Methods, systems, and computer program products for performing real-time quadrature projection based fourier domain optical coherence tomography

Description of equivalent WO 2007109127 A2

DESCRIPTION

METHODS, SYSTEMS, AND COMPUTER PROGRAM PRODUCTS FOR

PERFORMING REAL-TIME QUADRATURE PROJECTION BASED

FOURIER DOMAIN OPTICAL COHERENCE TOMOGRAPHY

GOVERNMENT RIGHTS

0006The present subject matter was made with United States Government support under Federal Grant No. EY013516 awarded by National Eye Institute

0007(NEI). The United States Government may have certain rights to this invention.

0008CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of U.S. Provisional Patent Application Serial No. 60/782,915, filed on March 16, 2006. The content of this provisional application is incorporated herein by reference in its entirety.

TECHNICAL FIELD

0010The subject matter described herein relates to the utilization of Fourier domain (FD) transforms and their application to the fields of interferometry and optical coherence tomography (OCT). More particularly, the subject matter described herein relates to performing real-time, quadrature projection based

0011FDOCT.

0012BACKGROUND .

0013The development of optical coherence tomography (OCT) in recent years has concentrated on Fourier domain (FD) techniques for high speed cross-sectional imaging of biological tissue. Namely, FD techniques provide increased signal-to-noise ratio (SNR) and increased robustness over traditional OCT techniques. The SNR advantage of FDOCT techniques may be employed for faster image acquisition, thereby enabling practical three-dimensional OCT imaging in living subjects. In. FDOCT, the locations of scatterers within a sample are obtained by Fourier transformation of real-valued spectral interferograms, which are generated, by mixing light backscattered from the sample with reference light. The Fourier transform of the interferogram is Hermititan symmetric, thereby introducing a complex conjugate artifact in which positive and negative distances are not resolved. In practice, this symmetric artifact may be avoided by locating the sample entirely within the positive or negative displacement range, thus utilizing only one half of the potential total imaging depth. Such one sided imaging is suitable for thin objects, but imaging of extended objects is limited by the characteristic roll-off in sensitivity that is typically associated with the finite spectral resolution of FDOCT systems.

0014Full-range imaging, in which positive and negative distances are resolved, can be achieved by indirectly measuring the complex component of the interferometric OCT signal using techniques borrowed from phase shift interferometry. The imaginary component of the interferogram is obtained by shifting the phase of the reference reflection in increments of 90 degrees. Phase shifting has been demonstrated in spectrometer-based FDOCT systems using a discretely stepped piezoelectric transducer (PZT) mounted reflector or an electro-optic modulator in the reference arm. One drawback of sequentially shifting the interferogram is that significant image corruption results from small deviations (e.g., chromatic deviations) in the actual phase shift obtained or from small (i.e., sub-wavelength) sample motion between the phase shifted acquisitions. Recently, the instantaneous acquisition of two phase shifted signals was demonstrated using linearly polarized light. This technique was limited by having only two phase shifted signals (which limits complex signal reconstruction) as well as by potential image corruption present in birefringent samples. Methods to instantaneously acquire three phase-shifted interferograms using 3x3 fused fiber couplers for both spectrometer-based and swept source (SS) FDOCT systems have also been employed. However, the performance of these types of systems for complex conjugate image reconstruction may be limited by the wavelength dependence of the splitting ratios associated with the fiber couplers. In addition, an approach to full range imaging based on frequency shifting may be considered, but this method would not compatible with spectrometer based systems.

0015Notably, all of the aforementioned phase shifting FDOCT techniques suffer image corruption arising from the mis-calibration of the phase shifts as well as from the wavelength dependence of the phase shifter. Numerical techniques to improve the suppression of symmetric artifact by compensating for the phase shift irregularities and for by accounting for axial sample motion in between phase shifted acquisitions have been previously been presented. However, there have not been any methods that address the removal of the complex conjugate artifact present in FDOCT images.

0016Accordingly, there exists a need for an effective method for eliminating the complex conjugate artifact in FDOCT images.

SUMMARY

0018The subject matter described herein includes methods, systems, and computer program products for performing real-time quadrature projection based FDOCT. One method includes phase shifting a plurality of interferogram signals. A Fourier transform is applied to each of the plurality of interferogram signals. Depth dependence of the plurality of transformed interferogram signals is then removed. A real quadrature component and an imaginary quadrature component for each of the plurality of transformed interferogram signals are subsequently calculated. The real quadrature components of the transformed interferogram signals are combined to obtain a derived real component and the imaginary quadrature components of the transformed interferogram signals are combined to obtain a derived imaginary component. A full-range depth profile of the object is constructed by adding the derived real component to the product of the derived imaginary component and a scaling factor. A full-range depth image of the object is then generated using the full-range depth profile. The subject matter described herein for performing real-time quadrature projection based FDOCT may be implemented in hardware, software, firmware, or any combination thereof. As such, the terms "function" or "module" as used herein refer to hardware, software, and/or firmware for implementing the feature being described. In one exemplary implementation, the subject matter described herein may be implemented using a computer program product comprising computer executable instructions embodied in a computer readable medium. Exemplary computer readable media suitable for implementing the subject matter described herein include disk memory devices, chip memory devices, programmable logic devices, application specific integrated circuits, and downloadable electrical signals. In addition, a computer program product that implements the subject matter described herein may be located on a single device or computing platform or may be distributed across multiple devices or computing platforms.

BRIEF DESCRIPTION OF THE DRAWINGS

0020Preferred embodiments of the subject matter described herein will now be explained with reference to the accompanying drawings of which: Figure 1 is a diagram illustrating an exemplary OCT interferometry system according to an embodiment of the subject matter described herein;

0021Figure 2 is a flow chart depicting an exemplary process for performing quadrature projection processing according to an embodiment of the subject matter described herein; Figure 3 is a diagram depicting phase-shifted interferogram signals according to an embodiment of the subject matter described herein;

0022Figure 4 is a diagram depicting Fourier transformed signals that are represented vectorally in the complex plane according to an embodiment of the subject matter described herein; Figure 5 is a diagram depicting the projection of vectors onto the real and imaginary axes according to an embodiment of the subject matter described herein;

0023Figure 6 is a diagram depicting the derived real and imaginary signals according to an embodiment of the subject matter described herein; Figure 7 is a diagram depicting the cancellation of symmetric complex conjugate artificial peaks to determine the full range depth profile according to an embodiment of the subject matter described herein;

0024Figure 8 is a chart depicting a complex conjugate resolved A-scan according to an embodiment of the subject matter described herein; and Figure 9 is a pair of in vivo full depth images of the ocular anterior segment according to an embodiment of the subject matter described herein. DETAILED DESCRIPTION

0025The present subject matter relates to a method for removing complex conjugate artifacts in FDOCT images by projecting phase shifted signals onto an orthogonal basis set using Fourier decomposition. In one embodiment quadrature projection processing is utilized. Quadrature projection involves the representation of phase shifted signals as vectors. The vectors are typically derived from the real and imaginary components of a phase shifted signal that has been subjected to a Fourier transform. These vectors are then projected onto the real and imaginary axes. One important aspect pertaining to quadrature projection processing is that it is insensitive to the mis-calibrated phase shifts in 90 degree-shift interferometry and only requires predetermination of the quadrant location for each phase shift in non-90 degree phase shift techniques.

0026Figure 1 depicts an exemplary OCT system that may be used to implement the present subject matter. Although Figure 1 describes the present subject matter in the context of an exemplary 3x3 swept source OCT (SS OCT) system used for high speed, real-time imaging an ocular anterior segment, the subject matter described herein may be applied generally to other OCT applications or may be implemented in any suitable interferometer device or system.

0027Referring to Figure 1 , OCT system 100 includes a source 102, a semiconductor optical amplifier (SOA) 104, a plurality of detectors IO6₁...₃, a fiber coupler 108, a reflectance mirror 110, a sample 112, a 4^th detector 114, a first optical circulator 120, a second optical circulator 122, and a recalibration interferometer 116. OCT system 100 may also include a processing unit 132 and a display unit 136.

0028Source 102 may include any broadband short-coherence length light source (e.g., a source having multiple wavelengths or modes) or any comparable light source mechanism. In one embodiment, source 102 may be a tunable Fabry-Perot fiber ring swept laser. Alternatively, source 102 may also comprise a narrowband source if desired. Source 102 is followed by SOA 104.

0029In one embodiment, SOA 104 may include a booster semiconductor optical amplifier. After being amplified by SOA 104, the source signal proceeds to fiber coupler 108 via optical circulator 120. Fiber coupler 108 may include a 3x3 fiber coupler or equivalent mechanism. In one embodiment, fiber coupler 108 is a fiber optic based splitter that is designed to split the source signal between three optical paths. Optical circulator 120 is a component that used to route light more efficiently from source 102 to a sample arm 128 and a reference arm 126. Optical circulator 120 aids with the routing of backscattered light from the sample back to the detector 106i. In an exemplary embodiment, the three optical paths from fiber coupler 108 lead to the sample arm 128, reference arm 126, and recalibration interferometer 116 via the second optical circulator 122. Similarly, fiber coupler 108 may be configured to induce a given phase shift to interferogram signals depending on the coupler's splitting ratio.

0030In the sample arm 128, the source signal strikes sample 112 and a small amount of light is reflected back into fiber coupler 108. In the reference arm 126, the source signal strikes a moving reference mirror 110 and a portion of the source signal light is reflected back to fiber coupler 108. The light beams reflected back from both the sample arm 128 and the reference arm 126 are then recombined in fiber coupler 108 (in either a constructive or destructive manner) depending on the path lengths of reference mirror 110 and sample 112. The interference pattern of light reflected or backscattered from sample 112 and light from reference mirror 110 contains information about the location and scattering amplitude of the scatters in sample 112.

0031After recombining the reflected light, fiber couple 108 splits the light signal and provides the divided light (i.e., detector signals) to detectors 1061...3. Detectors IO61...₃ may include spectrometers, array detectors, or any like components that processes detector signals in order to derive a complex interferometric signal. Specifically, as reference mirror 110 modifies the path length in the reference arm, the detectors observe a series of constructive and destructive combinations of reflected light. Notably, the series of light combinations may be^' used to generate an interferogram. The detector signals from detectors IO6₁...3 are ultimately received by processing unit 132. In one embodiment, the detector signals may be filtered prior to arriving at processing unit 132. Processing unit 132 may include any processing mechanism, such as a desktop computer, a laptop computer, a personal digital assistant (PDA), and the like. Notably, processing unit 132 is responsible for performing the computations for deriving a complex interferogram signal from the filtered detector signals. Processing unit 132 then produces a signal output which is then provided to display unit 136. Display unit 136 may include any display screen, such as a computer monitor, laptop screen, television, and the like. Specifically, display unit 136 is capable of displaying a complex conjugate resolved image that is constructed from the complex inteferogram signal. System 100 may be readily configured to display complex conjugate resolved FDOCT images. Notably, processing unit 132 may be adapted to perform a number of computation intensive tasks that are required to perform the method 200 below. In one embodiment, the processing unit 132 may execute a software program or any other computer readable medium to accomplish this goal.

0032Figure 2 is a flow chart that depicts an exemplary method 200 for performing real-time, quadrature projection based Fourier domain (FD) optical coherence tomography (OCT) according to an embodiment of the subject matter described herein. Referring to Figure 2, in block 202, each of a plurality of interferogram signals is separated by an induced phase shift. In one embodiment, an interferogram signal is derived from interferometric components of a detector signal in a phase-shifting FDOCT system. The interferogram component of the detector signal may be represented as ^s_« (^) ∞ ∑ A, cos(2Λz_mk + Θ_m +φ_n) , where Az_m, A_n, and θ_m respectively represent the axial distance, the reflectance, and relative phase of the m^th scatterer (e.g., a light deflecting particle or pit) in the sample to be analyzed. Similarly, φ_n represents an additional phase shift introduced for the n^th phase shifted acquisition.

0033Consider the scenario where three separate interferogram signals (or interferograms), Si(k), S₂(k), and S3(k), are separated by a nominal phase shift of 120 degrees. More specifically, each interferogram signal is phase shifted by 120 degrees in relation to the previous signal. For example, s∑ik) is shifted 120 degrees in relation to Si(k) and S₃(k) is phase shifted 120 degrees from s₂(k). Figure 3 is an exemplary illustration of the phase-shifted interferogram signals, Si(k), S₂(Zc), and s₃(k). In one embodiment, interferogram signals are provided as output from detectors 106. In block 204, the Fourier transform of each interferogram signal is taken.

0034Specifically, each of the interferogram signals Si(k), s₂(k), and s₃(k) is subjected to a Fourier transformation operation. In one embodiment, the

0035Fourier transform of each phase shifted interferogram signal may be a complex array whose values are the coefficients of a least squares fit of the interferogram signal by a summation of sine and cosine functions. For discretely sampled systems, a discrete Fourier Transform (DFT) or Fast Fourier Transform (FFT) (when the number of sample is a power of two) may be used. Because the interferogram signals differ only by an external phase shift (i.e., corresponding to less than a wavelength), the Fourier transform of each signal, S_n(z) = FT{Sπ(k)}, has the same intensity depth profile. However, the phase at each corresponding depth differs by the induced phase shift, φ_n as shown in Figure 4. Figure 4 depicts Fourier transformed phase-shifted interferogram signals in the complex plane. The Fourier transformed interferogram signals, S_n(Z), may be represented vectorally in the complex plane by plotting the depth resolved magnitude ||s_π(z_m))| at an angle θ_m+φ_π, where the possible values of θ_m range from 0 to 2π. Although additional reflectors may be utilized to generate interferograms in accordance to the present subject matter, Figure 4 only illustrates results for a single reflector for clarity purposes.

0036Returning to Figure 2, in block 206, the depth dependence of the phase θ_m is removed in order to align the phase at all depths with the induced phase shift φ_n. Knowledge of θ_m, which fills a 2τr distribution, is required to correctly sum the quadrature components of the phase shifted interferogram signals. To simplify accounting for the phase of each reflector, the phase profile of one detector signal may be artificially aligned to zero for one phase shifted signal by subtracting its phase profile from each of the phase shifted signals, thus maintaining the phase relationship induced by the fiber coupler. In one embodiment, this may be accomplished by subtracting the phase profile of the first transformed interferogram signal, which is arbitrarily selected to be aligned with the positive real axis, from each of the transformed phase shifted signals. Specifically, the expression s_nχz)=\\§_n(z)\\exp[arg(s_n(z))-arg(s₁(z))] may be used. As an example, the aligning of the second transformed signal (s₂'{z)) may be determined by subtracting the phase profile of the second signal from the phase profile of the first signal (i.e., s₂ (z) = ||s₂(z)||exp[arg(s₂(z))-arg(si(z))]). The result of this alignment (i.e., phase subtraction) process for a single reflector is illustrated vectorally in Figure 5, which shows that transformed signal S₁Xz) is aligned with the positive real axis while the remaining phase shifted signals are forced onto the angle induced by the phase shift, φ_n. Notably, the phase relationship between the shifted interferogram signals is maintained. This phase subtraction results in a negligible axial shift of each scatterer by up to half a wavelength (e.g., up to x), which is much smaller than the coherence length for practical light sources used in OCT. Note that the phase is referenced relative to the first transformed detector signal, and does not require knowledge

0037Returning to Figure 2, in block 208, the quadrature components of each phase referenced signal s_n'(z) are determined. As illustrated by the dashed arrows in Figure 5, this step is analogous to calculating the projection of each vector (associated with each transformed phased shifted signal) onto the real and imaginary axes. In one embodiment, this may be performed via Fourier decomposition by simply expressing the complex valued transformed signals s_π \z) (e.g., sY(z), S₂ '(z), and s₃ (z)) in rectangular coordinates and determining the respective real and imaginary parts, s_n^RE(z) = Re{s_π'(z)} and s_n'^M(z) - lm{s_π Xz)) for each interferogram.

0038Referring to Figure 5, each of the transformed signals έi(z), S₂(Z), and s₃(z) are broken down into rectangular coordinates, or "quadrature components." For example, έ₂(z) is broken down to quadrature components s₂^Rε(z) and s₂^IM(z) and s₃(z) is broken down to s₃^RE(z) and s₃^IM(z). Although not explicitly depicted in Figure 5, έi(z) is broken down to quadrature components έ-ι^RE(z) and si ^M(z) as well, but component si'^M(z) has no value because S₁(Z) lies completely on the real axis (and thus, is entirely equivalent to component Si^RE(z)). Notably, the derived s_n^RE(z) and s_n^IM(z) are in quadrature irrespective of the induced phase shift φ_n because the cosine and sine basis functions are orthogonal. More specifically, this operation corresponds to Fourier decomposition of the phase-referenced signals, with the cosine and sine basis functions aligned with the positive real and imaginary axes, respectively. These decomposed signals are thus separated in phase by precisely 90 degrees, and are thus said to be in quadrature. The choice of Fourier decomposition with cosine and sine basis functions is convenient but not critical to the processing quadrature projection algorithm which is compatible with any set of orthogonal basis functions. Also, dependence of φ_n on wavenumber (i.e., inverse of wavelength) does not affect the orthogonality of the projections. Note that both s_π^RE(z) and s_π'^M(z) are real valued, and that the depth profile of s_n^RE(z) is symmetric about the reference position, whereas s_n^IM(z) is anti-symmetric.

0039Returning to Figure 2, in block 210, the quadrature components of the transformed phase shifted signals are combined. When utilized in conjunction with 90 degree-phase shifting techniques, only the projection along the real or imaginary axis is retained. Notably, a non-zero value of the perpendicular component represents a combination of a mis-calibrated phase shift and/or sample drift. When quadrature projection is used with non-90 degree phase shifting techniques, the real and imaginary components are of similar amplitude and both are retained. An estimate of each φ_n is required to predetermine which quadrature projected components are aligned parallel or anti-parallel to the axes, represented using δ_n^Rε = ±1 and δ_n^IM = ±1 ■ Since the sign of each δ_n is dependent only on the quadrant location of φ_n, the quadrature projection algorithm is insensitive to mis-calibration or drift of the actual induced phase shift. The derived real and imaginary signals are thus determined using the relations s*^ε(z)= ∑δϊ^εs«^ε , and s'^M{z) = ∑δ™sf' , as illustrated in Figure 6.

0040Returning to Figure 2, in optional block 212, the derived quadrature components s^RE(z) and s^IM(z) are scaled to account for unequal contributions to the real and imaginary quadrature components. Specifically, the derived real and imaginary signals may require scaling to normalize their respective amplitudes. A variety of techniques may be used to determine the optimal scaling factor. For example, a normalizing scalar may be obtained by taking the ration of the summed magnitude of the real and imaginary vectors. Alternatively, an iterative process, either manual or computational, may be used to determine the optimal scaling factor β for suppression of any possible complex conjugate artifact from a strong reflector. In one embodiment, the derived quadrature components may be scaled relative to each other by calculating β as the ratio of the maximum value of s^RE to s'^M. In another embodiment, the scaling coefficient β may represent the ratio of the total energy in the derived real signal to that of the derived imaginary signal. Specifically, β may be represented as β = -J j|i ^RE (zf dzj /|i^/M (z| ^" dz .

0041Returning to Figure 2, in block 214, the complex conjugate resolved image is obtained. In one embodiment, the complex conjugate resolved image (e.g., an A-scan ultrasound biometry) is obtained by directly adding the derived real component and the scaled imaginary component, i.e., s(z) - s^RE(z)+β s'^M(z). This allows for unambiguous discrimination of positive and negative distances. The summation is illustrated in Figure 7, which shows the cancellation of the symmetric complex conjugate artifact peaks in s^Rε(z) by the anti-symmetric peaks in s^IM{z), thereby resulting in the full range depth profile s^RE{z) (which may ultimately be used to construct a complex conjugate resolved image).

0042Quadrature projection processing as described herein was demonstrated using three non-90° phase shifted interferograms acquired simultaneously from the ports of the 3x3 Michelson type interferometer as illustrated in Figure 1. The source was a tunable Fabry-Perot fiber ring swept laser (e.g., Micron Optics) followed by a booster semiconductor optical amplifier (e.g., InPhenix), providing a source bandwidth of 84 nm FWHM centered at 1310 nm, and an average power in excess of 8mW at the sample. The source was driven with a 3.33 kHz triangular wave, providing an effective 6.67 kHz line rate by processing both forward and backward sweeps. The sample arm was mounted on a customized slit lamp with galvanometer mounted scanning mirrors. A calibration signal from a 2x2 fiber Michelson interferometer was used to re-sample the data channels using a "nearest-neighbor" algorithm. All four channels were digitized simultaneously at 10MHz (e.g., National Instruments PCI 6115), from four photodiode detectors (e.g., New Focus; D_1-3, model 1817, and D₄, model 1617).

0043The optical power at the sample was reduced to 3.75mW for ocular anterior segment imaging, and this value was used to measure the system sensitivity. The complex conjugate suppression and system sensitivity measurements are presented in Figure 8 for a -5OdB attenuator in the sample arm. The complex conjugate resolved A-scans achieve >30 dB suppression of the symmetric artifact. The double-sided image depth was 6.6mm, as indicated in Figure 8. The peak system sensitivity, accounting for re-coupling losses, was measured to be 103 dB near DC and decreased by 8 dB at the ends of the depth scan.

0044Figure 9 depicts in vivo full depth images of the ocular anterior segment of human volunteers. The images were obtained using a combination of a 3x3 SS OCT system with quadrature projection processing. Corruption of the complex conjugate resolved image due to sample motion was not observed, since the phase-shifted interferograms from the 3x3 interferometer were acquired simultaneously. In one embodiment, the image processing may be performed by a standard desktop computer (e.g., Intel Pentium D 3.2 GHz) executing C++ programs or any other suitable program language. The nearest neighbor resampling and quadrature projection algorithm may be performed and displayed in real time on 1 ,024 point A-scans, for 800 lines per frame at 6.7 frames per second. In Figure 9, a generated complex conjugate resolved image 901 is compared against an unresolved image 902 (obtained by averaging the Fourier transformed detector outputs). Notably, image 901 demonstrates a full depth, artifact free image of a human eye anterior segment acquired in vivo. Conversely, image 902 depicts the human eye anterior segment that is corrupted by the complex conjugate artifact.

0045In conclusion, a method for complex conjugate resolved FDOCT with arbitrarily spaced phase shifts is presented. Although the method was described for -120 degree phase shifted interferometry using a 3x3 fiber coupler, the present subject matter is generally applicable for correction of mis- calibrated phase shifts obtained by other phase shifting techniques. It will be understood that various details of the subject matter described herein may be changed without departing from the scope of the subject matter described herein.. Furthermore, the foregoing description is for the purpose of illustration only, and not for the purpose of limitation.