Nanosecond SRS fiber amplifier for label-free near-infrared photoacoustic microscopy of lipids

Abstract

Near-infrared photoacoustics receives increasing interest as an intravital modality to sense key biomolecules. One of the most central types of biomolecules of interest are lipids as they constitute essential bio-hallmarks of cardiovascular and metabolic diseases and their in-vivo detection holds insightful information about disease progression and treatment monitoring. However, the full potential of near-infrared photoacoustic for high-resolution and high-sensitivity biomedical studies of lipids has so far not been exploited due a lack of appropriate excitation sources delivering short-pulses at high-repetition-rate, high-pulse-energy, and wavelength around 1200 nm. Here, we demonstrate a custom-built SRS fiber amplifier that provides optical excitations at 1192.8 nm, repetition rates of 200 kHz, pulse durations below 2 ns, and pulse energies beyond 5 μJ. We capitalize on the performance of our excitation source and show near-infrared photoacoustics resolving intrinsic lipid contrast in biomedically relevant specimens ranging from single cells to lipid-rich tissue with subcellular resolution.

1. Introduction

The concentration, spatial distribution, and molecular composition of lipids are essential biomarkers for investigating atherosclerosis, coronary heart disease, obesity, and diabetes, among other diseases, as they define the onset, severity, and progression of the disease course. Technologies typically employed for studying lipids such as histopathology [1], mass spectroscopy [2], [3], [4], optical frequency domain imaging [5], [6], [7], and Raman [8], [9] or near- and mid-infrared microscopy [10], [11], [12], [13] require biopsy extraction, harmful labelling strategies, are based on indirect lipid detection, or are characterized by poor spatial resolution and penetration depths and, thus, prevent application in cells, animals, and humans in vivo. As a new emerging modality, photoacoustics has a unique capability for label-free and dynamic monitoring lipids at high resolution and penetration depth in biological tissue by accessing the intrinsic molecular absorption contrast [14], [15], [16], [17], [18], [19], [20].

Label-free photoacoustic imaging of lipids is based on exciting the intrinsic vibrational absorption fingerprint of the aliphatic chain (i.e., stretch and bend of the CH molecular bonds) having the fundamental vibrational mode located in the mid-infrared range at ~3500 nm (~2860 cm⁻¹) [21], [22], [23] and its first and second overtones at ~1700 nm (~5880 cm⁻¹) and ~1200 nm (~8330 cm⁻¹), respectively [22], [24], [25], [26], [27], [28], [29]. Whereas increasing overtones are characterized by decreasing absorption coefficient and, thus, weaker photoacoustic signals when compared to the fundamental mode [24], [25], [27], imaging at wavelengths larger than 1700 nm requires special optical elements (e.g., reflective objectives, hollow-core fibers) for beam guidance, removing water vapor (e.g., N₂ flushing), and yield poorer spatial resolution due to longer wavelengths, and lower penetration depth due to water absorption [22], [25], [27]. Hence, photoacoustics for sensing lipids is best applied at the lipid's second overtone at ~1200 nm, also referred to as the NIR-II or SWIR spectral band, for high-resolution imaging of biological specimens with strong contrast and standard optics [30], [31], [32]. Near-infrared photoacoustic microscopy (NIR-PAM) based on laser technology emitting ~ns-pulses with ~µJ pulse energies at ~1200 nm and at ~100 kHz repetition rates would allow photoacoustics to assess lipid accumulations of biological specimens in micro- and endoscopic arrangements, and, thus, translate findings from basic research to the clinic. However, exploiting the full potential of PAM for imaging lipids on the microscale is currently hindered by a lack of light sources that deliver the required optical excitation in the NIR range with sufficiently high repetition rate, high pulse energy, and short pulse duration at low cost.

Available laser sources so far employed for time-domain NIR-PAM at ~1200 nm are diode-pumped solid-state lasers (DPSS) equipped with optical-parametric oscillators (OPO) [22], [25], [30], [33], [34], [35], [36]​, supercontinuum fiber lasers [26], [36], [37], [38] with wavelength selection filters, and stimulated Raman scattering (SRS) fiber amplifiers [36], [38], [39], [40]. OPO-based laser technology allows for optical excitations of up to µJ pulse energies, < 10 ns-pulse durations [25], [34], and up to kHz repetition rates ([25], [33]; NIR-PAM using OPO-based excitation was shown to measure lipid depositions in atheromatous biopsies [25], [30], Drosophila larva [25], murine ears [33] and adipose tissue and excised murine peripheral nerves [34], but has so far been limited in its application by a low repetition rate and spatial resolution.

Supercontinuum fiber lasers deliver up to 10 nJ/nm, sub-ns pulses, and high repetition rates of up to hundreds of kHz. For wavelength tuning, supercontinuum lasers are typically equipped with acousto-optic tunable filters (AOTF), acousto-optic modulators (AOM), variable linear filters (VLF), or bandpass filters [37]. The former two only allow for bandpass filtering below 5 nm bandwidth, which therefore leads to pulse energies below 50 nJ, and VLFs are so far not commercially available in the ~1200 nm region. Thus, only bandpass filters can be used for wavelength selection which limits specific fine-tuning of the optical excitation for photoacoustics. NIR-PAM using supercontinuum excitation was used for imaging lipids in intramuscular fat and Drosophila larva, but was demonstrated so far with > 10 µm resolution, < 10 kHz repetition rate, low power density, and > 10 ns pulse duration [37], which does not support high-speed, high-contrast, and high-resolution microscopy. Although optical-fiber based light sources are typically characterized by high efficiency, a small footprint, broad accessibility, and low cost, there were only a few attempts to develop fiber laser sources for NIR-PAM because the 1200 nm band is not supported by rare-earth-ion-doped fiber amplification. SRS fiber amplifiers are promising alternatives to overcome the lack of direct gain media for generating optical excitations around 1200 nm by utilizing a nonlinear optical process that produces one or more Stokes frequency downshifts to pump light. When the pump light's peak power reaches the Raman threshold, cascaded frequency shifts arise with an interval of 13.2 THz (in typical silica single mode fiber). Wavelength selection can be realized either by bandpass filters (>5 nm) [39], [40], [41], [42] or Raman seed lasers based on four-wave mixing (FWM) seeding the SRS and leading to narrow spectral bandwidths (<0.1 nm) [43], [44]. Fiber SRS sources have been shown to deliver optical excitation for NIR-PAM with < 10 µJ/pulse, < 50 kHz repetition rates, and wavelength coverage between 1064 and 1325 nm [39], [41], [42]. NIR-PAM using SRS fiber amplifiers have so far been used for lipid phantoms with ~1 µm resolution [41] and fixed Drosophila larva with ~13 µm spatial resolution [42], but not for imaging specimens relevant for biomedical research due to low speed, low power, large bandwidth, or low spatial resolution. In conclusion, the lack of an excitation source that delivers optical excitation with high pulse energy and repetition rate, near 1200 nm, and with short pulse duration limits the applicability of NIR-PAM for biomedical research.

Building from past fiber laser development [45], [46], we designed a versatile excitation source for NIR photoacoustic lipid sensing at ~1200 nm. In this study, we report on an SRS fiber amplifier for NIR-PAM of lipids having a repetition rate beyond 200 kHz, pulse energy of 5 μJ, and pulse duration below 2 ns at 1192.8 nm, selecting this wavelength as one of the discriminating lipid peaks in intravital sensing [47]. We employed the custom-designed SRS fiber amplifier in a laser-scanning transmission-mode photoacoustic microscope, allowing diffraction-limited imaging with 2.4 µm lateral resolution. We capitalized on this imaging performance to demonstrate the applicability of the SRS fiber amplifier for label-free NIR-PAM at 1200 nm to investigate lipid accumulations in biomedical specimens with subcellular resolution. The presented excitation source introduces a new concept for providing optical excitation for photoacoustics at 1200 nm and expands the applicability of high-speed, high-contrast, and high-resolution NIR-PAM in biomedical research.

2. Materials and methods

2.1. Nanosecond-pulsed FWM seeded SRS fiber amplifier

Fig. 1(a) schematically depicts the developed SRS fiber amplifier. The Raman pump was configured with a master oscillator fiber amplifier (MOFA) seeded by a distributed feedback laser diode (DFB-LD, DFB-1068-PM-50, Innolume Inc.) operating at 1068.1 nm. The Raman pump was controlled by an electro-optic modulator (EOM, NIR-MX-LN-10-PD-20 dB, IXblue) with a 3 dB bandwidth of 12 GHz driven by a fast pulse generator (28A240–9, Highland Tech.) for delivering pulse duration adjustable between 1 and 5 ns. We pre-modulated the DFB-LD with 10 ns pulses to suppress leaking continuous-wave (CW) light, which could deplete the efficiency of the following ytterbium-doped fiber amplifiers (YDFAs) considering a limited ~20 dB extinction ratio of the EOM. The pulsed Raman pump was pre-amplified with two single-mode YDFA stages, each consisting of a ~1 m YDF (Yb300–6/125-PM, Liekki) that was core-pumped by its own 976 nm laser diode (LD, LU0975M500, Lumics) in the backward direction via a wavelength division multiplexer (WDM). Isolators before each YDF protect fiber components from backward propagating light and 2 nm bandpass filters (Oz Optics Inc.) at 1068.1 nm remove the amplified spontaneous emission (ASE) between gain stages. The total gain of the two pre-amplifiers was measured to be ~25 dB (Fig. S1(a) in the Supplementary Material), although each amplifier had a gain of 15 dB due to a saturation of the single-mode amplifiers. Before the power amplifier, a fiber-Bragg-grating (FBG)-stabilized LD was coupled via a 1068.1 nm/1130 nm WDM. The LD was used as the seed light source of the FWM-seeded SRS amplification for a spectrally narrow output. It was pre-modulated to 100 ns pulses to prevent depletion of the power amplifier. The 1068.1 nm Raman pump as well as the seed light were linearly polarized by an in-line polarizer to reduce polarization-dependent effects and ensure environmental stability. The power amplifier utilized a 4 m double-clad (DC) YDF (Yb1200–10/125-PM-DC, Liekki) pumped in the forward direction by a high-power multi-mode 976 nm LD via a multi-mode pump combiner. The Raman threshold power, P_th, for generating the SRS can be described as follows [40]:

$${\text{P}}_{\text{th}}\text{≈}\frac{16{\text{A}}_{\text{eff}}}{{\text{L}}_{\text{eff}}{\text{g}}_{\text{r}}}$$ (1)

where A_eff is the effective core area, L_eff is the effective fiber length, and g_r is the Raman gain coefficient in the medium. In the absence of Raman seed light and/or passive components such as FBG mirror sets, the 13.2 THz Raman frequency downshift would dominate and generate a broad SRS spectrum with a frequency range from 11 to 15 THz [48] typical for fused silica fiber (Fig. S1(b) in the Supplementary Material). Inclusion of the Raman seed LD allowed for a spectrally narrow output considering the Raman gain coefficient of the SMF-28. We selected 1127 nm as the seed LD's wavelength to generate the 2nd Stoke wavelength at 1192 nm for NIR-PAM of lipids. When the Raman pump's peak power reached the 1st Raman threshold power, the optical energy was transferred to the seed wavelength (1st Stoke) and amplified within a narrow spectral region until the power reached the next Raman threshold. Subsequently, the 2nd Stoke light was generated at an equal frequency spacing (14.7 THz) between the Raman pump and the seed frequencies by FWM, and was further amplified by the SRS process.

Fig. 1.

Schematic depiction of the NIR-PAM system based on an SRS fiber amplifier. (a) Nanosecond SRS fiber amplifier experimental setup. (b) NIR-PAM schematic. (c) Output spectrum of the SRS fiber amplifier (red box: expanded scale) (d) Time trace of the SRS fiber amplifier (red box: expanded scale). Abbreviations: ADL, achromatic doublet lens; AMP, amplifier; BPF, bandpass filter; CMS, cladding mode stripper; COL, collimator; DC, double clad; EOM, electro-optic modulator; GMs, galvanometric mirrors; ILP, in-line polarizer; ISO, isolator; LD, laser diode; M, mirror; MLS, motorized linear stage; MMLD; multi-mode laser diode; MMPC, multi-mode pump combiner; NDF, neutral density filter; OC, optical coupler; OL, objective lens; OSA, optical spectrum analyzer; PC, polarization controller; SL, scan lens; UT, ultrasound transducer; WDM, wavelength division multiplexer; YDF, ytterbium doped fiber.

2.2. Optical resolution near-infrared photoacoustic microscope

As schematically depicted in Fig. 1(b), 99% of the light source was coupled through free space optics via a fiber-coupled FC/APC collimator. The beam was attenuated by neutral density filters and guided over galvanometric mirrors (GVS202, Thorlabs). Subsequently, the beam was enlarged by a telescopic arrangement of a scan lens (LSM03, Thorlabs) and an achromatic doublet lens (AC254, Thorlabs) and focused by a microscope objective lens (10x, NA: 0.26, Mitutoyo) through the bottom facet of a petri dish, held by a set of two motorized linear stages (SGSP20–20, Sigma Koki) (see 2.3 Sample preparation). For PA signal detection, a spherically-focused, 25 MHz ultrasound transducer (UT, V324-SM, focal distance: 15 mm, Olympus) was located in transmission mode and acoustically coupled to the specimen by a water or PBS droplet. The UT was further positioned at ~500 µm defocus for capturing a larger area of the specimen than the acoustic focal area and for avoiding artifacts that could originate from laser pulses transmitting through the sample and acting directly on the UT's piezoelectric element. All cables are double shielded and additionally sheathed with a metal cable shielding sleeve to suppress electronic noise. The acquired photoacoustic signals were amplified by 48 dB using two RF-amplifiers (ZFL-500LN-BNC+, Mini-Circuits, 500 MHz bandwidth) and recorded using a high-speed 12-bit DAQ card (Alazar ATS9373, sampling rate per channel: 500 MS/s, +/- 400 mV Input Range). The acquisition as well as the movement of both scanning units, i.e. galvanometric mirrors and sample holding stages, were triggered by a 200 kHz TTL signal from the arbitrary function generator (AFG, AFG3012C, Tektronics) that controlled the SRS source, thereby synchronizing the excitation and detection of NIR-PAM signals. Raw signals were averaged, bandpass filtered in the range of 5–45 MHz, and projected in 2D or 3D using the maximum amplitude projection. The entire system was controlled using a custom-developed Matlab code (Matlab, Mathworks, Natick, USA). Image post-processing was carried out in ImageJ (ImageJ 1.53 m, National Institutes of Health, Bethesda, USA [49]) including gamma correction, outlier removal, and histogram normalization.

2.3. Sample preparation

All animal experiments and care conformed to the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Massachusetts General Hospital Institutional Animal Care and Use Committee (IACUC) and the Animal Care and Use Review Office (ACURO) of the U.S. Army Medical Research and Development Command. All phantoms and biological specimens were positioned in 170 µm-thick glass-bottom petri dishes (20 mm micro-well, #1.5 cover glass, Cellvis).

Reference phantoms: for characterizing the imaging performance of the NIR-PAM setup regarding resolution, field of view (FOV), imaging speed, sensitivity field, and 3D imaging capabilities, we used 1 µm polystyrene beads (PolyScience) embedded in low melting temperature agar, 50 µm black polyethylene suture phantoms (Dafilon, 6/0, B.Braun), and black varnish. The UT was coupled to the specimens by a droplet of deionized water.

Human adipocytes: human brown and white adipocyte progenitors were isolated from a human cadaver specimen (neck) and immortalized as previously described [50]. Immortalized progenitor cells were cultured in Dulbecco's Modified Eagle Medium (DMEM) with high glucose supplemented with 10% Fetal Bovine Serum (FBS). For adipocyte differentiation, cells were grown in glass-bottom petri dishes until they reached full confluency and then treated with adipogenic induction media (DMEM with high glucose with 10% FBS, 33 μM biotin, 0.5 μM human insulin, 17 μM pantothenate, 0.1 μM dexamethasone, 2 nM T3, 500 μM IBMX, and 30 μM indomethacin) for 18 days. Differentiated cells were fixed in 10% formalin solution for 15 min and washed twice with phosphate-buffered saline (PBS). The UT was coupled to the specimens by a droplet of PBS buffer in order to prevent osmotic imbalances. After recording NIR-PAM images label-free, the fixed adipocytes were stained using Oil Red O (Cat3 O0625–100 G, Sigma Aldrich) for 15 min in darkness, washed three times with deionized water, counterstained with hematoxylin for 1 min (Mossberg), and washed three times with deionized water. After air-drying, drops of Ultramount Permanent Mounting Medium (CAT# S1964, DAKO) were added. After drying again, brightfield microscopy images of stained adipocytes were acquired using a photomicroscope (AxioPhot, Zeiss).

Sheep adipose tissue: specimens were obtained post mortem from four to eight months old Polypay-specific pathogen-free sheep that were euthanized following the completion of unrelated device (infrailiac bypass graft) research. The specimens were cut into ~1 × 1 mm² samples and sandwiched between a glass bottom petri dish and a cling film by centrifuged ultrasound gel (Aquasonic, Parker) to allow acoustic signal transmission. The UT was coupled to the cling film using a water droplet.

Mouse ear: specimens were obtained post mortem from six-months old male C57BL/6 mice that were euthanized following an unrelated research protocol (investigating liver fibrosis). Ears were excised from the murine carcasses by hot scalpel to cauterize the incision line. For imaging, entire excised ears were sandwiched between a glass bottom petri dish and a cling film by centrifuged ultrasound gel (Aquasonic, Parker) to allow acoustic signal transmission. The UT was coupled to the cling film using a water droplet.

3. Results and discussion

3.1. Characterization of nanosecond SRS fiber amplifier

Fig. 1(c) shows the output spectrum of the amplified 2nd Stokes shifted light of the laser source. The FWM seeded SRS amplification processes are as follows: (1) amplification of 1068 nm pump light in YDF, Fig. S2(a) to (b); (2) energy transfer (e.g., Raman amplification) to 1127 nm (1st Stokes) and FWM generation (2nd Stokes), Fig. S2(b) to (c); 2nd Raman amplification to 1192.8 nm (2nd Stokes), Fig. S2(c) to (d). We verified that the 2nd Stokes shifted light is generated at an equal frequency spacing of 14.7 THz between the Raman pump and seed light. The measured spectral bandwidth and pulse energy at 1192.8 nm are 70 pm, and 5 µJ, respectively, at the single-mode-fiber output. In contrast with general SRS sources [39], [40], the FWM seeded SRS amplifier shows a higher power density (over 70 μJ/nm) within the narrow spectral bandwidth and the highest peak power is at the 2nd Stokes shifted wavelength with a side-mode suppression ratio (SMSR) of 25 dB. Pulses with a duration of 2 ns were emitted at a repetition rate of 200 kHz (Fig. 1(d)) with a stability of less than 5% (see Fig. S3 in the Supplementary Material).

3.2. Characterization of the NIR-PAM system

The excitation optical fluence incident on the sample was measured to be 467 nJ/pulse. Such high optical excitation fluence was required due to the narrowband frequency coverage and long working distance of the ultrasound transducer used in the current system. While repeatedly imaging the samples presented in this report, we have not noticed laser-induced changes in tissue composition or microstructure.

Fig. 2(a) shows raw and frequency filtered signals from a black suture phantom (100-pulse average; 231 mV amplitude, 3.33 mV noise, SNR = 20 log(signal/noise) = 36.83 dB). The associated frequency response (Fig. 2(b)) shows a − 12 dB frequency coverage spanning 5–40 MHz. The resolution was determined by measuring 1 µm beads embedded in low melting temperature agar. Imaging was performed with 50 × 50 pixels across a FOV of 42 × 42 µm², averaging 50 frames and using a neutral density filter with an optical density of 0.5 to reduce potential damages to the bead. The data was projected onto the xy- and xz-plane using a maximum amplitude projection and assuming 1530 m/s as the speed of sound in water (see Fig. 2(c)). The resolution was calculated by fitting a Gaussian model to the axial and lateral line plots and using:

$$\text{Resolution=}\sqrt{{\text{d}}_{\mathit{exp}}^2\text{-}{\left(\frac{1}{3}\sqrt{2\text{∙}\mathit{ln}\left(2\right)}\text{∙}{\text{d}}_{\text{nom}}\right)}^2}$$ (2)

where $d_{\mathit{exp}}$ = 2.40 µm is Gaussian the full width at half maximum (FWHM) of the imaged profile (R² = 0.9729). Eq. (2) assumes that the optical focus as well as the bead's diameter can be assumed as Gaussian profiles with the nominal diameter $d_{\mathit{nom}}$ = 1 µm corresponding to ± 3σ. The lateral resolution was thus approximated as the convolution of two Gaussian profiles and was found to be 2.4 µm, comparing well to the theoretical limit of 2.29 µm. After Hilbert-transformation and projection along the optical propagation direction, the axial resolution was found to be 23.3 µm (assuming 1530 m/s as the speed of sound in water; R² = 0.9915; see Fig. 2(c)). The 2D sensitivity field was measured using a layer of black varnish and positioning the UT at ~500 µm positive defocus to capture a larger area than the acoustic focus and avoid noise from laser pulses directly acting on the UT. For testing different scanning modes, suture phantoms were used to characterize and calibrate the scanning speeds. Considering a drop of sensitivity from the UTs' central field to 1/e², an area of 465 × 465 µm² could be acquired when using only laser-scanning or a stripe of 465 µm and arbitrary length when using combined laser-and-sample scanning. When using sample scanning or stitched laser-scanning arbitrary large FOVs can be acquired (Fig. 2(e)). Whereas for pure laser-scanning and combined laser-and-sample-scanning with a pixel averaging of N_Avg, the high-speed data acquisition allows recording all laser pulses and, thus, a pixel scanning frequency of 200 kHz/N_Avg, sample-scanning is limited by the maximum step-scanning speed of the sample holding stage. For typical imaging settings of 1 µm² pixel size and N_Avg = 20, the former two capture areas at 1/100 mm²/s and the latter at ~1/1000 mm²/s.

Fig. 2.

Characterization of the imaging performance of the developed NIR-PAM system. (a) NIR-PAM signal and corresponding (b) frequency response of a suture phantom. (c) xy- and xz-image of a 1 µm polystyrene bead and (d) resolution determination. (e) Sample-scanning, laser-scanning, and combined sample-and-laser scanning of a suture phantom.

3.3. NIR-PAM images of biomedical specimens

In order to demonstrate the applicability of the SRS fiber amplifier for NIR-PAM of endogenous lipid contrast, we performed imaging experiments in biological and biomedical specimens typical in the field. Fig. 3(a) shows intracellular lipid droplets in unstained (see stepwise Zoom-ins of a green box in Fig. S4 in the Supplementary Material), differentiated, fixed white human adipocytes imaged at 500 averages (see analogous imaging of brown human adipocytes in Fig. S5 in the Supplementary Material). The system's high resolution resolved lipid droplets down to 3.2 µm in diameter and achieves an SNR of 27.1 dB for white adipocytes and 25.7 dB for brown adipocytes, caused by smaller lipid droplets and lower lipid concentration in the droplets. Histopathological imaging using Oil Red O staining after the label-free NIR-PAM imaging and comparison to other microscopy studies of lipid contents of adipocytes [21], [51], [52], [53] is in very good agreement and confirms the ability to detect intrinsic lipid concentration in lipid droplets with subcellular precision. Dark areas in the histology of Fig. 3(a) are attributed to hematoxylin staining of nucleic acids. Fig. 3(b) shows label-free imaging of clavicular adipose tissue of sheep imaged at 100 averages recorded in stitching laser-scanning FOVs (see analogous imaging of sheep omentum adipose tissue in the Fig. S6 in the Supplementary Material). Magnified insets in Fig. 3 highlight single-cell structures of the adipose tissue surface with an SNR of 35.0 dB, which are in very good agreement with other published microscopy studies of adipose tissue [54], [55], [56], [57]. Fig. 3(c) shows the lipid contrast of a mouse ear ex vivo recorded at 50 averages. Expanded insets illustrate lipid content dominated of epidermal keratinocytes and hair follicles with an SNR of 22.2 dB, which resembles other lipid microscopy studies of mouse ears to a high degree [58], [59], [60].

Fig. 3.

NIR-PAM imaging of intrinsic lipid contrast in biological and biomedical specimens. (a) Fixed differentiated human white adipocytes show intracellular lipid accumulations in lipid droplets. Zoom-ins and Oil Red O staining confirm spatial co-registration of lipid contrast. (b) Stitched-imaging of ex vivo sheep clavicular adipose tissue and Zoom-ins show adipocyte structure. (c) Ex vivo imaging of a mouse ear illustrates lipid content in epidermal keratinocytes and hair follicles. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

4. Conclusion

We demonstrate a new approach for NIR-PAM for label-free imaging of intrinsic lipid contrast with high-resolution, high-sensitivity, and high-speed by exciting the second CH₂ vibrational overtone. Whereas the NIR-II or SWIR spectral band is advantageous for biomedical photoacoustic microscopy due to wavelengths allowing for standard optics and high resolution, a lack of versatile excitation sources previously limited its full potential. The demonstrated NIR-PAM system is based on a SRS fiber amplifier as a new type of excitation source that addresses the need for near-infrared photoacoustics at ~1200 nm by offering a 200 kHz repetition rate, pulses of < 2 ns duration, and 5 µJ pulse energies. We capitalize on the unique performance of our NIR-PAM system and showcase the applicability to resolve intrinsic lipid contrast in biomedically relevant specimens ranging from single cells to lipid-rich tissue. Our NIR-PAM resolves lipid contrast label-free with subcellular resolution and SNR up to 35 dB at high speed.

The SRS fiber amplifier design may open new applications for near-infrared photoacoustics by enlarging the excitation spectral range in a cost-effective manner. In here we demonstrate the general concept of using SRS fiber amplifiers for near-infrared photoacoustics to qualitatively sense lipid embeddings in cells and tissues. In future studies, we will redesign the optical assembly and signal detection to characterize the system’s sensitivity and specificity quantitatively. Utilizing ultrawide bandwidth transducers with shorter working distances and/or all-optical detection of ultrasound will allow the collection of higher frequencies and, thus, similar SNRs at significantly lower optical fluences. We anticipate that these modifications will permit imaging in vivo while adhering to ANSI exposure limits as is typical for NIR-PAM ([61], [62]). Future generations of this excitations source will target other vibrational biomolecular contrast, especially to sense water at 1400 nm, proteins at ~1500 nm (combination overtones of Amide I and II), and carbohydrates around ~1600 nm by equipping additional high-nonlinear Raman fibers. We aim to capitalize on the narrow spectral band of the SRS fiber amplifier technology to characterize the molecular composition of cells and tissues. Furthermore, a straightforward integration for endoscopic imaging system will facilitate the fiber amplifier's application for in vivo intravascular lipid detection for cardiovascular research. Besides adapting the excitation wavelength to specific molecular contrast of interest, the flexible triggering scheme of the laser allows for freely tuning the repetition rate or interpulse time delay. Thus, we aim to exploit this unique capability for exploring time- and frequency-domain photoacoustic microscopy, burst and chirp acquisitions, and non-linear photoacoustic effects such as Grueneisen-imaging. SRS fiber amplifiers thus pave the way for high-end translational near-infrared photoacoustic microscopy, mesoscopy, and endoscopy for label-free lipid detection.

Funding

This work was supported by the National Institutes of Health, USA (P41 EB015903). MRS was supported by the OSA Thomas F. Deutsch Fellowship, Optica (formerly known as The Optical Society), USA.

Disclosures

HL, MRS, NL, and BEB declare no conflict of interest. GvS is an advisor to and holds equity in Kaminari Medical BV. SKN is an inventor of coagulation sensing technology licensed to Coalesenz, Inc., a company that is developing a point-of-care system to measure coagulation parameters. SKN has equity and serves on the board of the Coalesenz, Inc, and her interests were reviewed and are managed by Massachusetts General Hospital and Partners Healthcare in accordance with their conflict-of-interest policies.

Code availability statement

The code that supports the findings of this study is available from the corresponding author upon reasonable request.

CRediT authorship contribution statement

HL, MRS, and BEB conceived the idea of near-infrared photoacoustic microscopy of lipids in biological tissue. HL, NL, and BEB conceptualized the SRS fiber amplifier. HL built and characterized the laser’s performance. MRS built the photoacoustic microscope, conceptualized and performed the imaging experiments, data and image processing, and characterized the imaging characteristics. SKN and GvS guided the project with helpful discussions regarding the construction of the laser and the imaging experiments. HL and MRS wrote the manuscript and all authors read and edited the paper.

Declaration of Competing Interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Brett E. Bouma reports financial support was provided by National Institutes of Health. Markus R. Seeger reports financial support was provided by The Optical Society. Gijs van Soest reports a relationship with Kaminari Medical BV that includes: equity or stocks. Seemantini K. Nadkarni reports a relationship with Coalesenz Inc. that includes: equity or stocks.

Biographies

Hwidon Lee received his PhD degree in the Department of Cogno-Mechatronics Engineering at Pusan National University, South Korea, in 2015. He is currently working at Massachusetts General Hospital and Harvard Medical School as a Research Fellow (2019). His current research interests are the developments of novel light sources for biomedical applications including optical coherence tomography and photoacoustic imaging.

Markus R. Seeger is a biomedical microscopist developing advanced intravital imaging methods using optical and photoacoustic modalities for translational research. During his training as a biophysicist (MSc Goethe University Frankfurt 2015) and his research stay abroad (Fudan University Shanghai 2014), he started working on using light-matter-interactions as microscopic and spectroscopic methods for biomedical applications. Seeger conducted his doctoral research in Medical Life Science and Technology (PhD Technical University Munich and Helmholtz Zentrum Munich 2020), investigating multimodal microscopy for biomedical imaging. He joined the Wellman Center (Harvard Medical School and Massachusetts General Hospital) in January 2021 as a postdoctoral research fellow and as the recipient of the 2020/2021 OSA Thomas F. Deutsch Fellowship, where he develops next-generation photoacoustic biomicroscopy for label-free and dynamic interrogations of an organism’s structural, functional, molecular, and metabolic (micro)environment.

Norman Lippok currently holds a position as an Instructor at Harvard Medical School. He received his undergraduate training in Physics Engineering in Jena (Germany) and completed his dissertation research at the Physics Department, University of Auckland (New Zealand). His research focuses on new imaging signatures for functional and molecular tissue contrast as well as high speed swept sources, stepped frequency combs, and integrated photonics for biomedical imaging applications.

Seemantini K. Nadkarni is an Associate Professor at Harvard Medical School and directs her laboratory at the Wellman Center for Photomedicine at Massachusetts General Hospital. Her work encompasses both fundamental and translational areas of research, primarily focused on the development and investigation of novel optical technologies for applications in Hematology, Cardiology, cancer research and in vitro diagnostics.

Gijs van Soest leads the “Invasive imaging” research group, which investigates catheter-based imaging technologies, primarily aimed at guidance of cardiovascular interventions but with an eye to other applications (diagnostic imaging, imaging in lung or gastrointestinal tract) as well. Van Soest is an experimental physicist by training (MSc Rijksuniversiteit Groningen 1997). He started working in optics and light scattering during his PhD (Universiteit van Amsterdam, 2001). He worked as a postdoc at the Royal Netherlands Meteorology Institute (KNMI) to investigate atmospheric ozone profiles using satellite-borne spectrometers. In 2005 he joined the Biomedical Engineering group of the Thorax Center at Erasmus MC. Starting a new research topic in optical imaging and taking on the existing research in intravascular ultrasound, he developed his own research line. He was appointed Full Professor in 2020.

Brett E. Bouma is a professor in Health Sciences and Technology at Harvard Medical School and the Massachusetts Institutes of Technology. His doctoral studies were in physics, with a focus on ultrafast and high-intensity laser physics, at the University of Illinois at Chicago. As a Post-doctoral Fellow and Research Scientist at MIT in the laboratory of James Fujimoto, he worked on the first time-domain endoscopic OCT technologies and as an early faculty member in Harvard Medical School, he conducted the first human gastrointestinal and cardiovascular OCT imaging studies. His laboratory developed the first high speed frequency domain OCT instruments and his patents have been licensed by five companies, leading to several commercial medical instruments.

Appendix A. Supplementary material

Supplementary material

.

Data availability statement

The data of this work is available from the corresponding author upon reasonable request.