1. Introduction

Hyperspectral imaging is rapidly becoming an essential tool for investigating the world around us [1,2]. By effectively providing the chromophore specificity present in a spectrometer along with the spatial discrimination afforded by an imaging technique, these devices generate a wealth of information useful across fields and disciplines. Conventional hyperspectral imaging via light transmission, reflection, and fluorescence can be used across a wide range of clinical, pre-clinical, and nondestructive testing pursuits. These techniques have demonstrated efficacy in applications from predicting crop health [3] or monitoring of food spoilage [4] to providing noninvasive disease diagnosis and surgical guidance [1]. However, spectroscopy of optical absorption contrast provides additional challenges. As many traditional approaches rely on measurement of a relative transmitted optical fraction, they become inappropriate for characterizing optical absorption on comparably thick and opaque samples. The ability to characterize and image such contrast would be highly valuable given the extensive range of endogenous optical absorption contrast available within biological tissues [5–7].

A potential solution to this conundrum comes from the field of photoacoustic modalities. This family of optical techniques excite thermo-elastic pressure waves within a sample using a short (commonly nanosecond) excitation pulse. As these pulses are absorbed by specific chromophores, the generated acoustic profiles encode spatial information regarding the distribution of said chromophores. Lower-resolution high-penetration embodiments of this technique commonly utilize a tunable flash lamp driven optical parametric oscillator (OPO) providing high pulse energies (mJ) at low repetition rates (<50 Hz) appropriate for use in photoacoustic tomography (PAT) [8]. However, such optical sources become inappropriate when moving to higher-resolution laser-scanning-based devices such as optical-resolution photoacoustic microscopy (OR-PAM) [9] or photoacoustic remote sensing (PARS) microscopy [10,11]. Since these devices scan point-by-point, they require higher repetition rates, and demand high beam quality to achieve near-diffraction-limited lateral performance on the sample. This has commonly led to the implementation of white-light optical sources [12] which may not provide sufficient optical fluence at the sample when a narrow-line excitation is desired. Moreover, for some sensitive samples such as open wounds or open surgical sites, the use of contact-based approaches such as OR-PAM may be impractical or incompatible with the desired target.

Here, we present a hyperspectral optical absorption imaging microscope based on a PARS detection pathway. The PARS approach replaces the acoustic transducer common to conventional OR-PAM devices with a co-focused detection beam that encodes photoacoustic thermal and pressure effects as back-reflected intensity perturbations. This non-contact photoacoustic device features a high-speed OPO driven by a nanosecond diode-pumped solid-state laser configured to deliver pulses at up to 1 kHz in the range from 210 nm to 1550 nm. The tunable nature of this device provides access to a wealth of endogenous biological contrast and structure. Exciting in ultraviolet (UV) (∼260 nm) targets DNA contrast showing subcellular nuclear structural information. Exciting in the visible range (∼400 nm to ∼600 nm) targets hemeproteins such as cytochromes providing cytoplasmic structure, and hemoglobin highlighting erythrocytes and blood vessel structure. Finally, exciting in the infrared (IR) (∼1200 nm) targets lipids and collagen. This is of course not an exhaustive list. To date, infrared PARS devices have demonstrated contrast from oils samples [13], however, no studies have thus far been completed with regards to its efficacy in adipose tissues.

In this article we explore the efficacy of this device operating at salient wavelengths for various specimen. We have previously reported works related to an earlier embodiment of this system looking at formalin-fixed paraffin-embedded (FFPE) human tissue slides, FFPE human tissue blocks, frozen sections of human skin taken from Mohs micrographic surgery (MMS), and freshly excised unprocessed murine organs [11]. In these works, primarily DNA (250 nm) and hemeproteins (420 nm) were investigated highlighting cellular structure, erythrocytes, and cytoplasm. Moreover, the capabilities of the previous device were demonstrated operating as both a non-contact spectrometer and a hyperspectral absorption microscope in simple chromophore solutions and FFPE targets [14]. Here, this second-generation system with improved wavelength range was used to investigate live specimens by imaging chicken embryo chorioallantoic membrane (CAM) to explore in vivo DNA, hemoglobin, and lipid contrast for the first time in a single system. As well, unprocessed freshly resected murine adipose tissue was used to investigate lipid-rich samples with the newly implemented infrared excitation arm marking the first such report of a non-contact photoacoustic modality that is capable of visualizing individual adipocytes. These new capabilities are explored and discussed.

2. Method

2.1 Hyperspectral PARS system

The hyperspectral photoacoustic remote sensing (HS-PARS) system explored in these investigations is similar to that previously reported in [14] with several key differences. (i) The available wavelength range has been extended from 210 nm to 680 nm by adding an additional infrared band from 1050 nm to 1550 nm. This allows access to commonly reported lipid absorption regions (∼1210 nm) along with paraffin and water absorption. Consequently, the detection wavelength was changed from 1310 nm to 970 nm to facilitate the new infrared excitation range. 970 nm was selected as it represented an intermediate wavelength (between visible and lipid-infrared peaks). This facilitated the use of much longer excitation wavelengths from 1050 nm up to 1550 nm with readily available commercial dichroics which were not accessible in previous 1310 nm-based architectures. As for potential advantages and limitations of this new detection wavelength, the shorter wavelength is likely to provide reduced penetration depths in tissues and similar turbid media. The 970 nm beam would be expected to provide a smaller diffraction-limited focal spot size which may improve sensitivity due to improved overlap between the excitation and detection focal spots as compared with previous 1310-based architectures [14]. However, such potential benefits can be outweighed by relative responsivities of various InGaAs photodetectors. As well, since extremely deep penetration depths [15] were not a central goal of this project, introduced depth limitations were not expected to cause any notable performance issues. (ii) The previous reflective objective lens was replaced by a right-angle parabolic mirror. This offered superior power transmission characteristics and improved Gaussian beam propagation both of which are due to the lack of the central occlusion present in reflective objective systems. (iii) Finally, the mechanical scanning support system required redesign to accommodate tighter mechanical clearances which arose due to the effective shorter working distance provided by the parabolic mirror.

A diagram of the apparatus is shown in Fig. 1(a). The system consisted of a 3 ns wavelength-tunable excitation provided by a 1 kHz OPO driven by a diode-pumped solid-state laser (NT242, Ekspla). The power of this source was controlled via a variable attenuator. The output was then passed through three dichroic mirrors (HBSY134, DMSP505, and DMLP900, ThorLabs) to separate four wavelength bands (210 nm to 405 nm (UV); 405 nm to 505 nm (blue); 505 nm to 680 nm (green-red); and 1050 nm to 1550 nm (IR)). This was done such that we could manage each sub-range with appropriate achromatic optic systems and spatial filters. Spatial filters were required due to the highly astigmatic and elliptical output from the OPO. Three of these spatial filters (UV, blue, and IR) were implemented using pairs of air-spaced or cemented doublets (ACA254-100-UV, ACA254-100-A, and AC254-100-C-ML respectively, ThorLabs) each surrounding a pinhole. Meanwhile the green-red pathway spatial filter utilized an objective lens (RMS10X, ThorLabs) to focus into a single-mode optical fiber (S405-XP). These four excitation bands passed through either a variable beam expander (BE02-UVB, GBE02-A, and GBE02-C for the UV, blue, and IR paths respectively, ThorLabs) or a variable collimator (C40FC-A for the green-red path, ThorLabs) to provide variable divergence allowing for compensation of chromatic effects at the sample. Splitting the source range also allowed for normalization of pulse energies which varied significantly across the wavelength range. This was performed using an independent set of variable attenuators in each pathway. Another of the same series of dichroic mirrors was used to recombine the excitation pathways along with an additional dichroic (DMLP1000, ThorLabs) to combine the detection beam. The detection source was provided by a 970 nm superluminescent diode (SLD970P-A40W, ThorLabs). This fiber-coupled source was collimated using a variable collimator (C40APC-B, ThorLabs) and passed through a polarizing beamsplitter (PBS) (CCM1-PBS252, ThorLabs) and a quarter-waveplate (QWP) (WPQ05ME-980, ThorLabs) before joining the excitation pathways in the dichroic stack. Back-reflected detection from the sample passed back through the QWP and PBS being redirected through a bandpass filter (FELH0900 and FESH1000, ThorLabs) before being focused on to a photodiode (PDB425C-AC, ThorLabs) using an aspheric condenser lens (ACL25416U-B, ThorLabs). Final focusing of all beam pathways on to the sample was provided by a 90° off-axis parabolic mirror (MPD00M9-F01, ThorLabs). The imaging head is oriented as an inverted microscope to facilitate measurement and imaging of soft tissues and liquids.

 

Fig. 1. (a) System diagram of the reported system. Shorthand notation is defined: (AD) Achromatic Doublet; (Att.) Attenuator; (BPF) Bandpass Filter; (C) Collimator; (Cond.) Condenser; (M) Mirror; (OL) Objective; (PBS) Polarizing Beamsplitter; (PD) Photodiode; (PM) Parabolic Mirror; (QWP) Quarter Waveplate; (SMF) Single-mode Fiber. Dichroic mirrors are featured with their part number. (b) Diagram highlighting how water spectroscopy experiments were performed. (c) Diagram showing experimental layout used to acquire CAM frames. (d) Diagram of freshly resected rat adipose tissue experiments.

Download Full Size | PDF

Care was taken to ensure pragmatic achromatic performance across the extensive wavelength range. However, some chromatic contributions could not be fully removed including sample cover slips along with small chromatic variations within refractive components. To help minimize these issues each of the four excitation wavelength bands were aligned at their central wavelengths to maximize consistency across the full range. Additional wavelength-dependent variation was also introduced in the form of pulse energy variation. This aspect, along with spectroscopic correction approaches were previously discussed [14].

Mechanical scanning of the sample about the optical interrogation point was provided by two linear stepper motor stages (MFA-PPD powered by a XPS-D4, Newport). Photodiode signals and stage position feedback were captured using a digitizer (CSE161G4, Gage Applied). Acquisition and reconstruction approaches match previously reported methods [11,14]. The system could be used as both an absorption spectrometer by scanning a given region over fine excitation wavelength increments or could be used as a hyperspectral imaging platform by imaging at several salient wavelengths. Multiple wavelength acquisitions were acquired sequentially by switching the pump wavelength between frames or by sweeping the pump wavelength within a given region for performing spectral sweeps.

2.2 Samples

Several sample types were explored with this system including water, paraffin, paraffin-embedded human adipose tissue, in vivo CAM from chicken embryos, and freshly resected murine tissues. Water spectroscopy measurements were conducted by placing a drop of water onto a UV-transparent cover slip then focusing the interrogation point at the upper air-water interface (Fig. 1(b)). The CAM models used in these experiments were developed in house. Fertilized White Leghorn eggs were placed in a hatcher for 72 hours before being cracked and incubated for an additional 10 days in a vessel covered with a perforated film. These were then flipped into containers featuring UV-transparent cover slips such that the CAM could be pressed against this window to stabilize it for mechanical scanning. This allowed for scanning of the live subject by the system (Fig. 1(c)). Freshly excised murine tissues were obtained with the aid of collaborators at the Central Animal Facility, University of Waterloo under protocol ID: 41543 (Photoacoustic Remote Sensing (PARS) Microscopy of Resected Rodent Tissues, University of Waterloo). All murine tissue experiments were performed in accordance with the relevant guidelines and regulations. Freshly excised tissues were immediately placed in phosphate buffer solution and imaging was conducted within 3 hours of devitalization. For these experiments adipose tissues were taken from the Gluteal region of female Sprague Dawley rats. Imaging was performed by placing the adipose tissue onto a UV-transparent cover slip (Fig. 1(d)). For both the CAM and adipose tissues, no additional fluids or coupling media was added.

2.3 Blood oxygen saturation measurement

Given the multi-wavelength potential of this device a brief look at functional imaging was conducted. This consisted of extracting blood oxygen saturation (sO₂) from the micro vasculature of the CAM model. As we were interested to simultaneously show the efficacy of easily achievable 532 nm-pumped stimulated Raman scattering bands [16], 532 nm and 560 nm were selected. Measurement of sO₂ was performed by imaging the same sample region at each wavelength then assuming collected signals represented a linear super position of constituent chromophores based off their relative concentrations [17]. In addition, absorption of the detection wavelength was also considered in this model as it too will vary depending on chromophore concentration resulting in an inverse dependence on the collected PARS signals. Each point could then be solved by a simple linear system of equations.

3. Results

3.1 PARS infrared absorption spectra

The performance of the infrared excitation band was examined first. Three chromophores of interest in this range include water, adipose tissue, and paraffin which were interrogated at 5 nm wavelength increments. These sweeps are shown in Fig. 2. All three of these sweeps demonstrate similar absorption behavior with previously reported trends [6,18,19]. Measurements suggest that the HS-PARS system should be able to distinguish lipid-rich tissues from the background water signal by targeting the distinct 1210 nm peak. However, the lipid and paraffin signals overlapped significantly within this spectral region, likely preventing distinct recovery of adipocytes in paraffin-embedded samples. This was later confirmed in experiments with formalin-fixed paraffin-embedded (FFPE) blocks and slides where no distinct lipid contrast was observed at 1210 nm or other nearby wavelengths. Moreover, experiments involving formalin-fixed tissue chunks also yielded poor infrared contrast suggesting crosslinking from formalin fixation may significantly alter available lipid contrast in these samples.

 

Fig. 2. Several absorption measurement sweeps using the infrared wavelength band showing water, FFPE adipose tissue, and paraffin.

Download Full Size | PDF

3.2 Multi-wavelength CAM imaging

The hyperspectral imaging system next explored available contrast from chicken embryo CAM. Acquisitions were performed using 300 nm lateral step sizes. One such dataset is presented in Fig. 3. First, ultraviolet 250 nm was used to excite DNA contrast Fig. 3(a). The wavelength was selected as previous studies [11] have found 250 nm to provide optimal contrast between cell nuclei and the background tissue. This is despite the slight displacement from the ∼260 nm DNA absorption peak. These results represent the first such report of label-free visualization of live cell nuclei in a CAM model. Moreover, multiple distinct cell layers can be seen in this image with several more yet slightly out of focus of the system. This is consistent with the CAM anatomy which features two relatively cell-dense epithelium layers surrounding the sparser intermediate mesodermal internal layer where larger blood vessels are located [20]. Next, several visible wavelengths were selected to target hemeprotein absorption (420 nm, 532 nm, and 560 nm; Fig. 3(b)-(d) respectively) with hemoglobin likely being the dominant such chromophore in these instances. These acquisitions highlighted the fine capillary beds of the CAM. Discrepancies between these three wavelengths likely result from small sample motion as each wavelength acquisition required ∼16 min to acquire, and the live chicken embryo is free to move during imaging. Previous efforts with imaging CAMs directly with non-inverted imaging heads have been fraught with motion artifacts. This new flipped CAM model appears to heavily mitigate such substantial sample motion, providing reasonable lateral consistency between acquisitions while providing a flat imaging plane.

 

Fig. 3. Results from several wavelengths looking at a single region of a live chicken embryo CAM. (a-d) Shows each individual 300 µm x 300 µm acquisition at the respective wavelengths. (e) Shows a processed overlay of the UV acquisition highlighting DNA along with extracted sO2 values. Scale bar: 100 µm.

Download Full Size | PDF

Consistency between the 532 nm and 560 nm acquisitions was sufficient to extract sO₂. Unmixing was performed assuming extracted PARS signals represented linear super positions of constituent chromophores (oxy- and deoxyhemoglobin in this case) while accounting for wavelength-dependent excitation fluence. The extracted sO₂ is shown in overlay along with the DNA contrast in Fig. 3(e). Near-infrared wavelengths were also used on the subjects to target the 1210 nm lipid absorption peak. However, no signal could be recovered with these excitation wavelengths from the CAM. This could be due to several factors and will be explored in the discussion section.

3.3 Multi-wavelength adipose tissue imaging

To explore and characterize the lipid visualization capabilities of the system, freshly resected murine adipose tissue was imaged using 1210 nm excitation along with 250 nm excitation to highlight DNA contrast (Fig. 4). These captures were taken using 900 nm lateral step sizes. The 250 nm acquisition recovered sparse point structures of similar scale to those expected from adipose tissues consisting of large adipocytes in the 30 µm to 60 µm, each with their respectively smaller (< 10 µm) nuclei [21]. Likewise, the 1210 nm acquisition recovered larger amorphous structures in the expected scale and generals shape of the adipocyte fat reservoir. The field of view of these adipose acquisitions are substantially larger than those of the CAM, with each frame requiring roughly 52 minutes to acquire.

 

Fig. 4. HS-PARS image of freshly resected murine adipose tissues. (a,b) Shows individual 1.6 mm x 1.6 mm frames using 250 nm excitation to target DNA contrast and 1210 nm excitation to target lipid contrast. (c) Shows an overlay of these two datasets. Scale bar: 500 µm.

Download Full Size | PDF

Quantitative metrics from the CAM and adipose tissue acquisitions are summarized in Table 1. Resolution measurements were taken from edge-spread functions of recovered structures within the images. As such, they may be over reporting values due to a lack of ideal edges in the images. Calculated Rayleigh criterion $r = 0.61\lambda /NA$ for wavelength $\lambda $ is included in Table 1. These show that the recovered resolutions are larger than expected which may be attributed to poor beam quality incident on the main objective, or the over estimation of resolution due to the measurement method. Calculated ideal axial resolutions (Abbe’s diffraction formula $d = 2\lambda /N{A^2}$) for each wavelength are also included in Table 1. However, both the lateral and axial resolutions of the system were sufficient to adequately recover desired biological structures. Signal-to-noise (SNR) metrics were acquired by taking the ratio of either the maximum signal found in a given region or the mean signal found in a given region over the standard deviation of the noise taken from another region not exhibiting PARS signals. Regions were manually selected where substantial PARS signals were observed and compared against regions where minimal/no PARS signals were observed.

Table 1. Quantitative imaging metrics

View Table

4. Discussion

When comparing the performance of this hyperspectral PARS device against previously reported fixed-wavelength techniques such as those in [11] it is clear that there has been some tradeoff in system performance to achieve the highly tunable excitation. The resolution of the system has been reduced quite substantially (Table 1). This would partly be attributed to a reduction in numerical aperture but may be primarily attributed to reduction in beam quality. Given that the system alignment on the sample along with the respective alignment of each of the spatial filters now must accommodate such a wide range of wavelengths, aligning for consistent performance came at the expense of imaging resolution. This appeared to be particularly true in the UV range which was met with substantial optical dispersion variation. Based on our input beam diameter of ∼6.0 mm and equivalent numerical aperture of ∼0.2, we would expect to see an aperture-diffraction-limited resolution near 400 nm for 250 nm excitation. The measured resolution of 2.4 µm was substantially above this value. This may prove problematic when attempting to perform a hyperspectral imaging sweep through this range where high lateral resolution (<500 nm) is desirable to elicit clear cellular and sub-cellular morphology. However, the reported resolution still afforded meaningful structural recovery providing bulk morphology for each individual nucleus, along with other diagnostically relevant information such as intra-nuclear spacing and nuclear density. Likewise, resolution performance in the infrared range was relatively poor. With an input beam diameter of ∼8.2 mm and equivalent numerical aperture of ∼0.27, we would expect to see an aperture-diffraction-limited resolution near 1.4 µm for 1210 nm excitation. However, the measured resolution of 8.4 µm was quite a bit larger than this value. This may be attributed to several factors. The resolution values were pulled as edge-spread functions from the images themselves. This may suggest that the fat reservoirs on the adipocytes do not represent ideal edges for this purpose. As well, the observed beam quality in this range was particularly poor as compared to the other wavelength arms. Future efforts on similar devices may require stricter spatial filters to provide additional beam quality conditioning. Resolutions in the visible regimes were initially higher on earlier versions of the apparatus (<∼1 µm), however, when trying to visualize the capillary beds such high-resolution recovered only a sparse collection of erythrocyte-scale (∼6 µm) structures. This was likely attributed to the true sparse nature of these cells within the capillaries where they would be oriented in a single file separated by plasma. As such, visualizations of more connected capillary structures required reduction of lateral resolution which provided more consistent and connected structures. This reduction in lateral resolution was accomplished by narrowing the visible wavelength beams relative to the UV beams before entering the common objective.

For all tissue experiments ultraviolet and visible focal exposure levels were maintained below around 500 mJ/cm2, a metric commonly used in the OR-PAM community for 532 nm exposure [22], which is still well below the reported erythrocyte damage threshold of 22 J/cm² at 840 nm [23]. Infrared skin exposure levels (from 1050 nm to 1400 nm) permit five times the energy deposition, and this aspect was considered when selecting 1210 nm excitation pulse energy for targeting lipid absorption in the adipose samples which were calculated to be around 700 mJ/cm². Such relatively high fluences may come as a result of the relatively long excitation pulsewidth used of 3 ns. Given that the stress confinement condition for a ∼2 µm region will be on the order of ∼1.5 ns, such a long excitation event likely resulted in inefficient thermo-elastic generation. Future efforts should therefore aim for shorter pulsewidth tunable sources. The 970 nm detection source was maintained at around 5 mW for all acquisitions. The beam size incident on the main objective was around 7 mm diameter for the 970 nm detection, giving an NA of roughly 0.23 and a Rayleigh criterion around 2.5 µm.

In both the CAM and adipose studies, the wavelength contrast behavior aligned well with expectations. Ultraviolet excitation appeared to recover primarily cell nuclei, although, in the case of the adipose tissue it also appeared to have captured some of the lipid contrast. Given that the majority of biological tissues provide some degree of absorption at these wavelengths, some such signal bleed-through is inevitable but may require special attention when conducting ultraviolet-excitation photoacoustic microscopy on freshly excised adipose samples. Visible excitation (420 nm, 532 nm, and 560 nm) was assumed to primarily recover hemoglobin as it would be expected as the primary chromophore in this region. Comparing with previous reports of PARS and OR-PAM [10,24] we observed similar structures of tightly interconnected capillaries. Oxygenation measurements also match closely to reported values appearing between the near ∼100% sO₂ in arteries and ∼70% sO₂ in veins previously observed [25]. There indeed may be minor biological changes occurring on acquisition timescales. However, as the specimen was alive during the entire procedure and continued to live there after, it was assumed that substantial variations in blood oxygen saturation were not likely to have occurred. Moreover, extracted values provided reasonable biological blood oxygen levels matching previously reported values in [25]. Variations in oxygenation were recovered across the frame which is consistent with previously reported oxygenation measurements in CAM capillaries [25]. Some of the observed variation may have arose as errors introduced due to observed small lateral sample motion between frames.

In the CAM model we did not expect significant lipid contrast recovery as a result of infrared excitation as this tissue does not feature adipocytes, and the lipid bilayers of cells within this region may not constitute sufficient contrast for the current sensitivity and resolution of the system. Water absorption may have also been expected for this target. However, tissue-glass interfaces and tissue-tissue interfaces may not provide sufficient scattering contrast to amplify water signals above the noise floor of the system. Standalone water measurements utilized an air-water interface to amplify PARS signals. Meanwhile, tissue-tissue interfaces may provide roughly two orders of magnitude or less scattering contrast as compared to an air-water interface.

Acquisition of lipid contrast in the adipose tissue required substantially higher pulse energies as compared to targeting hemoglobin in the live subjects. This is likely attributed to the significant reduction in absorption coefficient. At the local absorption peak of 1210 nm, lipid is expected to provide around two orders of magnitude less absorption as compared to hemoglobin at 532 nm. This, when combined with the larger focal spot provided by the infrared wavelengths as compared to their visible counterparts, demanded substantial pulse energies nearing single micro-Joule levels. As well, this contrast appeared to generally give more vague structure as compared to the well-defined cell nuclei and micro vessel structures seen from the CAM models. Despite these challenges. this represents the first such report of a non-contact photoacoustic modality which can recover lipid contrast from freshly resected tissues. Adipose samples were explored using visible excitation wavelengths in addition to the ultraviolet and infrared, however, little contrast was recovered. This is likely attributed to extremely low concentrations of hemeproteins (cytochromes, hemoglobin) in these regions. As such, these results were not included.

Acquisition times with the current architecture are quite long, however, we are moving forward with fixed-wavelength MHz-range excitation lasers which aim to drastically reduce these frame times by around 3 orders of magnitude. For example, the adipose frames which currently take 52 minutes using the current architecture may be acquired in a matter of seconds using optical scanning techniques.

5. Conclusion

Here we have presented recent work surrounding an improved hyperspectral PARS microscope. This demonstrates the first report of a non-contact photoacoustic microscope that is capable of recovering DNA, hemeprotein, and lipid contrast from a single device. Multiplex acquisition efficacy was demonstrated by recovering functional sO₂ measurements from a live subject. Meanwhile, sensitivity of the device was such that it was able to characterize infrared absorption of water. The HS-PARS microscope is poised to represent an essential material investigation tool capable of non-destructive recovery of a substantial range of endogenous contrast. Work will continue towards further improving system sensitivity and consistency while maintaining, or yet further extending, its wide excitation wavelength range.