|Home | About | Journals | Submit | Contact Us | Français|
We demonstrate ambient light coherent anti-Stokes Raman scattering (AL-CARS) microscopy that allows CARS imaging to be operated under environment light for field use. CARS signal is modulated at megahertz frequency and detected by a photodiode equipped with a lab-built resonant amplifier, then extracted through a lock-in amplifier. Filters in both spectral domain and frequency domain effectively blocked room light contamination of the CARS image. In situ hyperspectral CARS imaging of tumor tissue under ambient light is demonstrated.
Coherent anti-Stokes Raman scattering (CARS) microscopy is a recently developed vibrational spectroscopic imaging technology with broad applications in biology and medicine [1–4]. CARS is a third-order nonlinear optical process in which a pump-probe beam (ωp) and a Stokes beam (ωs) interact with molecules in the specimen. When the beat frequency (ωp–ωs) is tuned to be resonant with a given vibrational mode, a strong anti-Stokes signal is generated at the frequency of ωas=2ωp–ωs [1,4]. CARS microscopy holds the promise for noninvasively imaging complex systems with high spatial resolution, high sensitivity, and label-free chemical specificity. Its clinical applications have been promoted by advanced developments , including epi-detection , multiplex acquisition [7–9], single-beam excitation [10,11], spectral focusing [12,13], nonlinear fiber wavelength conversion , time-resolved probing  and interferometry [16, 17]. Despite these advances, current CARS microscope can only be operated in a dark environment because the CARS signal currently relies on highly sensitive photomultiplier tubes (PMT) which can be easily saturated by room light. This remaining challenge has blocked the in situ applications of CARS microscopy in an operating room and in the field.
Here, we address this challenge through the development of ambient light CARS (AL-CARS) microscopy. In our scheme, the laser is modulated at megahertz frequency. Under ambient light, a photodiode equipped with a lab-built resonant amplifier selectively picks up the CARS signal at the same modulation frequency. Bandpass filters are further used to spectrally block the laser and the room light. Together, by filtering in both frequency domain and spectral domain, we have succeeded in blocking the environment light and sensing CARS signal with sufficient signal-to-noise ratio (SNR). By our AL-CARS setup, spectroscopic images of human breast cancerous tissue in situ were obtained. Through multivariate curve resolution (MCR) analysis , the fibrosis and cytoplasm were distinguished.
In AL-CARS, both band pass filters and the modulation serve as the obstacle to ambient light. The band pass filters block the excitation laser as well as most of the ambient light, releasing only the CARS signal and room light around the CARS wavelength to the detector. By modulation of the laser and demodulation of the signal at megahertz, we further reject the environment light in the frequency domain. The ambient light entering the detector has an optical frequency of about several hundred THz and is accompanied with the alternating current frequency of 60 Hz. Since the modulation frequency is quite different from either the optical frequency of the ambient light or the alternating current frequency, we are able to extract the CARS signal selectively by a resonant circuit and a lock-in amplifier.
The system is depicted in Fig. 1a. An ultrafast laser (InSight, Spectra Physics) with dual outputs provides two synchronized pulse trains. The tunable beam with 120 fs pulse duration served as the pump beam and was tuned to 798 nm for excitation of C-H bonds. The 1040 nm beam with 200 fs pulse duration serves as the Stokes beam and is modulated by an acousto-optic modulator (AOM, 1205-C, Isomet) at the frequency of 2.31 MHz. The two beams are combined and chirped by two 15 cm long SF57 glass rods. In this way, the Raman shift can be controlled by the time delay between the pump and Stokes beams. The pump and Stokes beams are sent into a lab-built microscope . A 60× water immersion objective lens (NA=1.2, UPlanApo/IR, Olympus) is used to focus the light onto the sample, and an oil condenser (NA=1.4, U-AAC, Olympus) is employed to collect the scattering signal. Two band pass filters (D680/100, HQ650/60, Chroma) are housed in a lens tube and attached to the photodiode case. The signal is collected by a large-area photodiode (S3994-01, Hamamatsu) and amplified by a lab-built resonant circuit with a 2.25 MHz central frequency and 300 kHz bandwidth (Fig. 1b). The lock-in amplifier (HF2LI, Zurich Instrument) extracts the modulated CARS signal. Hyperspectral CARS images are acquired by scanning the temporal delay between the chirped pump and Stokes beams [12, 20]. By this spectral focusing scheme, the Raman shift of each frame in hyperspectral images varies as the time delay changes.
The calibration of the Raman shift with respect to the temporal delay was completed by using Raman peaks of known chemicals. A modified Kramers-Kronig method [21–23] was employed to extract the equivalent Raman spectra from measured CARS spectra which contain a vibrationally resonant signal and a nonresonant background. By comparing the phase retrieved CARS spectral profile of known chemicals with their spontaneous Raman spectra, the relation between motor step position and Raman shift is obtained and shown in Fig. 1c. The relation can be described by a linear fitting with R2=0.99.
In order to compare our AL-CARS setup with “dark” CARS detected by PMT, we recorded the AL-CARS image of Dimethyl sulfoxide (DMSO) (Fig.2a) and the “dark” CARS image of the same sample (Fig.2b), respectively, under the 60 Hz room light. The AL-CARS setup worked well in the lighted room, where the photodiode was not saturated and the modulated CARS signal was successfully extracted from the ambient light. In contrast, Fig.2b revealed that the PMT was saturated by the lamp light when the voltage for PMT was 250 V. Because the CARS signal from a biomedical sample is much weaker than that of DMSO, for CARS imaging of tissues, the voltage for PMT must be much higher than 250 V. Under such condition, the environment light would definitely saturate the PMT or even damage the detector. Together, this comparison shows the advantage of our scheme for ambient light applications of CARS microscopy.
To characterize the signal and the source of noise in AL-CARS, we measured the CARS signal and noise level as a function of the excitation laser power. The intensity of CARS signal can be expressed as :
χ(3) is the third-order nonlinear susceptibility containing a vibrationally resonant and a nonresonant component. Fig. 3a shows that the CARS intensity is proportion to Ip2IS, which is in agreement with equation (1). There are three major noise resources in a CARS microscopic system, namely the shot noise, the laser intensity noise and the detector electronic noise . The relationship between these noise resources can be written as:
The electronic noise is a constant, and the shot noise is known to be proportional to the square root of the power, here, the intensity of the CARS signal. In our AL-CARS microscope, the total noise in the signal raises with the increase of the excitation intensity (Fig. 3b). When Ip2IS is below 5×105 mW3, the electronic noise is dominant and the total noise increases slowly with the laser intensity. As Ip2IS exceeds 5×105 mW3, the total noise level increases linearly with the laser intensity. As the CARS signal level has a linear relationship with Ip2IS, the shot noise is proportional to the square root of Ip2IS accordingly. Thus, the linear relationship between noise and Ip2IS in Fig. 3b indicates that the AL-CARS setup is laser noise dominant when Ip2IS exceeds 5×105 mW3. Consistently, the SNR increases quickly at the very beginning, and becomes a constant as the excitation intensity is raised (Fig. 3c).
To evaluate the detection sensitivity of our AL-CARS imaging setup, we used DMSO at different volume concentrations as a test bed. Fig. 4a shows a representative phase-retrieved image of DMSO solution diluted by Deuterium Oxide (D2O). The CARS spectra of different concentrated DMSO samples can be found in Fig.4b. When the DMSO volume concentration is as low as 1%, the peak of 2913cm−1 Raman shift is still clear, which indicates that the detection limit of the setup can reach 1% DMSO.
In equation (1), χ(3) is the term that generates CARS signal. From microscopic view of molecules, χ(3) is proportional to the bulk number density [24,25]. Connecting the macroscopic tensors and molecular tensors, there should be a quadratic relationship between CARS signal intensity and DMSO density. The results in Fig. 4c are consistent with the theory. Meanwhile, the SNR of each image is related to the DMSO concentration by linear fitting, resulting in a linear correlation with R2=0.99 (Fig. 4d).
Finally, we demonstrated the capability of AL-CARS scheme for in situ spectroscopic imaging of breast tissue. Forward-detected hyperspectral CARS imaging mapped the human breast cancer cells and the stroma based on their distinct chemical composition. The CARS image at 2920 cm−1 shows the morphology of the cancerous tissue (Fig. 5a). By phase retrieval and MCR analysis, we were able to decompose the hyperspectral CARS data set into a chemical map containing two major components (Fig. 5b). The spectral profile (Fig. 5c) with a strong peak around 2930cm−1 is assigned to the fibrosis in the stroma, and the weaker and broader peak is assigned to the protein-rich cytoplasm. The nuclei showed a dark contrast.
In summary, the current work demonstrated a new scheme that allows CARS imaging under ambient light, in which a photodiode with a lab-built resonant amplifier is used as the detector. The band pass filters in the spectral domain as well as modulation in the frequency domain blocks the environment light. We note that stimulated Raman scattering (SRS) [26, 27] works well under ambient light. Nevertheless, SRS microscopy suffers from laser noise and requires a high-quality solid state laser as the excitation source. Because the CARS signal appears at a new wavelength, the laser noise is not a big issue. Our AL-CARS scheme is applicable to a compact fiber laser as the excitation source. Moreover, epi-detected CARS imaging is doable. These features collectively render AL-CARS a promising tool for in situ and in vivo clinical applications and field uses.
Funding. National Institutes of Health (NIH) grants (GM104681, CA182608). National Natural Science Foundation of China (NSFC) (61505143)
The authors thank Professor Masanobu Yamamoto of Purdue University for constructive suggestions on the signal detection, Chi Zhang and Delong Zhang of Purdue University for helpful discussion. Y.Z. is grateful to China Scholarship Council (CSC) for financial support in the United State (No. 201406255077), and thankful to Professor Zhanhua Huang of Tianjin University for the support for the study abroad.
OCIS codes: (300.6230) Spectroscopy, coherent anti-Stokes Raman scattering, (300.6380) Spectroscopy, modulation, (170.1610) Clinical applications