experimental confirmation of potential swept source optical coherence tomography performance...

Experimental confirmation of potential sweptsource optical coherence tomography

performance limitations

Kathy Zheng,1 Bin Liu,1,2 Chuanyong Huang,1,2 and Mark E. Brezinski1,2,*1Center for Optical Coherence Tomography and Optical Physics, Department of Orthopedic Surgery,

Brigham and Women’s Hospital, 75 Francis Street, Boston, Massachusetts 02115, USA2Harvard Medical School, Boston, Massachusetts 02115, USA

*Corresponding author: [email protected]

Received 28 May 2008; revised 30 September 2008; accepted 6 October 2008;posted 13 October 2008 (Doc. ID 96416); published 11 November 2008

Optical coherence tomography (OCT) has demonstrated considerable potential for a wide range of med-ical applications. Initial work was done in the time domain OCT (TD-OCT) approach, but recent interesthas been generated with spectral domain OCT (SD-OCT) approaches. While SD-OCT offers higher dataacquisition rates and no movable parts, we recently pointed out theoretical inferior aspects to its per-formance relative to TD-OCT. In this paper we focus on specific limitations of swept source OCT (SS-OCT), as this is the more versatile of the two SD-OCT embodiments. We present experimental evidenceof reduced imaging penetration, increased low frequency noise, higher multiple scattering (which can beworsened still via aliasing), increased need to control the distance from the sample, and saturation ofcentral bandwidth frequencies. We conclude that for scenarios where the dynamic range is relatively low(e.g., retina), the distance from the sample is relatively constant, or high acquisition rates are needed, SS-OCT has a role. However, when penetration remains important in the setting of a relatively high dynamicrange, acquisition rates above video rate are not needed, or the distance to the tissue is not constant, TD-OCT may be the superior approach. © 2008 Optical Society of America

OCIS codes: 170.4500, 290.9210, 170.3660, 110.4280.

1. Introduction

Optical coherence tomography (OCT) is now beingused to examine a wide range of human pathologies[1,2]. Based on low coherence interferometry, OCTallows micrometer scale imaging at or in excess ofvideo rate [3,4]. The original OCT studies were per-formed with time domain OCT (TD-OCT), while re-cently Fourier domain OCT (FD-OCT) and sweptsource OCT (SS-OCT), referred to collectively asspectral domain OCT (SD-OCT), have attracted aconsiderable amount of attention [1,5–9]. The theorybehind each has been reviewed elsewhere but, ingeneral for TD-OCT, an optical group delay is pro-duced in the reference arm, although some methods

work via a variant of this approach [1]. For SD-OCT,which generally has no moving parts, the whole spec-tral interferogram is recorded from the sample, anddepth/amplitude information is obtained via theFourier transform of the interferogram [1]. Amongthe major reasons for the recent attraction to SD-OCT are the high data acquisition rates and theclaimed superior performance with respect to detec-tion sensitivity. We recently demonstrated that, atleast with current embodiments, the theoretical per-formance in terms of dynamic range and penetrationof SD-OCT could be inferior to that of TD-OCT [1,10].In this work, experimental data further supportingthese conclusions is presented using a phantom ofconsistent composition and high scattering. In addi-tion, the deleterious effects of high multiple scatter-ing, complex conjugate ambiguity, and aliasing are

0003-6935/08/336151-08$15.00/0© 2008 Optical Society of America

20 November 2008 / Vol. 47, No. 33 / APPLIED OPTICS 6151

demonstrated for SS-OCT in a moderately scatteringsample. We conclude that for scenarios where thesample dynamic range is relatively low (ex: retina),the distance from the sample is relatively constant,or high acquisition rates are needed, SS-OCT may bethe superior embodiment. However, when penetra-tion remains important particularly in the setting ofa relatively high dynamic range, acquisition ratesabove video rate are not needed, or the distanceto the tissue is not constant, TD-OCT has distinctadvantages.

2. Methods

The focus of this paper is to expand upon SS-OCTlimitations identified in our previous work, particu-larly providing supporting experimental evidence forthe conclusions [10]. Since direct comparisons aremade between SD-OCT and TD-OCT, schematics ofboth SS-OCT and TD-OCT are shown in Fig. 1.The principles of both are described elsewhere in de-tail [1,10]. Spectral radar or FD-OCT, whose dynamic

range is the least of the three embodiments andwhose use is the most limited for reasons discussedpreviously, is not addressed here [10]. As stated, TD-OCT performs ranging primarily by mechanically in-troducing an optical group delay in the referencearm. With SS-OCT, the source is frequency sweptacross its bandwidth, and the total interferogramfor each A scan is obtained. Several aspects of SS-OCT imaging are relevant to expand upon. First,with SS-OCT, the power on the sample at any givenfrequency is higher than TD-OCT because the fre-quencies are swept. The importance of this is shownin Section 4, particularly with respect to multiplescattering and saturation of the central frequenciesvia the detector/preamplifier. In the latter case, thisresults in only the off-center frequencies being de-tected and Fourier transformed, resulting in a sinu-soidal artifact. Second, while the reference armmirror does not move, the path length in the refer-ence arm is kept shorter than that in the samplearm (z offset, typically 500 μm), which is designedto better isolate the autocorrelation function [1,7].The fact that fixing the mismatch at a specific dis-tance may not be ideal for imaging is discussed inSection 4. Third, the system is low pass, so low fre-quency noise is less effectively removed than in TD-OCT, reducing performance [1,10].

We have previously defined our use of terms suchas dynamic range, signal to noise ratio (SNR), andpenetration to reduce ambiguity, a problem in theliterature [10]. Defining these terms is critical, as un-conventional definitions have led to significant mis-interpretation of many past publications. They arealso briefly reviewed here:

Dynamic range: In a system or device, the ratio of aspecified maximum level of a parameter, such aspower, current, voltage, or frequency, to the mini-mum detectable value of that parameter, is usuallyexpressed in dB. Here, we are interested in the dy-namic range of the final digitized image as it corre-lates with penetration in nontransparent tissue.

Signal to noise ratio (SNR): This is the ratio of theamplitude of the maximal signal to the amplitude ofnoise signals at a given point in time. SNR is usuallyexpressed in dB and in terms of peak values for im-pulse noise and root-mean-square values for randomnoise. In defining or specifying the SNR, both the sig-nal and noise should be characterized (e.g., peaksignal-to-peak noise ratio) to avoid ambiguity. Photocurrent is the detected parameter in TD-OCT andSS-OCT, whereas photo-generated charge numbersare the detected parameter in FD-OCT. The differentparameters do not affect the SNR if the parametersare appropriately defined. Here SNR describe theanalog or radio frequency (RF) signal prior to digiti-zation, which can be a point of confusion in theliterature. Furthermore, groups have termed the de-scribed parameter the dynamic range, which againcan be a considerable source of misinterpretation.

Sensitivity: In an electronic device (e.g., a commu-nications system receiver) or detection device (e.g.

Fig. 1. Schematics of both TD-OCT and SS-OCT are shown.Images courtesy of [1].

6152 APPLIED OPTICS / Vol. 47, No. 33 / 20 November 2008

PIN diode), the minimum input signal required toproduce a specified output signal having a specifiedSNR, or other specified criteria is termed “sensitiv-ity”. With OCT, the SNR is often chosen as 1. So theminimum input signal, usually defined as the mini-mum sample reflection intensity, equals the amountof signal that can produce the same amount of outputas noise. As discussed below, image penetration is thecontrast as a function of depth, dependent more onthe dynamic range than on sensitivity, as opposedto other technologies. In other words, effective OCTimaging is not measuring the total backscatteredphotons but the contrast between the point spreadfunction (PSF) at different spatial points.Decibel: We define decibel as

decibels ¼ 10 log x: ð1ÞHowever, this can be confusing since some authorsuse a definition found in the electronics literature:

decibels ¼ 20 log x: ð2ÞThe reasons for this are discussed elsewhere [10].Therefore, this can result in the second yieldingSNR and dynamic range values twice that of the firstfor the same system.Resolution: The resolution here is defined as the

FWHM of the point spread function measured froma totally reflective surface in a vacuum (air for ap-proximation).Contrast: There are many definitions of contrast,

which are distinct from resolution [1]. It is contrastat depth which defines penetration. One methodfor calculating the contrast C is using the standardformula

C ¼ f − bf þ b

; ð3Þ

where f is the mean (average) gray-level value of aparticular object in an image (foreground) and b isthe mean gray-level value of the surrounding regionmedia (background).Penetration: In describing penetration with OCT,

we are not discussing individual photon penetrationbut the contrast as a function of depth. Essentially,we are looking at the ability to discriminate struc-ture and not assessing backreflected absolute photoncounts. This penetration is determined by the dy-namic range, multiple scattering, resolution, and, toa lesser degree, total power. For OCT, high sensitivitywill allow photon numbers of lower intensity to bedetected, but is not equivalent to image contrast asa function of depth, as either low dynamic range orhigh multiple scattering would yield poor imagingpenetration in spite of the increased photon counts.Both TD-OCT and SS-OCT systems used are com-

mercially available. These commercial systems werechosen because, with the extensive amount of engi-neering put forward in their development, they ap-pear to have optimized performance to a point at

least as good as anything quoted in the literature,particularly the detection electronics [11,12]. Systemparameters between each modality were matched asclosely as possible. The Lightlab imaging engine(Westford, Massachusetts) (TD-OCT) used has amedian wavelength of 1287nm, an axial resolution of13 μm, total power of 13mW, and an A-scan rate of3:13kHz [11]. The OCM1300SS SS-OCT system(Thorlabs, Newton, New Jersey), was used for SS-OCT imaging [12]. This system had a median wave-length of 1325nm, an axial resolution of 12 μm, totalpower of 10mW, and an A-scan rate of 16kHz. Thespot size at the beam waist for the TD-OCT systemis 24 μm, compared with 15 μm for the SS-OCT sys-tem, which corresponded to confocal parameters of0.7 and 0:3mm, respectively.

3D reconstructions of samples were made duringpostprocessing using Image J. The saved JPEGstream was converted to gray scale. The parametersset were a projection method of the brightest point,and a rotation angle increment of 3°. The reconstruc-tions were saved as MPEG movies.

Choosing the appropriate sample to assess SS-OCT and TD-OCT reliably is challenging. Tissueswith high dynamic range requirements, such as cor-onary atherosclerotic plaque and gastric mucosa, arevery heterogeneous [3,13–15]. This would undoubt-edly leave the results subject to criticisms that differ-ences are solely the result of this variability. Wetherefore elected to use a tissue engineering scaffoldwith a well-defined biological composition, poly(lactide-co-glycolide) (PLGA) (Boehringer Ingelheim,Ingelheim, Germany) [16,17]. It is also known to be arelatively high scatterer, leading to a large dynamicrange for achieving maximum penetration. Thispolymer consists of biomolecules present in the bodybut in a well-defined configuration, making it anideal phantom. While the scattering coefficient hasnot been published, it is known to have a penetrationof 300 μm by confocal microscopy and minimal ab-sorption, supporting the assertion that it is represen-tative of tissue with moderate to high dynamic range[18]. The scaffolding was selected based on experi-ence with a wide range of materials, both biologicaland nonbiological. The scaffolding has the advantageof having homogeneous composition as well as repro-ducible structure and a high dynamic range. A totalof five scaffolds were examined with a minimum of200 B scans obtained per sample. The scaffolds weresubmersed in normal saline. The scaffold bundlewidths by OCT were assessed by the average FWHM� the standard error (SE) at the surface and at 1mmdepth (n ¼ 6 for each of the five samples). The drop inintensity for both technologies at 1mm was obtainedby dividing the peak intensity at that depth by theintensity at the surface (n ¼ 6 for each of the fivesamples). Data is expressed as percent reduction �the SE. Statistical significance was assessed usinga pairwise Student’s t-test.

Focusing issues are also addressed here.While thistopic does not involve limitations in dynamic range


or penetration, there are difficulties with SD-OCTthat do not occur with TD-OCT, adding further chal-lenges to the use of the technology. To demonstratedifficulties with focusing, particularly the generationof inverted images (complex conjugate ambiguity), abovine meniscus was used as heterogeneity was notproblematic in identifying the effect [19,20]. The me-niscus was obtained within three hours postmortemand kept in normal saline prior to imaging.

3. Results

Direct comparisons are made of SS-OCT with TD-OCT on the PLGA scaffolding. 2D images are shownof the scaffolding imaged with both the TD-OCT andSS-OCT systems in Fig. 2. In the TD-OCT images,penetration extends beyond four layers, with thescaffolding bundles being easily resolvable at thisdepth (over 1:5mm) while results were substantiallyworse with SS-OCT. A scanning electron microscopymicrograph of the image is shown in Fig. 3 [21]. Theporous structure and layering of the scaffoldingis shown, where the pore size ranges from 100 to250 μm. The intensity dropped 24%� 2% at 1mmfor TD-OCT versus 46%� 3% over the same distancefor SS-OCT (p < 0:001). With respect to the FWHMof the bundles at the surface, for TD-OCT they were0:17mm� 0:05mm versus 0:24mm� 0:05mm forSS-OCT (p < 0:05). At 1mm, the bundle widths were0:24mm� 0:06mm for TD-OCT versus 0:35mm�0:08mm for SS-OCT (p < 0:005). In addition to re-duced penetration and FWHM, multiple artifactsare noted in the SS-OCT images that affect perfor-mance. The mechanisms are theorized below butbriefly: the middle arrow in the SS-OCT image showsdetector saturation of the central frequencies, the leftarrow shows heavy multiple scattering, the lowerright and farthest left arrow illustrates aliasing ofthe multiply scattered light, and the lower right cor-ner arrow points to a loss of structural detail likelyfrom limited dynamic range (resulting from a combi-nation of mechanisms as described below).The effect seen by the insert and the line shown by

the red arrow is well known to those using SD-OCTand occurs primarily at sites of high reflection. Theimportant aspect of this line is shown in the insert ofthe figure, which is that it is sinusoidal in nature andnot constant in intensity. The relevance of this is dis-cussed in Section 4.

In the 3D images shown in Fig. 4, the scaffoldingappears artificially wide or blurred out for SS-OCTand structural definition is lost less than 1mm intothe sample. Qualitatively, structural detail is seenfar deeper and is more sharply defined with TD-OCT.

Figure 5(a) shows SS-OCT imaging with the focuson the surface of a bovine meniscus. The bovinemeniscus has a smooth surface and discrete bandingpatterns from birefringence [19,22]. The arrows iden-tify the tissue surface. However, when the focus hasbeen 1mm below the surface (as shown in Fig. 5(b),the image becomes severely distorted by the invertedimage of the surface (complex conjugate ambigu-ity) [23].

SS-OCT is a low pass system that does not removeas effectively low frequency noise sources such as 1=fand thermal noise in the detector compared to TD-OCT, as we have previously described [1,10,24].

Fig. 2. (Color online) 2D images are shown of the scaffolding im-aged with TD-OCT (left) and SS-OCT systems (right).

Fig. 3. Scanning electron microscopy micrographs of the porousstructure of PLGA scaffolding. The pore size ranges from 100 to250 μm.

Fig. 4. Still frames of the 3D projection of PLGA scaffolding usingTD-OCT (left) and SS-OCT (right).


Figure 6 demonstrates low frequency noise that de-grades the image and is noted most obviously in thehigh background signal. The postprocessing removalof noise results in a further reduction of dynamicrange.

4. Discussion

Technological advances with SD-OCT represent oneof the most important progresses in OCT researchover the last decade, particularly its high data acqui-sition rate relative to TD-OCT. However, as we havepointed out in a previous publication [10], theoreticalanalyses as well as evaluation of data in the litera-ture indicate that current embodiments have sub-stantial detection limitations when compared toTD-OCT under the appropriate circumstances. Wedemonstrated that much of the previous work claim-ing essentially universally superior detection cap-

abilities compared with TD-OCT are misleading, aseither the experimental conditions (e.g., tissue usedof relatively low dynamic range) or the inconsistentdefinitions of parameters result in comparison lim-itations. In the current work, by using a tissueengineering scaffold (PLGA), chosen because of thehomogenous organization of its biological constitu-ents and high scattering (i.e., large dynamic range),we are able to demonstrate in a relatively straight-forward manner some of these limitations. The syn-thetic scaffold was used rather than human tissue toprevent variability and the criticism that differenceswere due to tissue heterogeneity. The literaturequotes the average PLGA fiber size of 12 μm, whichis consistent with the 17� 5 μm found with TD-OCT[25]. The slightly greater PSF of SS-OCT at the sur-face is consistent with multiple scattering, even withthe superficial bundles seen in the images. Thelimitations of SS-OCT for the samples examined in-clude a statistically significant reduced penetration,statistically significant reduced resolution with pe-netration, increased multiple scattering, aliasing ofmultiply scattered light, central wavelengths satura-tion, high background noise, and mirroring from ne-gative z offsets (complex conjugate ambiguity).

The lines identified by the farthest left arrows inFig. 2 are well known to those using SS-OCT, parti-cularly when imaging a highly reflective surface.They do not occur with TD-OCT. The localized struc-ture of this line has been demonstrated more clearlyin an enlargement in the picture in right bottom cor-ner of Fig. 2. Analysis of the data and an understand-ing of the underlying theory behind SS-OCT stronglysuggest a mechanism. The phenomenon is mostlikely due to central wavelength saturation or “chop-ping out” of frequencies in the backreflected light byintensity limitations of the detector/preamplifier.When these lines are examined at high resolution,consistent with theory, it is noted that the linesare not DC in intensity but are actually sinusoidal.When saturation occurs with TD-OCT, blindness re-sults in that, for a given A scan, the detector/preamplifier is exposed to ALL frequencies at once,resulting in complete loss of signal from the reflectorand typically the area behind it. TD-OCT then ana-lyzes backreflected data detected time sequentially.If a backreflected signal from a place in the tissueis above the saturation point of the detector, thattime point (and subsequent recovery time of the de-tector) will be lost from the image. In other words, alocalized collection of pixels in the vicinity of the sa-turating surface is “blinded”. In contrast, with SS-OCT, the frequencies are swept across the sample,and then all frequencies are used to reproduce theA scan via the fast Fourier transform (FFT). There-fore, if the central frequencies of the swept radiationare above the threshold of the detector/preamplifier(i.e., the center frequencies are removed), only theywill be lost and those at the edges of the spectrumwill be processed (i.e., again the source is swept).The FFT is therefore only performed on these

Fig. 5. (a) SS-OCT imaging with the focus on the surface of a bo-vine meniscus. The bovine meniscus has a smooth surface and dis-crete banding pattern from birefringence. The arrows identify thetissue surface. (b) When the focus has been moved 1mm below thesurface, the image becomes completely distorted by the invertedimage of the surface (complex conjugate ambiguity).

Fig. 6. This figure demonstrates low frequency noise that is notremoved in the detection electronics because SS-OCT is a low passsystem. Postprocessing to reduce this noise is required, which de-grades the image by removing the low frequency components of theautocorrelation function.


frequencies, which are well below and above the cen-tral frequencies. This principle is seen in Fig. 7. Whatresults is a quasi-delta function pair, the FFT ofwhich is almost sinusoidal (rather than DC) in nat-ure [26]. This is the same pattern seen in the SS-OCTimages. Since the sides are not true delta functionsbut have a finite width, the FFT is not purely sinu-soidal. This phenomenon is a function of intensity(does not occur below saturation) and does not occurwith TD-OCT.Because SS-OCT sweeps the bandwidth in small

frequency intervals, the amount of power that can beand is delivered at a given frequency is much higherthan with TD-OCT. In other words, the interfero-gram (single A scan) is determined from the FFTof a spectrum with much higher powers for eachcontributing frequency. This can result not only incentral frequency saturation, but also increased mul-tiple scattering [27–29]. Since multiple scatteringincreases much faster than penetration with power,

a well described phenomenon, the multiple scatter-ing deleterious effects outweigh any benefits pene-tration power has [28,29]. This is represented bythe farthest left arrow in Fig. 2, where the classictailing effect (shower curtain effect) [30] after thescatterer is seen due to multiple scattering by thedense scaffold. This results in reductions of both re-solution and penetration.

The z offset between the reference and sample arm(target surface) is typically set around 500 μm, withthe reference arm having the shorter length. Whenthe z offset between the target and reference arm in-creases, such as with deep structures in the sample,higher frequencies are present in the backreflectedsignal, and their capture is required for correct re-construction of the A scan. Therefore, resolving dee-per structures can be ultimately limited by thefrequency sensitivity of the detector/preamplifier. Ifthe detection of higher frequencies fails, aliasingand distortion of the PSF will occur. This is a phe-nomenon that does not occur with TD-OCT. In addi-tion to distorting normal structure, the aliasing ordistortion of the multiple scattering can also occur(right arrow in Fig. 2), based on the same principles,which further degrades the image [29].

As has been pointed out in the original theoreticalanalysis, two of the reasons that SD-OCT has a re-duced dynamic range are that it is a low pass systemand it does not undergo logarithmic demodulationbefore A–D conversion. The fact that SD-OCT islow pass results in low frequency noise, such as ther-mal and 1=f noise, being mixed with the autocorrela-tion function. This is seen in the background of Fig. 6.The low frequency noise is generally removed withdigital postprocessing, but this results in a furtherreduction in the dynamic range of the image as lowfrequency components of the autocorrelation func-tion are also removed. With TD-OCT, the interfero-gram is Doppler shifted to higher frequencies bymechanical movement in the reference arm, allowinga bandpass filter to be introduced after the detectorto reduce low frequency noise.

The other major reason for reduced dynamic rangeis the lack of logarithmic demodulation before A–D[1,10]. The A–D conversion and its respective signalloss are different for TD-OCT and SD-OCT. The max-imum level of the quantizer is 2MΔ, where M is thebits of an A–D converter andΔ is the intervals of dis-crete levels. In TD-OCT, digital processing is appliedto the analog autocorrelation A scan, which allowslog 10 demodulation to maintain a high dynamicrange. In a SD-OCT system, digital Fourier trans-form is conducted to calculate the A-scan signal fromthe quantized spectral interferogram. Prior to A–Dconversion, in TD-OCT, the RF signal undergoeslogarithmic demodulation. This leads to a theoreticalmaximal dynamic range near 80dB. In contrast, withSD-OCT, the RF signal does not undergo logarithmicdemodulation in current embodiments prior to A–Dconversion, resulting in a maximum potential (ex-cluding other sources of loss) dynamic range of less

Fig. 7. A demonstration that, if the central frequencies of theswept radiation are above the threshold of the detector (i.e., thecenter frequencies are removed), the FFT is only performed onthe frequencies well below and above the center, as can be seenin the figure. This FFT of the remaining two peaks, if their widthis sufficiently small, produces a sinusoidal like pattern.


than 40dB. For tissue with a low dynamic range,such as retina, this does not represent a major pro-blem. However, if the backreflected signal intensityis greater than 30dB between the surface and struc-tures deeper within the sample, penetration can belimited by this lower dynamic range.In Fig. 2, the ability to resolve structures greater

than 500 μm below the surface is severely compro-mised in SS-OCT compared with TD-OCT. This isconsistent with statistically significant reductionsin the point spread function as well as intensity withSS-OCT relative to TD-OCT. The scaffolding was cho-sen because of its homogeneity and high scattering(i.e., dynamic range) [18]. Therefore, the differencein performance between the two embodiments onthis medium, which was statistically significant at1mmpenetration, can be assumed to occur for a com-bination of several reasons as discussed above andare consistent with the hypothesis of this paper.These include reduced dynamic range, low pass nat-ure of the system, substantial multiple scattering,and potentially the large z offset resulting in aliasing(including aliasing of multiply scattered light).Figure 5 demonstrates mirroring or complex con-

jugate ambiguity, which does not deal directly witheither dynamic range or penetration, but representsa significant disadvantage of SD-OCT compared withTD-OCT. This occurs because, as a Fourier transformon any real-valued function (in the direct space) isHermitian symmetrical around the origin (in thetransform space), it generates a complex conjugatepart that mirrors with the targeted A scan aroundthe zero-optical-path-mismatching point. This couldlead to the overlapping of the targeted A scan and itsinverted/ghost image if the sample is too close to thezero-optical-path-mismatching point, or even if it in-tersects that point. In 5(a), an image of bovine menis-cus is presented where the tissue surface representsa 500 μm z offset from the reference arm. The surfaceas well as the banding pattern from cartilage bire-fringence is sharply defined. However, the complexconjugate ambiguity is noted when the focus is toofar into the tissue (reversed z offset) as a mirrorimage of the surface will develop. This is seen inFig. 5(b) and results because of the need for a shorterlength in the reference arm compared with the sam-ple arm because the surface is brought too close. Thefact that the point of focus cannot be too close, result-ing in image inversion, or too far, resulting in alias-ing and loss of resolution, makes it difficult toevaluate how it will be easily implemented in situa-tions where the tissue is moving, image depth is suf-ficiently large, or for radial images where a largeluminal diameter exists. Three major approaches at-tempting to rectify this problem, aiming on construc-tion of complex spectral interferograms, have beenput forward to reduce these issues, all of which arefor FD-OCT, though potentially applicable to SS-OCT. However, they all involve manipulation ofthe reference arm, either through mechanical move-ment or modulation such as electro-optically

[23,31,32]. However, they have only been of limitedsuccess and the addition of reference arm manipula-tion makes the techniques far from attractive and, atthe very least, reduces acquisition rate.

A significant limitation of this study was that,while very close with respect to optical parameters,the systems were not identical. In particular, therewas a difference in the confocal parameters betweenthe two embodiments. This means that the spot sizewill increase in diameter more rapidly away from thebeam waist for SS-OCT in this experiment comparedwith the TD-OCT setup. While this does not signifi-cantly affect axial resolution, central bandwidth sa-turation, low frequency noise, dynamic range, largez-offset frequency loss, or complex conjugate ambigu-ity, it can lead to a degradation of lateral resolutionand potentially some increase in multiple scattering.Although this is not likely to represent a majorproblem in data interpretation due to the fact the dif-ference is not substantially large and performancelimitations described are occurring at the beamwaist(in addition to off focus), this potential limitation ofthe study needs to be mentioned.

5. Conclusion

SS-OCT in its current embodiments, while having afaster data acquisition rate, has reduced perfor-mance relative to TD-OCT under the appropriate cir-cumstances. These limitations include a reduceddynamic range, increased multiple scattering, alias-ing of multiple scattering, loss of resolution withdepth due to a large z offset (high frequency loss),central wavelengths saturation, low frequency noisevulnerability, and complex conjugate ambiguity.Therefore, when penetration remains important inthe setting of a relatively high dynamic range, acqui-sition rates above video rate are not needed, or thedistance to the tissue is not constant, TD-OCT hasdistinct advantages. However, where the sampledynamic range is relatively low (e.g., retina), the dis-tance from the sample is relatively constant, or highacquisition rates are needed, SS-OCT may representthe superior embodiment.

M. E. Brezinski’s work is currently funded byNational Institute of Health (NIH) Grants R01AR44812, R01 HL55686, R01 EB02638/HL63953,R01 AR46996, and R01 EB000419. The authors alsothank Alicia N. Goodwin, Namita P. Kumar, andJulie A. Williams for their technical support. Theauthors have no financial conflicts with regard tothe data published in this paper.

References1. M. E. Brezinski, Optical Coherence Tomography, Principle

and Practice (Academic, 2006).2. M. E. Brezinski and J. G. Fujimoto, “Optical coherence tomo-

graphy: high-resolution imaging in nontransparent tissue,”IEEE J. Sel. Top. Quantum Electron. 5, 1185–1192 (1999).

3. M. E. Brezinski, G. J. Tearney, B. E. Bouma, J. A. Izatt,M. R. Hee, E. A. Swanson, J. F. Southern, and J. G. Fujimoto,“Optical coherence tomography for optical biopsy—properties


and demonstration of vascular pathology,” Circulation 93,1206–1213 (1996).

4. G. J. Tearney, M. E. Brezinski, B. E. Bouma, S. A. Boppart,C. Pitris, J. F. Southern, and J. G. Fujimoto, “In vivo endo-scopic optical biopsy with optical coherence tomography,”Science 276, 2037–2039 (1997).

5. A. F. Fercher, C. K. Hitzenberger, G. Kamp, and S. Y. Elzaiat,“Measurement Of intraocular distances by backscatteringspectral interferometry,” Opt. Commun. 117, 43–48(1995).

6. A. Wax, C. H. Yang, and J. A. Izatt, “Fourier-domain low-coher-ence interferometry for light-scattering spectroscopy,” Opt.Lett. 28, 1230–1232 (2003).

7. S. H. Yun, G. J. Tearney, B. E. Bouma, B. H. Park, and J. F. deBoer, “High-speed spectral-domain optical coherence tomogra-phy at 1:3 μm wavelength,” Opt. Express 11, 3598–3604(2003).

8. S. R. Chinn, E. A. Swanson, and J. G. Fujimoto, “Optical co-herence tomography using a frequency-tunable opticalsource,” Opt. Lett. 22, 340–342 (1997).

9. M. A. Choma, M. V. Sarunic, C. H. Yang, and J. A. Izatt, “Sen-sitivity advantage of swept source and Fourier domain opticalcoherence tomography,” Opt. Express 11, 2183–2189(2003).

10. B. Liu and M. E. Brezinski, “Theoretical and practical consid-erations on detection performance of time domain, Fourier do-main, and swept source optical coherence tomography,”J. Biomedical Optics 12, 044007 (2007).

11. http://www.lightlabimaging.com/12. http://www.thorlabs.com/13. H. Yabushita, B. E. Bouma, S. L. Houser, H. T. Aretz,

I.-K. Jang, K. H. Schlendorf, C. R. Kauffman, M. Shishkov,D.-H. Kang, E. F. Halpern, and G. J. Tearney, “Characteriza-tion of human atherosclerosis by optical coherence tomogra-phy,” Circulation 106 (13), 1640–1645 (2002).

14. C. Pitris, C. Jesser, S. A. Boppart, D. Stamper, M. E. Brezinski,and J. G. Fujimoto, “Feasibility of optical coherence tomogra-phy for high resolution imaging of human gastrointestinaltract malignancies,” J. Gastroenterol. 35, 87–92 (2000).

15. G. Zuccaro, N. Gladkova, J. Vargo, F. Feldchtein, E. Zagaynova,D. Conwell, G. Falk, J. Goldblum, J. Dumot, J. Pnsky,G. Gelikonov, B. Davros, E. Donchenko, and J. Richter, “Opticalcoherence tomography of the esophagus and proximalstomach in health and disease,” Am. J. Gastroenetrol. 96,2633–2639 (2001).

16. V. Karageorgiou and D. Kaplan, “Porosity of 3D biomaterialscaffolds and osteogenesis,” Biomaterials 26, 5474–5491(2005).

17. S. R. Frenkel and P. E. Di Cesare, “Scaffolds for articular car-tilage repair,” Ann. Biomed. Eng. 32, 26–34 (2004).

18. S. L. Ishaug Riley, G. M. Crane-Kruger, M. J. Yaszemski, andA. G. Mikos, “Three-dimensional culture of rat calvarial osteo-

blasts in porous biodegradable polymers,” Biomaterials 19,1405 (1998).

19. B. Liu, M. Harman, S. Giattina, D. L. Stamper, C. Demakis,M. Chilek, S. Raby, and M. E. Brezinski, “Characterizing oftissue microstructure with single-detector polarization-sensitive optical coherence tomography,” Appl. Opt. 45,4464–4479 (2006).

20. M. Wojtkowski, A. Kowalczyk, R. Leitgeb, and A. F. Fercher,“Full range complex spectral optical coherence tomographytechnique in eye imaging,” Opt. Lett. 27, 1415–1417 (2002).

21. T. Ren, J. Ren, and K. Pan, “The bone formation in vitro andmandibular defect repair using PLGA porous scaffolds,”J Biomed Mater Res A 74, 562–569 (2005).

22. W. Drexler, D. Stamper, C. Jesser, X. Li, C. Pitris, K. Saunders,S. Martin, M. B. Lodge, J. G. Fujimoto, and M. E. Brezinski,“Correlation of collagen organization with polarization sensi-tive imaging in cartilage: implications for osteoarthritis,”J. Rheumatol. 28, 1311–1318 (2001).

23. R. K. K. Wang, “In vivo full range complex Fourier domain op-tical coherence tomography,” Appl. Phys. Lett. 90, 054103(2007).

24. P. Horowitz and W. Hill, The Art of Electronics, 2nd ed.(Cambridge University, 1997), pp. 614–617.

25. H. Kin and H. W. Kim, “Sustained release of ascorbate-2-phosphate and dexamethasone from porous PLGA scaffoldsfor bone tissue engineering using mesenchymal stem cells,”Biomaterials 24, 4671–4679 (2003).

26. J. W. Goodman, Introduction to Fourier Optics, 2nd ed.(McGraw Hill, 1996).

27. D. J. Smithies, T. Lindmo, C. Zhongping, J. S. Nelson, andT. E. Milner, “Signal attenuation and localization in optical co-herence tomography studied by Monte Carlo simulation,”Phys. Med. Biol. 43, 3025–2044 (1998).

28. G. Yao and L. V. Wang, “Monte Carlo simulation of an opticalcoherence tomography signal in homogeneous turbid media,”Phys. Med. Biol. 44, 2307–2320 (1999).

29. R. A. Leitgeb, W. Drexler, A. Unterhuber, B. Hermann,T. Bajraszewski, T. Le, A. Stingl, and A. F. Fercher, “Ultrahighresolution Fourier domain optical coherence tomography,”Opt. Express 12, 2156–2165 (2004).

30. L. Thrane, H. T. Yura, and P. E. Andersen, “Analysis ofoptical coherence tomography systems based on the extendedHuygens–Fresnel principle,” J. Opt. Soc. Am. A 17, 484–490(2000).

31. J. Zhang, J. S. Nelson, and Z. P. Chen, “Removal of a mirrorimage and enhancement of the signal-to-noise ratio inFourier-domain optical coherence tomography using anelectro-optic phase modulator,” Opt. Lett. 30, 147–149 (2005).

32. A. M. Davis, M. A. Choma, and J. A. Izatt, “Heterodyne swept-source optical coherence tomography for complete complexconjugate ambiguity removal,” J Biomed. Opt. 10,064005 (2005).


experimental confirmation of potential swept source optical coherence tomography performance...

Documents