Open Access
Add to BibSonomy
Abstract
Digitally balanced detection (DBD) has been studied in Fourier domain optical coherence tomography systems to improve sensitivity. We report a technical advancement in a multiple reference optical coherence tomography (MR-OCT) system with the implementation of the DBD technique. We used a free-space, polarization-based balanced detection configuration, which is scalable and suitable for miniaturization. The efficiency of common-mode noise rejection is limited in free-space systems due to the non-uniform illumination of the sensors. We show that recording the signals separately and balancing them in the digital domain improves common mode rejection and signal quality. The application of the DBD scheme for MR-OCT achieves an average sensitivity improvement of 5 ± 0.5 dB over its analogue balanced detection counterpart. We also show that DBD improves the contrast on images of Scotch tapes and mouse eyes.
Published by Optica Publishing Group under the terms of the Creative Commons Attribution 4.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article's title, journal citation, and DOI.
1. Introduction
Optical coherence tomography (OCT) is an imaging technique that produces high-resolution cross-sectional images of biological tissues and materials. OCT is based on low-coherence interferometry and can perform live, non-invasive imaging. Time domain OCT was first demonstrated in 1992. Later, spectral domain OCT (SD-OCT) systems and swept-source-based OCT (SS-OCT) systems were introduced. The SD-OCT and SS-OCT systems are collectively called the Fourier domain OCT systems (FD-OCT) [1,2]. The commercial availability of miniaturized optics and the need for flexibility in many clinical applications have driven OCT systems to reduce the form factor and cost while maintaining sufficient imaging performance [3,4]. Since Fourier domain OCT (FD-OCT) systems offer advantages in respect of speed, sensitivity and stability over TD-OCT, most of the recent miniaturization efforts have concentrated on Fourier domain implementations. Also, due to the requirement of a mechanically scanning reference mirror, achieving sufficient imaging depth while keeping the optics mechanically stable is challenging in TD-OCT [4–7]. Multiple reference optical coherence tomography (MR-OCT) is an amendment on TD-OCT with a partial mirror in front of the scanning reference mirror. The re-circulation of light between the partial mirror and reference mirror leads to an enhanced imaging depth even with a relatively low scan range [8]. The MR-OCT design unlocks opportunities to realize low-cost and miniature OCT for defined applications using off-the-shelf components and conventional production methods. The transmission of reference beam through the partial mirror limits the achievable sensitivity for MR-OCT. Because of this, it is necessary to suppress as much noise as possible to achieve sufficient sensitivity for highly scattering samples such as skin [9,10]. MR-OCT technology has already demonstrated applications in dentistry, dermatology and non-destructive testing [11–14]. The MR-OCT targets a range of applications in which high-end imaging is not required, cost constraints demand the use of less expensive components, and reduced speed and sensitivity are acceptable. MR-OCT has the potential to fill the gap between larger high-end systems and wafer-level optically integrated systems. To accurately judge the future commercial viability, many more factors would need to be considered that relate to costs of production, availability of components, and required sensitivity to detect scattered light.
It is well known from the literature that the signal-to-noise ratio (SNR) of OCT systems can be improved by implementing a balanced detection scheme [15–18]. It can reduce any common-mode noise that originates from the reference and sample arms. Specifically, the noise from the light source can be significantly reduced using balanced detection. Additionally, in comparison to the unbalanced configuration, the balanced detection configuration can yield a current that is double that of the single detection [15,19].
Even though balanced SD-OCT is now possible with more than 100 times noise suppression [20], the use of balanced detection is more common among TD-OCT systems and swept source-based OCT system (SS-OCT) compared to SD-OCT systems. This is because the SD-OCT system demands a more complex configuration since it requires balancing an array of photo-detectors instead of a single detector as in TD-OCT or SS-OCT systems. The merits of fibre-based OCT systems along with balanced detection are well explored [21]. However, fibre-based systems are not suitable for miniaturization since the bending radius of fibre guides will limit miniaturization attempts. There are integrated optical coherence tomography systems on a chip proposed in the literature. However, the packaging challenges make their mass production difficult. That means the wafer-level optical integration will be challenging to access in the near future. Also, optimising it for specific applications is not possible once a wafer optical design is finalised. On the other hand, MR-OCT design is highly flexible, does not need expensive development steps, and can quickly be produced in smaller numbers [22–24]. In this study, we implement a polarization-based configuration to generate two optical channels with a phase shift of $\pi$ radians required for balanced detection [25]. The free-space configuration discussed is suitable for low-cost construction and miniaturization.
The conventional balanced detectors use a differential amplifier that electronically subtracts the input signals to generate the output signal. Perfect optimization of common-mode rejection requires identical photodetectors and an optimum match between the optical beams in terms of path length, intensity, and the illumination of the detector area. To achieve maximum common-mode rejection, photodiodes made on the same semiconductor chip or auto-balanced photodetectors, which compensates up to some extent the optical mismatch using an electronic circuit, may be used. However, these approaches cannot effectively compensate for optical mismatches and spectral differences between the input channels. Compared to free space systems, fibre-based setups can achieve high common-mode rejection. In a free-space optical system, even slight differences in the illumination of sensors on the detectors between the two channels will cause significant differences in the corresponding photocurrents generated by the two photo detectors. In practical settings, it is almost impossible to achieve optimum illumination between the two sensors of the balanced detector. To overcome this, the signals from the two channels can be separately acquired using two different digitizer channels and then balanced in the digital domain after scaling them to the same magnitude [16,26]. In this study, we evaluate the performances of unbalanced detection scheme (UBD), conventional analogue balanced detection (ABD) and digitally balanced detection (DBD) in the context of an MR-OCT system with an open-space, polarization-based two-channel optical configuration. This study mainly contributes to understanding how DBD is applied for MR-OCT. We demonstrate that the digitally balanced detection configuration (DBD) yields sensitivity advantages of $5 \pm 0.5$ dB over an analogue balanced detection scheme and performs more efficient removal of fixed pattern noises.
2. Theory and methods
2.1 Multiple reference optical coherence tomography
The basic theory of MR-OCT comes from the theory of TD-OCT [27–29]. The time-dependent power $P$ of the interference for TD-OCT is
Fig. 1. (a) Schematic of MR-OCT system. The red ellipses represent the scan ranges corresponding to each order of reflection. SLD: superluminescent diode, L: lens, BS: beam splitter, TM: turning mirror, PM: partial mirror, SRM: scanning reference mirror, VC: voice coil actuator, PD: photodetector. (b) Shows enhanced optical path delays generated by multiple reflections. The angle $\alpha$ is close to zero. $\delta l$ is the axial scanning range of SRM. The incident wavefront from the beamsplitter with a power $P_{\text {r}_0}$ is reflected on the PM with reflectivity $R_{\text {PM}}$ generating an optical DC with power $P_{\text {r}_0}\cdot {R_\text {PM}}$. The beam transmitted through the PM undergoes multiple reflections to give rise to higher-order reference beams with optical powers $P_\text {r (N = 1, 2,.., N)}$. (c) The scan range increases with higher orders of reflections (N). This causes the scan ranges of higher orders to for overlapping regions $(ol)$. D is the spacing between SRM and PM.
Depending on the reflectivity $R$ of the partial mirror, a portion of the reference intensity is reflected towards the beam splitter. Along with an unwanted background DC on the detector, This causes a reduction in intensity for higher orders. The reference power associated with the $N^\text {th}$ order reflection is given by
where $P_{{r}_0}$ is the intensity of the incident reference beam emerging from the beam splitter of the Michelson interferometer and directed towards the reference mirror and $N = 1, 2, 3,..$. In this study, we used a partial mirror that transmits 20% and reflects 80% of the reference intensity. As a consequence of multiple orders of reflections, Eq. (1) can be re-written asA superluminescent light emitting diode (SLED) source from Denselight Semiconductors Pte. Ltd. (DL-CS3207A) with a center wavelength of 1310 nm was used for this study. The maximum scanning mirror velocity $v_\mathrm {SRM}$ at the turnaround points was calculated to be 0.071 m/s for a scan range of 75 µm and a scan frequency of 300 Hz. The SNR of an OCT system is defined as ${i_s}^2/{i_n}^2$ where $i_s$ is the peak amplitude of the interference signal and $i_n$ is the overall noise in the detected signal [32]. The peak amplitude for MR-OCT interference signal on a single detector is given by, $i_s = 2\cdot \rho \sqrt {P_\mathrm {r} \cdot P_\mathrm {s}}$ where $\rho$ represents the responsivity of the detector and $P_\mathrm {s}$, $P_\mathrm {r}$ are the sample and reference arm powers, respectively. The responsivity is $\rho = \eta q_e/h\nu _0$ where $\eta$ is the quantum efficiency of the detector, $q_e$ is the electron charge, $h$ is the Plank’s constant and $\nu _0$ is the center frequency of the light source [21,29,33–35]. The shot-noise limited signal for TD-OCT is $\mathrm {SNR}_\mathrm {TD-OCT}=\frac {\rho \lambda }{ h c} P_\mathrm {s} \frac {1}{B}$ [36] where $B$ is the electronic detection bandwidth related to the bandpass filter for order separation. Using the shot noise from TD-OCT underestimates the actual shot noise for MR-OCT due to the sample power $P_s$ distributed over all orders. We must consider that the coherence regions for MR-OCT share the total power in the interferometer. As it is not trivial to expand $P_s$ for all orders, we must use the shot-noise as described for TD-OCT as an approximation. Although, in theory, an infinite number of reference reflections can be calculated, it is plausible that for the generation of a signal, the reference power per order of reflection must be at least above the noise power of the shot noise. Current measurements on the MR-OCT suggest that, on average, 12 orders generate an interference signal suitable for imaging. Based on the light source centre frequency of 1300 nm, the bandwidth of 40 nm, and a total sample power for all orders $P_s = 0.34$ µW, we estimated the shot-noise limited SNR for a single order, assuming that the power fraction per order is linearly split between interference signals, with a value of about 67 dB. This value correlates to the obtained data in Fig. 5. However, even this may be overestimating the actual value because the power split is not linear, and noise components are present in all signal orders. The overall noise current $i_n$ arises from the contributions of shot noise ($i_{\mathrm {sh}}$), excess noise ($i_{\mathrm {ex}}$), thermal noise ($i_{\mathrm {th}}$) and data acquisition (DAQ) ($i_{\mathrm {DAQ}}$) noise. Therefore, $i_n^2$ can be expressed as
2.2 MR-OCT with digitally balanced detection
An MR-OCT system with a free-space, two-channel polarization-based balanced detection configuration was used in the study (Fig. 2). In the Michelson interferometer, the sample and reference beams are encoded with mutually orthogonal polarization states. These beams are spatially combined at the polarization beam splitter (PBS). The intensity on each detector in a polarization-based balanced detection configuration is the same as the intensity available on the sensor in a single detector configuration that uses a non-polarizing beam splitter. Also, the interference-related oscillations on both detectors will have a mutual phase difference, $\pi$, in radians (Fig. 3). However, the noise components including the reflections from partial mirror, and auto-correlation due to self-interference within the sample will be in phase. Consequently, the summing operation at the balanced detector will reject this common-mode noise and thereby improve the overall SNR of the signal. In a balanced detection scheme, the signal $i_s$ becomes $4\cdot \rho \sqrt {P_\mathrm {r} \cdot P_\mathrm {s}}$ (instead of $2\cdot \rho \sqrt {P_\mathrm {r} \cdot P_\mathrm {s}}$ in an unbalanced configuration). Also, the common mode noises will be rejected from the overall noise current. Therefore, theoretically, in a balanced detection configuration, the SNR ($i_s^2/i_n^2$) improves compared to an unbalanced detection scheme.
Fig. 2. Schematic of MR-OCT with different detection systems. a) analogue balanced detection, b) digitally balanced detection. The rectangles in red color highlight the differences between the detection schemes. SLED: superluminescent diode, OF: optical fiber, CM: collimator, POL: polarizer, PM: partial mirror, RSM: reference scanning mirror, TM: turning mirror, RAA: reference arm attenuator, HWP: half-wave plate, PBS: polarizing beam splitter, CP: compensation plate, QWP: quarter-wave plate, GM: galvo mirror, SL: sample lens, RL: reference lens, DL: detector lens, D: detector
Fig. 3. MR-OCT first order interferograms were recorded by the A and B channels. The signals have almost same amplitude and a mutual phase difference of $\pi$ rad. The green interferogram was obtained by subtracting signal B from A in the digital domain. The zoomed in portion in the inset helps to clearly visualize the phase difference between the signals.
We used a Newport New Focus 2117 photoreceiver to realise conventional analogue balanced detection and two Newport New Focus 2053 photoreceivers for digitally balanced detection. All the photoreceivers were used in their free space versions. Both models have the same diode type (InGaAs/PIN) and active area ($0.08\ \mathrm {mm^2}$). Both detectors come with the same built-in amplifier. The noise equivalent power is $0.4\ \mathrm {pW/\sqrt {Hz}}$ with a responsivity of $0.98\ \mathrm {A/W}$ at a wavelength of 1300 nm. Based on the relation $\mathrm {NEC} = \rho \cdot \mathrm {NEP}[\mathrm {A/\sqrt {Hz}}]$, where NEP is the noise equivalent power, the expected noise equivalent current (NEC) value is $0.4\ \mathrm {pA/\sqrt {Hz}}$. The bandwidth is based on the frequency range required for acquisition which was $10\ \mathrm {MHz}$. For the data acquisition, a 14 Bit, 4 Channel, 125 MS/s waveform digitizer from AlazarTech (ATS9440) was used. Since the detectors used have similar specifications, we assume that the thermal noise and shot noise in Eq. (7) remain comparable between all the detectors used. However, while using digitally balanced detection, the DAQ noise will be double that of unbalanced or analogue balanced detection schemes as two separate digitizer channels are used for acquisition. The advantages of digitally balanced detection mostly comes from the effective suppression of excess noise. In this manuscript we compare between DBD and ABD based on sensitivity and imaging performances.
3. Results and discussions
3.1 MR-OCT signal processing and digitally balanced detection
The processing steps for MR-OCT are depticted as a flow chart in Fig. 4. The raw signal contains signals from multiple depths present as different orders ($N$). The MR-OCT uses a voice coil for axial scanning. Since the voice coil is driven using a sinusoidal signal, the axial motion of the reference mirror is not linear in time. Since the data are sampled at regular intervals of time, the spatial distribution of the data does not precisely correspond to the spatial distribution of the backscattered profile of the sample. Consequently, the resulting image will be distorted or 'warped'. After the DC removal, a phase linearization is performed to remove these distortions. The process involves remapping the sample points of the signal according to a look-up table that relates to the actual movement of the scanning mirror. More details on phase linearization are available in the literature [31,38]. The frequency of interference signal for each order of reflection is different according to Eq. (6). Each order of reflection is then separated out based on its beat frequency by using a suitable band-pass filter. After band-pass filtering, a Hilbert transform is performed to detect the envelope of the signal. The signal envelope for each order represents its respective intensity values. In MR-OCT, the increase in scan range for higher orders of reflections appears on the oscilloscope as squeezed signals. As a consequence, for the $N^{th}$ order, the Gaussian full width half maximum (FWHM) of the envelope would reduce to FWHM/$N$. To compensate this, the $N^{th}$ order is upsampled by a factor of $N$ to achieve the proper pixel spacing relative to the first order. Furthermore, to calculate the physical spatial position of each order, the zero position of each virtual scanning mirror is displaced by the spacing between SRM and PM ($D$) and a half of the scanning range ($\delta z$).
Fig. 4. MR-OCT signal processing steps. $N$ represents the total number of orders processed. $O_N$ means $N^{th}$ order after band-pass filtering. $H$ refers to the envelope of the signal after Hilbert transformation. $D$ is the spacing between PM and SRM. $\delta z$ is scan range of SRM. The operation of ’displace by’ means to displace the zero position of each virtual scanning mirror by the formula specified. The ’+’ sign denotes that all separate signal arrays are merged into a single array for final image reconstruction.
As balanced detection is based on subtraction of two signals, it is important to ensure that the signals have the same amplitude and a phase difference of $\mathrm {\pi }$ radians as much as possible. One of the advantages of digitally balanced detection over the analogue balanced detection scheme is that it allows us to manipulate the individual signals separately and apply balancing by subtraction at a later stage. To evaluate the performance of digitally balanced detection, we performed certain processing steps from the flow-chart (Fig. 4) on the individual signals separately, applied balancing by subtracting the signals, and then continued the remaining processing steps for reconstructing the final image. The motivation behind this exercise was to evaluate if the identicalness of the individual signals improved at any stage of the MR-OCT signal processing. The sensitivity results from the above exercise are listed in Table 1. For example, the column ’After linearization’ in Table 1 lists the sensitivity values for selected orders when the individual signals were processed separately until ’phase linearization’ step of Fig. 4, before balancing them. The rest of the processing steps (according to Fig. 4) were carried out on the balanced signal. The beat oscillations are not preserved after the Hilbert transform. Therefore, balancing the signals is not possible after this stage. A solution to minimize the spectral variation between the two signals is reported by Chen et al. [26]. A $5^{th}$ order polynomial fit of the channel ratio $R(n) = S_A(n)/S_B(n)$ is used as a compensation function where S denotes the signal, A and B are the two channels, and $n$ is the sample number. The balancing is then performed in the digital domain after scaling the two signals to the same magnitude using Eq. (10).
The sensitivity values after performing this compensation function (CF) are given in Table 1 in the column titled "CF method". We could not achieve any sensitivity improvements with the CF method compared to balancing the signal after DC subtraction. It could be an indication that the splitting ratio on the polarizing beam splitter is not spectrally dependent. The sensitivity results (Table 1) show that the balancing of signals was most effective when performed right after DC subtraction. Therefore, we use balancing after DC subtraction to compare the performance among digitally balanced, analogue balanced and unbalanced detection schemes.
Table 1. We processed the individual signals separately according to the flow chart in Fig. 4. After the processing steps mentioned in the respective columns, the two signals were balanced by subtraction. The rest of the processing steps were continued on the balanced signal. The sensitivity values obtained for selected orders are listed in the table. CF method refers to the balancing using the compensation function.
3.2 Sensitivity characteristics for unbalanced, digitally balanced and analogue balanced detection
3.2.1 Performance comparison based on sensitivity versus reference arm power characteristics
To compare the sensitivity characteristics among different detection schemes for different values of reference attenuation, a mirror was used in the sample arm along with an OD = 2 neutral density filter. A continuously variable ND filter (NDC-50C-2M) from Thorlabs, Inc. was used in the reference arm to adjust the reference arm attenuation. The sample mirror was positioned to maximize the first-order signal, and 100 buffers were recorded from a single point. The mean and standard deviation were calculated from the peaks of PSFs. The median noise was subtracted from the mean to calculate the SNR. 40 dB was added to the SNR to obtain the sensitivity value. The sensitivity values were recorded until saturation was reached. In the cases of digitally balanced and unbalanced detection schemes, the photodetector saturated first, whereas, in the case of analogue balanced detector, digitizer got saturated first. A detector amplifier gain of ${10^4}$ and digitizer voltage range of 2 V was used for the measurements. The results are plotted in Fig. 5. The average sensitivity for the first-order signal obtained using digitally balanced detection (DBD) was $5\pm 0.5$ dB more than that of analogue balanced detection (ABD).
Fig. 5. (a) Sensitivity of MR-OCT signal for different detection schemes versus reference arm attenuation. DBD: digitally balanced detection, ABD: analogue balanced detection, UBD: unbalanced detection. The DBD saturated at OD = 1.1 while ABD saturated at OD = 1.3.
3.2.2 Performance comparison based on sensitivity roll-off evaluation
To evaluate the sensitivity vs depth characteristics for MR-OCT, we collected signals at equally spaced depth positions of the sample mirror (OD =2 attenuated) and constructed a calibration line (Fig. 6(a)). The calibration line based on the series of A-lines provides a more comprehensive representation of noise as seen in a B-frame. Each A-line containing the point spread function (PSF) at a specific depth position allows the evaluation of order-specific PSF. All A-lines are assembled into a B-frame keeping the spacing between A-lines equal to the mirror steps. The resulting B-frame shows a diagonal line with sample mirror depth position on the x-axis and the reconstructed depth position on the y-axis. This calibration line serves as a visual aid to evaluate the intensity roll-off and inspect other imaging artefacts. The SNR for each order was calculated by taking the ratio of the mean of peaks from corresponding point spread functions and the mean noise floor plotted in dB scale. The sensitivity was then calculated by adding 40 dB to account for the attenuation provided by an OD = 2 neutral density filter. Figure 6 shows calibration lines (a, b) for different detection configurations and the corresponding sensitivity roll-off (c).
Fig. 6. a) The calibration lines for DBD, ABD, and UBD show different background noise and noise artefacts. In the plots showing calibration lines, the x-axis is the actual sample mirror depth position, and the y-axis is the imaged sample mirror depth position. The range of the x and y axes is 1500 µm. For UBD, the background noise is larger than the balanced detection signals. Although the background noise appears to be the smallest for the ABD signal, it also has a reduced signal intensity. The ABD shows some fixed pattern noise as two horizontal lines in the upper half, most likely originating from electronic amplified noise. The DBD successfully rejects noise patterns and has low homogeneous background noise. The dashed artefacts to the left of the main diagonal originate from interference between multiple reflections of the partial mirror. b) Shows the calibration lines normalized to the maximum intensity. Here, we can see that the background noise is the lowest for DBD and highest for UBD. c) Shows the sensitivity roll-off for MR-OCT system with analogue balanced detection (ABD), digitally balanced detection (DBD) and unbalanced detection (UBD) from one of the channels. It can be observed that the DBD scheme provides higher sensitivity compared to ABD. The spacing between two adjacent orders is 100 µm.
The ABD calibration line shows some horizontal fixed pattern noise. The intensity of fixed pattern noise is less in the DBD scheme and almost invisible in the calibration line from unbalanced signal. The fixed pattern noise is most likely originating from electronic amplified noise. The calibration lines for each detection scheme were acquired while the reference arm power was just below saturation. At the time of acquisition, the reference arm power was lower for the ABD scheme. Consequently, the overall background noise is lower in the ABD calibration line than in the DBD. Figure 6(b) shows the calibration lines normalized to the maximum intenisty. Figure 6(c) shows that the DBD scheme yields a sensitivity advantage of up to 5 dB with respect to ABD. The error bars represent the standard deviation among the peaks of PSFs representing the corresponding order. The ABD scheme reduces noise and yields higher sensitivity than the unbalanced (UBD) scheme.
A set of A-lines extracted from the calibration lines used for Fig. 6(a) are shown in fig. 7. We selected one A-line each for DBD, ABD and UBD detection schemes. All three A-lines were recorded from the same sample mirror position and correspond to the $4^{\mathrm {th}}$ order in MR-OCT. To compare the noise floors, we aligned the signals to match vertically at the peak and fitted a Gaussian function over the data points. We can see that the DBD scheme has the lowest noise floor. There is a difference of about 5 dB between the noise floors of DBD and ABD.
Fig. 7. Shows the MR-OCT signal ($\mathrm {4^{th}}$ order) from a sample mirror. The A-lines from DBD, ABD and UBD schemes were aligned vertically so that their peaks are at the same level. Then a Gaussian fit (continuous line) was applied to the data points. The flat region of the Gaussian fit represents the mean noise level for each detection scheme.
3.2.3 Performance comparison based on scattering samples
To compare the imaging performances in a scattering sample among different detection schemes, a sample was constructed by slicing off layers from a commercial 3M Scotch Magic Tape and then stacking them up on a coverslip. Scotch tape was chosen because it is readily available, and it is a material with good light scattering properties. To quantitatively compare the visibility of tape layers, we used Michelson contrast. The Michelson contrast describes the visibility of periodic intensity changes based on the minimum ($I_{\mathrm {min}}$) and maximum ($I_{\mathrm {max}}$) amplitude. Michelson contrast was used because the Scotch tape image has a regular pattern of bright and dark areas. We fitted exponential and sinusoidal functions to the axial intensity profiles in Fig. 8(b). The exponential function was used to compensate for the intensity roll-off vs depth. The sinusoidal model followed the intensity undulations and helped to calculate $I_{\mathrm {min}}$ and $I_{\mathrm {max}}$ for each layer. The Michelson contrast was then computed using the equation ($I_{\mathrm {max}} - I_{\mathrm {min}})/I_{\mathrm {max}} + I_{\mathrm {min}}))$ [39]. Table 2 shows the percentage improvement of the Michelson contrast for DBD over ABD, for different depths based on the tape layer steps. We show results from three data sets (Data 1, 2 and 3) acquired from slightly different locations. The results shown in Fig. 8 were computed from Data 1. The Fig. 8 equivalents of Data 2 and Data 3 are provided in the supplemental document (Supplement 1, Figs. S1 and S2). The increase of percentage improvements beyond layer six in Table 2 may be spuriously caused by increasing fitting errors. Nevertheless, the consistency of values suggests that some improvement occurs.
Fig. 8. A Scotch tape was imaged to compare the performance of different detection schemes in a scattering sample. The line artefact at the bottom of the image originates from the coverslip on which the tapes are stacked together. a) Shows the conventional intensity-based OCT images obtained for each detection scheme. The scale bar represents 225 µm. b) shows the averaged A-lines from intensity-based OCT images. c) compares the Michelson contrast for different layers of Scotch tape.
Table 2. The percentage improvement in the Michelson contrast for each layer while DBD was used instead of ABD. Values marked with (*) may need to be ignored due to large fitting errors. Results from three data sets (Data 1, Data 2 and Data 3) acquired from slightly different locations in the tape are shown.
To compare the imaging performance among different detection schemes with a biological specimen, the anterior chambers of mouse eyes were imaged, ex vivo (Fig. 9). To compare the visibility, contrast to noise ratio (CNR) for structures in the white boxes were calculated using $\vert { S_\mathrm {A} - S_\mathrm {B}}\vert /{\sigma }$ where $S_\mathrm {A}$ and $S_\mathrm {B}$ are signal intensities for signal producing structures A (region of cornea inside the inset) and B (median background intensity in the selected region), respectively and $\sigma$ is the standard deviation of the image noise. To calculate the error in the calculation of CNR, the inset was shifted to three different locations, each 30 pixels apart, and the standard deviation in the CNR calculation was computed. The CNR values from three different eyes (Samples 1, 2 and 3) are listed in Table 3. We performed background removal on all the images in Fig. 9 by subtracting the mean of A-lines. The images in Fig. 9 are acquired from sample 1 in Table 3. Equivalent images for samples 2 and 3 are provided in the supplemental document (figure S3). The contrast measurements from scattering samples may vary depending on the dehydration of the tissue, power fluctuations and drift of the light source, and alignment issues. The comparison of imaging performance among the three detection schemes was performed carefully in such a way that the above-mentioned parameters are controlled as much as possible. The light source remained switched on throughout the experiment and we performed the imaging as fast as possible to control the dehydration and voltage drift of the power source. Except for the switching of detectors, no alignment changes were introduced in the interferometric system. However, the sample position may have slightly changed.
Fig. 9. The anterior chamber of a mouse eye was imaged ex vivo using three different detection schemes. The horizontal and vertical scale bars represent 500 µm and 150 µm respectively. CNR values were calculated to quantitatively represent the visibility of structures in the insets. A background removal by subtraction of mean A-line was performed on all the three images before CNR calculation. The MR-OCT system has axial and transverse resolutions in the air of 13 µm and 27 µm, respectively. The structure in the ABD scheme is slightly different since the sample position was slightly shifted while swapping the detection scheme.
Table 3. Comparison of CNR values among different detection schemes using three mouse eye samples.
4. Conclusion
We demonstrated a sensitivity improvement of an MR-OCT system by using a digitally balanced detection scheme. In a polarization-based, free space, two-channel balanced detection configuration, the digitally balanced detection scheme provided an average $5 \pm 0.5$ dB improvement in sensitivity and reduced fixed pattern noise compared to analogue balanced detection. In conclusion, the removal of common-mode noise is more efficient with digital post-processing after the removal of DC, which comprises the digitally balanced detection scheme. We also compared the performance of the DBD scheme against analogue balanced detection in a scattering sample. We used the visibility of layers in a Scotch tape to obtain quantitative estimates of the sensitivity improvement. The image contrast of the cornea of a mouse eye was improved with DBD, demonstrating that MR-OCT can image biomedical ex vivo samples. The reported method uses time-domain reference scanning , and the generation for multiple scanning layers allows to enhance the imaging depth. The distinct availability of multiple scanning layers in depth and frequency can be of interest for interferometric depth measurement systems or specific OCT configurations requiring additional imaging depths for larger objects. The advantages of DBD have been confirmed to be applicable to the multi-band detection method of MR-OCT.
Funding
Horizon 2020 Framework Programme (761214, 779960).
Acknowledgments
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreements no. 761214 and no. 779960. The materials presented and views expressed here are the responsibility of the authors only. The EU Commission takes no responsibility for any use made of the information set out. The authors also acknowledge the University of Galway for facilities and funding. Sincere gratitude to my co-authors for their valuable insight and discussions about our shared research interests. The authors wish to thank the Bio-Resources Unit technical, veterinary and administrative staff at the University of Galway for facilitating ex vivo studies and for their ongoing assistance, advice and support in animal husbandry, care and welfare.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are available on request.
Supplemental document
See Supplement 1 for supporting content.
References
1. D. Huang, E. Swanson, C. Lin, J. Schuman, W. Stinson, W. Chang, M. Hee, T. Flotte, K. Gregory, C. Puliafito, and J. G. Fujimoto, “Optical coherence tomography,” Science 254(5035), 1178–1181 (1991). [CrossRef]
2. 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(6), 1206–1213 (1996). [CrossRef]
3. G. L. Monroy, J. Won, D. R. Spillman, R. Dsouza, and S. A. Boppart, “Clinical translation of handheld optical coherence tomography: Practical considerations and recent advancements,” J. Biomed. Opt. 22(12), 1 (2017). [CrossRef]
4. G. Song, K. K. Chu, S. Kim, M. Crose, B. Cox, E. T. Jelly, J. N. Ulrich, and A. Wax, “First Clinical Application of Low-Cost OCT,” Trans. Vis. Sci. Tech. 8(3), 61 (2019). [CrossRef]
5. A. F. Fercher, C. K. Hitzenberger, G. Kamp, and S. Y. El-Zaiat, “Measurement of intraocular distances by backscattering spectral interferometry,” Opt. Commun. 117(1-2), 43–48 (1995). [CrossRef]
6. B. Golubovic, B. E. Bouma, G. J. Tearney, and J. G. Fujimoto, “Optical frequency-domain reflectometry using rapid wavelength tuning of a Cr4+:forsterite laser,” Opt. Lett. 22(22), 1704 (1997). [CrossRef]
7. M. A. Choma, M. V. Sarunic, C. Yang, and J. A. Izatt, “Sensitivity advantage of swept source and Fourier domain optical coherence tomography,” Opt. Express 11(18), 2183–2189 (2003). [CrossRef]
8. M. Leahy, J. Hogan, C. Wilson, H. Subhash, and R. Dsouza, “Multiple reference optical coherence tomography (MR-OCT) system,” in Dynamics and Fluctuations in Biomedical Photonics X, vol. 8580 (International Society for Optics and Photonics, 2013), p. 85800L.
9. P. M. McNamara, R. Dsouza, C. O’Riordan, S. Collins, P. O’Brien, C. Wilson, J. Hogan, and M. J. Leahy, “Development of a first-generation miniature multiple reference optical coherence tomography imaging device,” J. Biomed. Opt. 21(12), 126020 (2016). [CrossRef]
10. A. Arangath, K. Neuhaus, Y. Zhou, S. Alexandrov, J. Hogan, C. Wilson, and M. Leahy, “Optimization of SNR, sensitivity values by evaluating the relative spacing between the partial mirror and scanning reference mirror for multiple reference optical coherence tomography,” in Optics and Biophotonics in Low-Resource Settings VI, vol. 11230 (International Society for Optics and Photonics, 2020), p. 112300X.
11. R. Dsouza, H. Subhash, K. Neuhaus, J. Hogan, C. Wilson, and M. Leahy, “Dermascope guided multiple reference optical coherence tomography,” Biomed. Opt. Express 5(9), 2870–2882 (2014). [CrossRef]
12. R. Dsouza, H. M. Subhash, K. Neuhaus, J. Hogan, C. Wilson, and M. Leahy, “3D nondestructive testing system with an affordable multiple reference optical-delay-based optical coherence tomography,” Appl. Opt. 54(18), 5634–5638 (2015). [CrossRef]
13. R. Dsouza, H. Subhash, K. Neuhaus, R. Kantamneni, P. M. McNamara, J. Hogan, C. Wilson, and M. Leahy, “Assessment of curing behavior of light-activated dental composites using intensity correlation based multiple reference optical coherence tomography,” Lasers Surg. Med. 48(1), 77–82 (2016). [CrossRef]
14. K. Neuhaus, S. O’Gorman, P. M. McNamara, S. A. Alexandrov, J. Hogan, C. Wilson, and M. J. Leahy, “Simultaneous en-face imaging of multiple layers with multiple reference optical coherence tomography,” J. Biomed. Opt. 22(08), 1 (2017). [CrossRef]
15. A. G. Podoleanu, “Unbalanced versus balanced operation in an optical coherence tomography system,” Appl. Opt. 39(1), 173–182 (2000). [CrossRef]
16. S. Moon and D. Y. Kim, “Normalization detection scheme for high-speed optical frequency-domain imaging and reflectometry,” Opt. Express 15(23), 15129–15146 (2007). [CrossRef]
17. M. Cooper, C. Söller, and B. J. Smith, “High-stability time-domain balanced homodyne detector for ultrafast optical pulse applications,” J. Mod. Opt. 60(8), 611–616 (2013). [CrossRef]
18. A. Bradu and A. G. Podoleanu, “Fourier domain optical coherence tomography system with balance detection,” Opt. Express 20(16), 17522–17538 (2012). [CrossRef]
19. G. Abbas, V. Chan, and T. Yee, “A dual-detector optical heterodyne receiver for local oscillator noise suppression,” J. Lightwave Technol. 3(5), 1110–1122 (1985). [CrossRef]
20. A. M. Kho and V. J. Srinivasan, “Proactive spectrometer matching for excess noise suppression in balanced visible light optical coherence tomography (OCT),” Opt. Express 29(25), 42037–42054 (2021). [CrossRef]
21. A. M. Rollins and J. A. Izatt, “Optimal interferometer designs for optical coherence tomography,” Opt. Lett. 24(21), 1484–1486 (1999). [CrossRef]
22. B. D. Goldberg, S. M. R. M. Nezam, P. Jillella, B. E. Bouma, and G. J. Tearney, “Miniature swept source for point of care Optical Frequency Domain Imaging,” Opt. Express 17(5), 3619–3629 (2009). [CrossRef]
23. S. Schneider, M. Lauermann, P.-I. Dietrich, C. Weimann, W. Freude, and C. Koos, “Optical coherence tomography system mass-producible on a silicon photonic chip,” Opt. Express 24(2), 1573–1586 (2016). [CrossRef]
24. B. I. Akca, B. Považay, A. Alex, K. Wörhoff, R. M. de Ridder, W. Drexler, and M. Pollnau, “Miniature spectrometer and beam splitter for an optical coherence tomography on a silicon chip,” Opt. Express 21(14), 16648–16656 (2013). [CrossRef]
25. A. Arangath, K. Neuhaus, S. Alexandrov, and M. Leahy, “Evaluation of signal degradation due to birefringence in a multiple reference optical coherence tomography system with polarization-based balanced detection,” IEEE Photonics J. 14(4), 1–12 (2022). [CrossRef]
26. Y. Chen, D. M. de Bruin, C. Kerbage, and J. F. de Boer, “Spectrally balanced detection for optical frequency domain imaging,” Opt. Express 15(25), 16390–16399 (2007). [CrossRef]
27. A. F. Fercher, “Optical coherence tomography,” J. Biomed. Opt. 1(2), 157–173 (1996). [CrossRef]
28. M. R. Hee, “Optical coherence tomography of the eye,” Thesis, Massachusetts Institute of Technology (1997).
29. A. F. Fercher, W. Drexler, C. K. Hitzenberger, and T. Lasser, “Optical coherence tomography - principles and applications,” Rep. Prog. Phys. 66(2), 239–303 (2003). [CrossRef]
30. K. Neuhaus, P. M. McNamara, S. Alexandrov, S. O’Gorman, J. Hogan, C. Wilson, and M. J. Leahy, “Performance Review of Multiple Reference Versus Time Domain Optical Coherence Tomography,” IEEE Photonics J. 10(3), 1–14 (2018). [CrossRef]
31. K. Neuhaus, “The multi-spectral signal properties of multiple reference optical coherence tomography,” Thesis, NUI Galway (2019).
32. P. H. Tomlins and R. K. Wang, “Theory, developments and applications of optical coherence tomography,” J. Phys. D: Appl. Phys. 38(15), 2519–2535 (2005). [CrossRef]
33. R. Leitgeb, C. K. Hitzenberger, and A. F. Fercher, “Performance of fourier domain vs. time domain optical coherence tomography,” Opt. Express 11(8), 889–894 (2003). [CrossRef]
34. W. Sorin and D. Baney, “A simple intensity noise reduction technique for optical low-coherence reflectometry,” IEEE Photonics Technol. Lett. 4(12), 1404–1406 (1992). [CrossRef]
35. J. F. de Boer, B. Cense, B. H. Park, M. C. Pierce, G. J. Tearney, and B. E. Bouma, “Improved signal-to-noise ratio in spectral-domain compared with time-domain optical coherence tomography,” Opt. Lett. 28(21), 2067–2069 (2003). [CrossRef]
36. J. F. de Boer, R. Leitgeb, and M. Wojtkowski, “Twenty-five years of optical coherence tomography: The paradigm shift in sensitivity and speed provided by Fourier domain OCT [Invited],” Biomed. Opt. Express 8(7), 3248–3280 (2017). [CrossRef]
37. M. E. Brezinski, Optical Coherence Tomography: Principles and Applications (Academic Press, 2006).
38. R. I. Dsouza, “Towards low cost multiple reference optical coherence tomography for in vivo and NDT applications,” Thesis, National University of Ireland, Galway (2016).
39. A. A. Michelson, Studies in Optics (University Press, 1927).
