>Supercontinuum Laser Applications
>Ultra-high-resolution optical coherence tomography based on supercontinuum light sources
（UHR-OCT）是一种依赖于低相干干涉测量（LCI）原理的非侵入性成像模式。 Ultra-high resolution optical coherence tomography (UHR-OCT) is a non-invasive imaging mode that relies on the principle of low-coherence interferometry (LCI). The main application field of UHR-OCT is biomedical imaging, and there are many applications in the field of non-destructive imaging (NDI). Although speed is one of the main parameters to consider when imaging in vivo samples, optical power, spectral range, and penetration depth are more important for NDI. SC is the only light source that can provide a wide range of free wavelength choices and large bandwidths, while providing high spatial coherence and high optical power. The SC source currently used for UHR-OCT is a mode-locked laser as the pump source.
Q-switched lasers can be used as pump sources instead of mode-locked lasers. Because Q-switched lasers can provide wider pulse width pulses, have sufficient peak power for SC generation, and are low cost.
This article uses a low-cost commercial SC source of Q-switched pump laser (QS-SC) (SuperK Compact, NKT Photonics) for UHR-OCT. The pump laser operates at 22.222 kHz and has a pulse width of 1.6 ns. High-resolution OCT images can be achieved by increasing the exposure time of the camera. We compare this QS-SC with a conventional SC source for OCT, which is based on a mode-locked pump laser (ML-SC) (SuperK Extreme, NKT Photonics) with a repetition frequency of 320 MHz and a pulse width of 10 ps. QS-SC sources currently cost less than 15% of the ML-SC price.
Figure 1 Schematic of UHR-OCT system. DC: directional coupler, PC: polarization controller, C1, C2: parabolic collimator, Disp.C: dispersion compensation block, NDF: neutral density filter, M1: plane mirror, OBJ: scanning lens.
As shown in Figure 1, the UHR-OCT device is a Michelson interferometer with an ultra-wideband 50/50 directional coupler (DC) that divides the optical path into a reference arm and a sample arm. The reference arm consists of a reflection collimator (C2-Thorlabs RC04APC-P01), a dispersion compensation block (Thorlabs LSM02DC), a variable neutral density filter (ND filter), and a plane mirror (M1). The sample arm consists of a reflection collimator (C1-Thorlabs RC04APC-P01), a set of galvanometer-based XY scanners (Thorlabs GVSM002 / M) and a scanning lens (OBJ-Thorlabs LSM02). The spot size is 11 μm at a wavelength of 1315 nm . The spectrometer is Cobra 1300 (Wasatch Photonics), with an optical bandwidth of 1070 nm to 1470 nm, a maximum line speed of 76 kHz, and 2048 pixels. The imaging range of this system is approximately 2 mm. The processing unit consists of a frame grabber (NI PCIe-1433) connected to a workstation (Dell-CPU Intel i7 = 3.33GHz-12 GB RAM). Use the home-designed LabVIEW interface and Matlab algorithm to obtain and process all data. No specific synchronization is applied between the SC source and the spectrometer readings to make the system as simple as possible in software and hardware.
Figure 2 Spectra of two SC sources measured using (a) a commercial OSA and (b) an interferometer with a spectrometer. (C) The normalized PSF is evaluated for each SC source and corresponding Fourier transform (FT) -limited PSF at an axial position (distance) of 150 μm.
Figure 2 (a) is the spectrum of two SC sources measured using an integrating sphere connected to an optical fiber and a spectrometer (OSA) at 1750nm. For QS-SC sources, the spectrum actually extends to 2.4 μm, and for ML-SC sources, the spectrum extends to 2.0 μm. The spectrum measured by the spectrometer after the interferometer of each SC source is shown in Figure 2 (b). Because their spectral shapes are similar, the point spread function (PSF) that characterizes the axial resolution of the UHR-OCT system is also similar. As shown in Fig. 2 (c), the measured axial resolution is slightly lower than 5 μm (in air), which corresponds to the resolution of 3.5 μm tissue (n = 1.4).
In terms of noise, it is mainly detector noise, shot noise and relative intensity noise (RIN). Detector noise includes noise caused by thermo-optic electrons and signal digitization. Scattered noise is noise caused by the random arrival of photons at the detector. RIN is the noise caused by the amplitude fluctuation of the light source. OCT systems should operate in areas where shot noise is dominant, as this will provide the highest signal-to-noise ratio (SNR). However, in the presence of noisy light sources, RIN can dominate, resulting in reduced SNR.
RIN can be measured using filters, photodiodes, and oscilloscopes. The RINs of the two SC sources measured in the wavelength range of 1100nm to 1450nm cover almost the entire spectrometer range of 1070nm to 1470nm. To this end, several 10nm bandpass filters (Thorlabs) are used to filter the light from the SC source with a center wavelength of 50nm and then detected by InGaAs photodiodes (Thorlabs-DET08CFC-800 to 1700nm, 5GHz). The pulse sequence was recorded with a fast oscilloscope (Teledyne LeCroy-HDO9404-10 bit resolution, 40 GS / s and 4 GHz). Figure 3 (a) summarizes the measured RIN as a function of wavelength. The QS-SC source has a low RIN, only between 2.5-5%. In contrast, ML-SC showed strong fluctuations, with RIN in the range of 10% to nearly 45% at longer wavelengths. For the repetition frequency of the light source, when the QS-SC source operates at a rate of kHz, the ML-SC source operates at a rate of MHz. Considering the camera's line rate in the kHz range, the ML-SC can provide thousands of pulses in one exposure time, while the QS-SC provides only a few pulses. For example, an exposure time of 100 μs corresponds to 32,000 pulses of ML-SC and only 2 to 3 pulses of QS-SC. However, since the repetition frequency is very different, because RIN is not Gaussian noise, its improvement is not easy.
Figure 3 SC noise analysis of different camera exposure times: (a) RIN vs. wavelength (b) Noise floor obtained only from the reference arm (the sample arm is blocked). (C) Sensitivity decreases with depth.
The noise in the OCT image is characterized as the background noise of the depth information profile (A-scan) obtained when the sample arm is blocked. To clearly show the trend between noise and exposure time, more than 500 A-scans were averaged. Figure 3 (b) shows the noise floor of the two SC sources and four different exposure times. The two sets of QS-SC and ML-SC curves are separated by about 20dB. In the case of ML-SC, due to the averaging of a large number of pulses, the system is within a shot noise limit of about 100 μs (noise floor from 100 μs to 150 μs does not improve the exposure time). For QS-SC, the longer the exposure time, the lower the noise floor, indicating the presence of RIN in the OCT system. The preliminary observation results were confirmed by the sensitivity chart in Fig. 3 (c), and the two sets of curves showed a sensitivity difference of 20 dB.
Figure 4 1 mm x 2.4 mm B-scan of an IR card, using a 1.3 mW sample power supply, the exposure time of each column is 20 μs, 40 μs, 100 μs and 150 μs: (ad) based on QS -SC and (eh) based on ML- B-scan for SC. (150 μm scale-400 μm horizontal)
Each B-scan in Figure 4 is an area of 1 mm (depth) x 2.4 mm (horizontal) and consists of 500 A scans with a sample power of 1.3 mW. All B-scans are displayed using the same black and white levels to encode dB values into gray levels. When the exposure time is 20 μs, the quality of the B-scan obtained using the QS-SC source is poor, and the SNR is low. Black stripes indicate readings without any light pulses. If the exposure time is increased to 40 μs, the mismatch between the light source pulse and the camera readout still exists, but the image quality is improved. When the exposure time is increased to 100 μs or 150 μs, both light sources can provide good final image quality with similar axial resolution and the same distinguishable structural information.
Figure 5 B-scan of palm skin (1.6mm × 4mm) from a healthy volunteer, using: (ad) ML-SC; (eh) QS-SC. Exposure time of each line: 20, 40, 100, 150 μs (450 μm scale).
Figure 6 4mm x 4mm C-scan from palm skin of healthy volunteers. Based on: (ad) ML-SC; (eh) QS-SC. The exposure time is 20, 40, 100, 150 μs. (Scale bar 1 mm)-(NaN means that the contrast of a 20 μs QS-SC image cannot be calculated due to lack of signal).
Figures 5 and 6 are examples of 4mW skin images collected in vivo from the palms of healthy volunteers with a volume size of 500 (A-scan) x 500 (B-scan) x 1024 (depth). These were obtained in 37.5 seconds, taking into account the longest exposure time (150 μs). But this exposure time is too long to apply imaging to the eyes, heart, and even skin.
This article proves that QS-SC can be used for UHR-OCT at 1300 nm. Even if the repetition frequency of the light source is in the kHz range, by slightly increasing the camera exposure time, an image quality comparable to that of the prior art system can be produced. For imaging of the skin, a tracking or compensation procedure is required if the exposure time is long. In addition to demonstration operations in the 1300 nm range, the QS-SC light source is suitable for operation at shorter wavelengths (800 nm range) or even longer wavelengths (1700 nm, 2000 nm range).
From <M. Maria, IB Gonzalo, T. Feuchter, M. Denninger, PM Moselund, L. Leick, O. Bang, and A. Podoleanu, "Q-switch-pumped supercontinuum for ultra-high resolution optical coherence tomography", Opt. Lett. 42, 4744 (2017).>