around 10 and higher, resulting in deviations from the actual χ

^{2}of µ

_{a}comparable to the DW result in Fig. 4(b).

_{L}(see Case 3).

## 5. Discussion

**17**(8), 6799–6812 (2009). [CrossRef] [PubMed]

*et al*[13

13. C. Xu, F. Kamalabadi, and S. A. Boppart, “Comparative performance analysis of time-frequency distributions for spectroscopic optical coherence tomography,” Appl. Opt. **44**(10), 1813–1822 (2005). [CrossRef] [PubMed]

_{a}) and the application of localized measurements of hemoglobin concentrations [tHb] and oxygen saturation (SO

_{2}) in tissue. Besides the spectral/spatial resolution, the spectral recovery of the analysis method is an important aspect for this application, since small changes in spectral shape and amplitude can lead to large errors in the parameters of interest.

^{2}values of the recovered spectra in Fig. 1(e), 1(f) and Fig. 2 (Table 2). As a consequence, the [tHb] and SO

_{2}of the blood volume that were derived from the DW spectra show more deviation from the actual values than the same parameters derived from the STFT spectra. Both the STFT and the DW method underestimate the actual SO

_{2}of the simulated blood volume (Fig. 3(a)), due to recovery of the spectra with a lower spectral resolution than the reference spectrum that was used for quantification of the SO

_{2}. Fortunately, this underestimation can be compensated for by spectral smoothing of the reference spectrum prior to the fitting procedure.

_{L}of the DW method is larger than the STFT spatial window Δz

_{M}, the DW method suffers more from this artifact than the STFT. Low-pass filtering is an effective solution to this artifact, but is limited to a cut-off frequency of 7 µm for applications involving hemoglobin absorption, due the features of its absorption spectrum. Hence, when reflectors are spatially separated by less than 7 µm, a reliable recovery of hemoglobin absorption is not possible by any method if window sizes exceed 7 µm. We expect that this problem is of minor influence in real tissue, since multiple reflectors at varying separations will average out this effect.

_{S}– seems to be better than that of the STFT (Fig. 1(a)), the recovered spectra by the DW method are influenced by reflectors that are located outside the theoretical spatial resolution (Fig. 4). Hence, the

*effective*spatial resolution of the DW method is worse than its theoretical spatial resolution. We explain this difference by the influence of the large spatial window Δz

_{L}on the effective spatial resolution (Fig. 8(a)) – i.e. Δz

_{L}and Δz

_{S}are not entirely decoupled in the spectral/spatial recovery. On the other hand, the effective spatial resolution of the STFT was better than its theoretical spatial resolution Δz

_{M}. This can be explained by the Gaussian shape of the STFT window, which favors those reflections that are probed in its center and attenuates those reflections that are probed in its tails. Note that the effective spatial resolution is a qualitative measure of resolution, since it depends on the relative power of the reflectors (Fig. 7), noise power and choice of maximum allowed χ

^{2}of µ

_{a}deviation of baseline. It is intended to facilitate comparisons between the STFT and DW methods, and is indicative of the spatial range over which spectral information is obtained.

_{2}in tissue with LCS (using STFTs: Δz

_{M}= 22 µm, as in Ref [7

7. N. Bosschaart, D. J. Faber, T. G. van Leeuwen, and M. C. G. Aalders, “In vivo low-coherence spectroscopic measurements of local hemoglobin absorption spectra in human skin,” J. Biomed. Opt. **16**(10), 100504 (2011). [CrossRef] [PubMed]

_{S}= 5 µm and Δz

_{L}= 75 µm, as in Ref [5

5. F. E. Robles, C. Wilson, G. Grant, and A. Wax, “Molecular imaging true-colour spectroscopic optical coherence tomography,” Nat. Photonics **5**(12), 744–747 (2011). [CrossRef] [PubMed]

_{L}. For a power ratio of 1 between the second and third reflector (which is a reasonable scenario in tissue given the small spatial variations in refractive index), the effective spatial resolution of the DW method is better than that of the STFT when Δz

_{L}≈2Δz

_{M}(Fig. 7(a)). However, in this case the spectral recovery (χ

^{2}of µ

_{a}) for the DW method only becomes better than that of the STFT when Δz

_{L}≈7 µm (i.e. Δz

_{L}≈1/3Δz

_{M,}or Δz

_{L}≈1.5Δz

_{S}, Fig. 8(b)) - although the spectral recovery for neither of the two methods is good enough for the quantitative interpretation of µ

_{a}in this situation. Hence, the DW method can perform better than the STFT, but only if the spatial distribution of tissue reflectivity is relatively homogeneous, and not for the condition that Δz

_{L}>> Δz

_{S}.

7. N. Bosschaart, D. J. Faber, T. G. van Leeuwen, and M. C. G. Aalders, “In vivo low-coherence spectroscopic measurements of local hemoglobin absorption spectra in human skin,” J. Biomed. Opt. **16**(10), 100504 (2011). [CrossRef] [PubMed]

19. J. Yi and X. Li, “Estimation of oxygen saturation from erythrocytes by high-resolution spectroscopic optical coherence tomography,” Opt. Lett. **35**(12), 2094–2096 (2010). [CrossRef] [PubMed]

**16**(10), 100504 (2011). [CrossRef] [PubMed]

_{t}= µ

_{a}+ µ

_{s}, with µ

_{s}the scattering coefficient) rather than the absorption coefficient only. Several models exist to accurately separate the scattering and absorption contributions to µ

_{t}and from that, the tissue chromophore concentrations [7

**16**(10), 100504 (2011). [CrossRef] [PubMed]

10. C. P. Fleming, J. Eckert, E. F. Halpern, J. A. Gardecki, and G. J. Tearney, “Depth resolved detection of lipid using spectroscopic optical coherence tomography,” Biomed. Opt. Express **4**(8), 1269–1284 (2013). [CrossRef] [PubMed]

_{M}(Δz

_{M},Δk

_{M}). Although we investigated its performance only for one size of w

_{M}, we showed that hemoglobin absorption spectra can be recovered accurately with this window within the expected spectral and spatial resolutions. Other sizes for w

_{M}may be more effective in other situations, e.g. for other wavelength ranges and other chromophores. This can be checked with simulations similar to those presented in this paper. The DW method requires more careful optimization in terms of window sizes, as we showed for the long spatial window size Δz

_{L}. Reducing the size of the small spatial window Δz

_{S}will further increase the quality of spatial (A-scan) recovery until Δz

_{S}equals the coherence length of the source spectrum. However, this will lead to a further decrease in spectral recovery. Since spectral shape is preserved quite well by the DW method (Fig. 4(b)), it is excellently suited for contrast enhancement and color coding of high resolution OCT images. However, based on our simulations, it is less well suited for the quantitative recovery of optical property spectra, because the spectral amplitude is less well preserved (Fig. 4(b)).

## 6. Conclusion

## Acknowledgments

## References and links

1. | N. Bosschaart, T. G. van Leeuwen, M. C. G. Aalders, B. Hermann, W. Drexler, and D. J. Faber, “Spectroscopic low coherence interferometry”, Chapter 23 in |

2. | J. M. Schmitt, S. H. Xiang, and K. M. Yung, “Differential absorption imaging with optical coherence tomography,” J. Opt. Soc. Am. A |

3. | U. Morgner, W. Drexler, F. X. Kärtner, X. D. Li, C. Pitris, E. P. Ippen, and J. G. Fujimoto, “Spectroscopic optical coherence tomography,” Opt. Lett. |

4. | C. Xu, J. Ye, D. L. Marks, and S. A. Boppart, “Near-infrared dyes as contrast-enhancing agents for spectroscopic optical coherence tomography,” Opt. Lett. |

5. | F. E. Robles, C. Wilson, G. Grant, and A. Wax, “Molecular imaging true-colour spectroscopic optical coherence tomography,” Nat. Photonics |

6. | N. Bosschaart, M. C. G. Aalders, D. J. Faber, J. J. A. Weda, M. J. C. van Gemert, and T. G. van Leeuwen, “Quantitative measurements of absorption spectra in scattering media by low-coherence spectroscopy,” Opt. Lett. |

7. | N. Bosschaart, D. J. Faber, T. G. van Leeuwen, and M. C. G. Aalders, “In vivo low-coherence spectroscopic measurements of local hemoglobin absorption spectra in human skin,” J. Biomed. Opt. |

8. | N. Bosschaart, D. J. Faber, T. G. van Leeuwen, and M. C. G. Aalders, “Measurements of wavelength dependent scattering and backscattering coefficients by low-coherence spectroscopy,” J. Biomed. Opt. |

9. | J. Yi and V. Backman, “Imaging a full set of optical scattering properties of biological tissue by inverse spectroscopic optical coherence tomography,” Opt. Lett. |

10. | C. P. Fleming, J. Eckert, E. F. Halpern, J. A. Gardecki, and G. J. Tearney, “Depth resolved detection of lipid using spectroscopic optical coherence tomography,” Biomed. Opt. Express |

11. | R. N. Graf and A. Wax, “Temporal coherence and time-frequency distributions in spectroscopic optical coherence tomography,” J. Opt. Soc. Am. A |

12. | F. Robles, R. N. Graf, and A. Wax, “Dual window method for processing spectroscopic optical coherence tomography signals with simultaneously high spectral and temporal resolution,” Opt. Express |

13. | C. Xu, F. Kamalabadi, and S. A. Boppart, “Comparative performance analysis of time-frequency distributions for spectroscopic optical coherence tomography,” Appl. Opt. |

14. | R. Leitgeb, M. Wojtkowski, A. Kowalczyk, C. K. Hitzenberger, M. Sticker, and A. F. Fercher, “Spectral measurement of absorption by spectroscopic frequency-domain optical coherence tomography,” Opt. Lett. |

15. | A. Wax, C. Yang, and J. A. Izatt, “Fourier-domain low-coherence interferometry for light-scattering spectroscopy,” Opt. Lett. |

16. | L. Cohen, “Time-frequency distributions – a review,” Proc. IEEE |

17. | Data tabulated from various sources compiled by S. Prahl, http://omlc.ogi.edu/spectra. |

18. | N. Bosschaart, M. C. G. Aalders, T. G. van Leeuwen, and D. J. Faber, “Spectral domain detection in low-coherence spectroscopy,” Biomed. Opt. Express |

19. | J. Yi and X. Li, “Estimation of oxygen saturation from erythrocytes by high-resolution spectroscopic optical coherence tomography,” Opt. Lett. |

**OCIS Codes**

(030.1640) Coherence and statistical optics : Coherence

(070.4790) Fourier optics and signal processing : Spectrum analysis

(160.4760) Materials : Optical properties

(170.6510) Medical optics and biotechnology : Spectroscopy, tissue diagnostics

**ToC Category:**

Optical Coherence Tomography

**History**

Original Manuscript: August 12, 2013

Revised Manuscript: September 28, 2013

Manuscript Accepted: October 15, 2013

Published: October 23, 2013

**Citation**

Nienke Bosschaart, Ton G. van Leeuwen, Maurice C. G. Aalders, and Dirk J. Faber, "Quantitative comparison of analysis methods for spectroscopic optical coherence tomography," Biomed. Opt. Express **4**, 2570-2584 (2013)

http://www.opticsinfobase.org/boe/abstract.cfm?URI=boe-4-11-2570

Sort: Year | Journal | Reset

### References

- N. Bosschaart, T. G. van Leeuwen, M. C. G. Aalders, B. Hermann, W. Drexler, and D. J. Faber, “Spectroscopic low coherence interferometry”, Chapter 23 in Optical Coherence Tomography – Technology and Applications, W. Drexler and J.G. Fujimoto, eds. (Springer Berlin Heidelberg New York USA), 2nd edition (2013)
- J. M. Schmitt, S. H. Xiang, K. M. Yung, “Differential absorption imaging with optical coherence tomography,” J. Opt. Soc. Am. A 15(9), 2288–2296 (1998). [CrossRef]
- U. Morgner, W. Drexler, F. X. Kärtner, X. D. Li, C. Pitris, E. P. Ippen, J. G. Fujimoto, “Spectroscopic optical coherence tomography,” Opt. Lett. 25(2), 111–113 (2000). [CrossRef] [PubMed]
- C. Xu, J. Ye, D. L. Marks, S. A. Boppart, “Near-infrared dyes as contrast-enhancing agents for spectroscopic optical coherence tomography,” Opt. Lett. 29(14), 1647–1649 (2004). [CrossRef] [PubMed]
- F. E. Robles, C. Wilson, G. Grant, A. Wax, “Molecular imaging true-colour spectroscopic optical coherence tomography,” Nat. Photonics 5(12), 744–747 (2011). [CrossRef] [PubMed]
- N. Bosschaart, M. C. G. Aalders, D. J. Faber, J. J. A. Weda, M. J. C. van Gemert, T. G. van Leeuwen, “Quantitative measurements of absorption spectra in scattering media by low-coherence spectroscopy,” Opt. Lett. 34(23), 3746–3748 (2009). [CrossRef] [PubMed]
- N. Bosschaart, D. J. Faber, T. G. van Leeuwen, M. C. G. Aalders, “In vivo low-coherence spectroscopic measurements of local hemoglobin absorption spectra in human skin,” J. Biomed. Opt. 16(10), 100504 (2011). [CrossRef] [PubMed]
- N. Bosschaart, D. J. Faber, T. G. van Leeuwen, M. C. G. Aalders, “Measurements of wavelength dependent scattering and backscattering coefficients by low-coherence spectroscopy,” J. Biomed. Opt. 16(3), 030503 (2011). [CrossRef] [PubMed]
- J. Yi, V. Backman, “Imaging a full set of optical scattering properties of biological tissue by inverse spectroscopic optical coherence tomography,” Opt. Lett. 37(21), 4443–4445 (2012). [CrossRef] [PubMed]
- C. P. Fleming, J. Eckert, E. F. Halpern, J. A. Gardecki, G. J. Tearney, “Depth resolved detection of lipid using spectroscopic optical coherence tomography,” Biomed. Opt. Express 4(8), 1269–1284 (2013). [CrossRef] [PubMed]
- R. N. Graf, A. Wax, “Temporal coherence and time-frequency distributions in spectroscopic optical coherence tomography,” J. Opt. Soc. Am. A 24(8), 2186–2195 (2007). [CrossRef] [PubMed]
- F. Robles, R. N. Graf, A. Wax, “Dual window method for processing spectroscopic optical coherence tomography signals with simultaneously high spectral and temporal resolution,” Opt. Express 17(8), 6799–6812 (2009). [CrossRef] [PubMed]
- C. Xu, F. Kamalabadi, S. A. Boppart, “Comparative performance analysis of time-frequency distributions for spectroscopic optical coherence tomography,” Appl. Opt. 44(10), 1813–1822 (2005). [CrossRef] [PubMed]
- R. Leitgeb, M. Wojtkowski, A. Kowalczyk, C. K. Hitzenberger, M. Sticker, A. F. Fercher, “Spectral measurement of absorption by spectroscopic frequency-domain optical coherence tomography,” Opt. Lett. 25(11), 820–822 (2000). [CrossRef] [PubMed]
- A. Wax, C. Yang, J. A. Izatt, “Fourier-domain low-coherence interferometry for light-scattering spectroscopy,” Opt. Lett. 28(14), 1230–1232 (2003). [CrossRef] [PubMed]
- L. Cohen, “Time-frequency distributions – a review,” Proc. IEEE 77(7), 941–981 (1989). [CrossRef]
- Data tabulated from various sources compiled by S. Prahl, http://omlc.ogi.edu/spectra .
- N. Bosschaart, M. C. G. Aalders, T. G. van Leeuwen, D. J. Faber, “Spectral domain detection in low-coherence spectroscopy,” Biomed. Opt. Express 3(9), 2263–2272 (2012). [CrossRef] [PubMed]
- J. Yi, X. Li, “Estimation of oxygen saturation from erythrocytes by high-resolution spectroscopic optical coherence tomography,” Opt. Lett. 35(12), 2094–2096 (2010). [CrossRef] [PubMed]

## Cited By |
Alert me when this paper is cited |

OSA is able to provide readers links to articles that cite this paper by participating in CrossRef's Cited-By Linking service. CrossRef includes content from more than 3000 publishers and societies. In addition to listing OSA journal articles that cite this paper, citing articles from other participating publishers will also be listed.

« Previous Article | Next Article »

OSA is a member of CrossRef.