1. Introduction

Raman spectroscopy (RS) is based on the Raman effect, which occurs when the frequency change of the incident light due to scattering exactly equals the frequency of molecular vibrations, thus providing molecular fingerprint information in a label-free manner [1–3]. The advantages of high chemical specificity, no special sample preparation, and non-invasive detection have made this technology invaluable in biology, medicine, and pharmacy over the past decades [4–6]. Especially in the pharmaceutical field, RS has emerged as a vital analysis tool as it can provide the detailed chemical composition and information about multi-dimensional molecular interactions [7–12]. For instance, RS technology has been widely used to study the distribution of active pharmaceutical ingredients, particle sizes of drugs, drug crystal forms, pharmacological screening, synthetic drugs, drug authenticity, etc. [13–17]. However, the current drug detection and analysis technology based on RS is largely applied for in vitro detection. This is owing to the fact that the conventional Raman spectrometer is generally based on the Gaussian beam to interact with the samples to stimulate the Raman signal. The Gaussian beam is easily interfered by scatters mixed in the investigated chemicals, thus, resulting in poor spectral quality and even inability to identify the useful components from the measured spectrum. On the other hand, the Gaussian beam generally only stimulates Raman signals in a small volume. When the concentration of the substance to be detected is low, there is a risk of missed detection, resulting in a poor quantitative result. Spatially offset Raman spectroscopy (SORS) enables the detection of chemical signals at different depths by setting a certain distance between the excitation and detection points. Due to the use of such a heterodyne detection, the SORS technique can detect the chemicals concealed in the opaque or sub-transparent media, making it well suited for in vivo detection [18]. However, it has the problem that it requires a certain thickness of the detected objects and it is difficult for us to know exactly where the Raman signal is coming from. When testing the samples with packaging or coatings, intense fluorescence or Raman scattering from the surface layer will annihilate the Raman spectrum of deeper sample. Furthermore, it has low accuracy for quantitative analysis of unknown targets. Transmission Raman spectroscopy (TRS) technique is separated from SORS, where the laser beam and the Raman collection end are separated to the extreme, i.e. both are located on opposite sides of the sample [19]. This technique is also suited for analyzing the component of turbid samples. However, it faces the same problem as the SORS and is totally unsuitable for in vivo detection.

Tissue scattering is a significant challenge for photonics-based sensing and imaging techniques. Many efforts have been devoted to this topic, including building mathematical models to calculate light propagation and distribution in tissues, using adaptive optics, wave manipulation, nonlinear or long-wavelengths excitation techniques, as well as in combination with other imaging methods [20–26]. The propagation and distribution of light in tissues can be accurately modelled by using the radiative transfer equation or Monte Carlo method [20,21]. These models are usually employed to characterize the features of the diffuse light and are not exactly suitable for the microscopic applications. Adaptive optics technology corrects the dynamic wavefront error resulting due to the tissue scattering in real time, thus, making the optical system adapt to the changes associated with the tissue scattering and maintain the best working state at all times [27]. Wave manipulation techniques, e.g., the time-reversed ultrasonically encoded light or the photoacoustically guided shaping technique [28–30], can stimulate the signal targets in deep scattering tissues, thus, finding successful application in fluorescence microscopy, photoacoustic microscopy and tomographic imaging [21–24]. Both adaptive optics and wave manipulation successfully extend the detection depth, however, these techniques are cumbersome and require complex instruments, which is not feasible in the case of in vivo Raman detection. Nonlinear excitation and combination of other imaging methods are also not suitable for the spontaneous Raman spectroscopy. Further, the detection depth can be extended by using long wavelength excitation, nonetheless, it is still affected by the tissue scattering, and its penetration ability is limited [25]. Recently, Bessel beam, a non-diffractive and self-reconstructing beam, has been successfully applied for microscopic imaging, so as to provide an extended depth of field and overcome the impact of tissue scattering. Examples of its successful applications include: the modalities of the light sheet microscopy [9,31,32], optical coherent tomography [33], photoacoustic microscopy [34], two-photon microscopy [35], coherent Raman scattering microscopy [36–38], etc. Due to the optimal self-reconstructing characteristics, it can also be expected to achieve in vivo sensing and deep imaging [39,40].

Conventionally, Gaussian beam is equipped in the Raman spectroscope as the excitation light, e.g., in the confocal Raman spectroscopy and portable Raman spectroscopy. Due to the high energy convergence at the focus, such RS system can provide good signal-to-noise ratio for the thin or transparent samples. However, the Gaussian beam is rapidly diffused as it hits the scatters, as shown in Supplementary Fig. S1(a), thus, leading to the large attenuation of light energy per unit volume at the chemical site, resulting in the poor spectral quality and bad quantitative analysis. In this study, we demonstrated the use of Bessel beam Raman spectroscopy (BRS) for chemical sensing of the targets in scattering media and biological tissue slices. The BRS method is based on the self-reconstructing characteristics of the Bessel beam when encountering scatters (Supplementary Fig. S1(b)), thus, overcoming the influence of scatters to a certain extent and providing a long sensing depth. Compared to the Gaussian beam Raman spectroscopy (GRS), BRS is able to provide Raman spectra with higher quality in the scattering media as well as better quantitative capabilities, which indicated significant potential for in vivo Raman detection of drugs. We first theoretically calculated the Bessel beam and compared it with the experimentally generated one to attain agreement. The BRS system was subsequently homebuilt and its feasibility was validated by using dimethyl sulfoxide (DMSO) and acetaminophen as test sample. Further, by using acetaminophen embedded in the handmade scattering samples, the potential of the BRS method for in vivo chemical sensing and quantitative detection of the drugs in the scattering media were exemplified. Finally, we explored the potential of BRS for in vivo chemical sensing of acetaminophen under biological tissue slices, demonstrating significant development towards the evaluation of drug in vivo.

2. Materials and methods

2.1 Bessel beam Raman spectroscope

A schematic of the homebuilt Bessel beam spectroscope has been depicted in Fig. 1. The excitation beam generated from a semiconductor continuous wave laser (lambda 532-200 DFSS, RGB Photonics) was collimated and subsequently expanded by using a 4F system. L1 (f = 50 mm) and L2 (f = 50 mm) were used to collimate the excitation laser. L3 (f = 30 mm) and L4 (f = 150 mm) were used to expand the excitation light by five times. Afterwards, the expanded beam having 5 mm in diameter was delivered to the axicon (AX252-B, Thorlabs) to convert the incident Gaussian beam into a Bessel beam. The axicon has a wedge angle of $\gamma = {2^o}$. The angle between the wavefront of incident beam after passing through the axicon and propagation direction was 0.9°, whereas the refractive index of the axicon was 1.4497. The full width at half maximum of the non-diffractive distance of the experimentally generated Bessel beam can be 106.80 mm. The generated Bessel beam was used to directly irradiate on the sample and produce the Raman scattering signals. A fiber-optic probe (ATR20105, OPTOSKY) was used to collect the Raman scattering signals at an angle less than 90 degrees from the excitation beam path, followed by its transmission to a spectrometer (ATP5020, OPTOSKY) to generate the Raman spectrum for analysis. The detection of the fiber-optic probe is focused on the same point as the illumination of the Bessel beam. To facilitate the comparison between the BRS and GRS schemes, a lens (AC254-100-B-ML, Thorlabs) with a focal length of 100 mm was placed in front of the axicon at a distance of 30 mm. The lens and axicon were affixed on the movable brackets, by which the arm between the BRS and GRS schemes could be swiftly switched. This design ensured that the excitation light under the two schemes could irradiate the same position of the sample to the utmost extent, so as to eliminate the interference of the external factors. The detailed description of the Bessel beam Raman spectroscope was presented in Supplementary Note 1, and the theory of the Bessel beam generation as well as its quality assessment were detailed in Supplementary Note 2, Note 3, and Supplementary Fig. S2.

 

Fig. 1. The schematic of the homebuilt Bessel beam Raman spectroscope (M: mirror; L: lens; A: axicon; and S: sample).

Download Full Size | PDF

2.2 Sample preparation

Dimethyl sulfoxide (DMSO) was purchased from Tianjin Tianli Chemical Reagent Co. Ltd.. Acetaminophen tablets were purchased at a local pharmacy, which was manufactured by Tonghua Wantong Pharmaceutical Co. Ltd..

In experiments involving DMSO, an appropriate amount of DMSO was pipetted in a quartz cuvette and then fixed on a sample holder to detect the Raman signal. Acetaminophen was finely ground into a powder and packed in the plastic bags for Raman signal detection.

The scattering medium was generated as follows [37,41]. 0.4 g agarose was weighed in a beaker, followed by the addition of 40 mL of deionized water. The mixture was stirred uniformly, and then heated in a microwave oven until an extremely transparent solution was obtained. Afterwards, 250 µL of polystyrene beads (2 µm or 10 µm in diameter) were added to the beaker. After thorough shaking, the mixture was quickly poured into a flat mold. The mixture was cooled and allowed to solidify, followed by cutting into layers of the same size and thickness for subsequent experiments.

Different concentrations of drugs were made by mixing different amounts of acetaminophen with starch. The total amount of the mixture was fixed, and the proportion of acetaminophen to starch was varied to achieve different concentrations. Each sample was stirred with a spoon to achieve as homogeneous as possible. The drug concentrations used in the experiments in the absence of a scattering medium were 95%, 90%, 85%, 70%, 40%, and 20%. These used in the experiments in the presence of a scattering medium were 90%, 70%, 60%, 50%, 40%, 30%, and 20%. These samples of all concentrations were evenly divided into three groups and detected in batches for experimental reproducibility.

The animal experiment was performed according to the Ethical Guidelines on Animal Care with approval from the animal welfare committee of the Fifth Affiliated Hospital of Sun Yat-sen University (00020). A Kunming mouse at ∼8 weeks old was dissected, whose inner skin, ear, and liver were taken out. The thickness of the ear was about 0.44 mm, and the inner skin was about 1.3 mm. The liver was cut into a slice of about 1 mm thick. Besides, the lamb rolls slice at 2 mm was purchased from the local supermarket.

2.3 Spectral processing and analysis

All spectra were acquired at three different positions of the sample, and then averaged for subsequent analysis. Prior to analysis, the Raman spectra were preprocessed with denoising and baseline correction [42,43].

2.4 Evaluation indicators

In order to quantitatively evaluate the accuracy of the Raman spectra measured by the BRS system, the evaluation indicators including the recognition rate of Raman peaks (RRRP) and localization error of Raman peaks (LERP) were used. RRRP is defined as the ratio of the number of Raman peaks identified by the measurements to the total number of Raman peaks provided by the reference. LERP indicates the position deviation of the Raman peak, i.e., the position of the Raman peak obtained by the BRS system compared to that provided by the reference. Since the spectral resolution of the spectrometer is roughly 9 cm^-1, the Raman characteristic peak was considered to be misidentified in cases where the LERP was greater than 9 cm^-1.

We also calculated the spectral similarity (SSIM) between the Raman spectra measured by the BRS and GRS systems. The definition can be described as [44]:

3. Experiments and results

3.1 Feasibility validation of the Bessel beam Raman spectroscope

The feasibility of the homebuilt BRS system was verified by selecting the GRS system as a comparison on detecting the DMSO and acetaminophen samples. Figures 2(a) and 2(b) present the Raman spectra of the DMSO and acetaminophen samples, respectively, acquired by the homebuilt BRS and the conventional GRS systems. The Raman spectra of both samples measured with the BRS system were noted to be completely consistent with those obtained by the GRS system. The SSIM between them was as high as 0.9940 and 0.9979 for DMSO and acetaminophen, respectively. In the spectral range of 200 to 3000 cm^-1, the GRS system identified nine Raman peaks of DMSO (313 cm^-1, 335 cm^-1, 382 cm^-1, 670 cm^-1, 699 cm^-1, 1048 cm^-1, 1424 cm^-1, 2908 cm^-1, 2989 cm^-1) and twenty-one Raman peaks of acetaminophen (328 cm^-1, 393 cm^-1, 466 cm^-1, 509 cm^-1, 630 cm^-1, 653 cm^-1, 711 cm^-1, 799 cm^-1, 836 cm^-1, 861 cm^-1, 969 cm^-1, 1170 cm^-1, 1239 cm^-1, 1280 cm^-1, 1327 cm^-1, 1447 cm^-1, 1518 cm^-1, 1566 cm^-1, 1617 cm^-1, 1654 cm^-1, 2925 cm^-1), respectively. Fortunately, all of these peaks were also identified by the BRS system. To quantitatively assess the accuracy of identification, the standard spectra obtained from a public database (Spectral Database for Organic Compound SDBS and RRUFF) were selected as references (as plotted in Supplementary Fig. S3), and the evaluation indicators of RRRP and LERP were calculated between the BRS (or GRS) system and the reference spectrum. Figures 2(c) and 2(d) show the number of Raman characteristic peaks of DMSO and acetaminophen, and the associated LERP value at each peak position. The RRRP and mean LERP of the two samples have also been plotted in Figs. 2(e) and 2(f). Because the BRS and GRS systems identified all the Raman peaks provided by the reference spectrum, the RRRP of the BRS and GRS systems both reached 100% for sensing the DMSO sample (Fig. 2(e)). However, for the acetaminophen, there are three Raman peaks (1250 cm^-1, 1609 cm^-1, 1612 cm^-1) cannot be identified by both the BRS and GRS systems. Of all the identified Raman peaks, the largest LERP reached 11 cm^-1 for BRS system (653 cm^-1), slightly higher than the system spectral resolution of 9 cm^-1 (magenta line in Figs. 2(c) and 2 (d)). The bad thing is that two of the Raman peaks identified by the GRS system have LERP values greater than 9 cm^-1 (664 cm^-1, and 2934 cm^-1). Thus, the RRRP value of the BRS system for acetaminophen sample was 83.33%, and the GRS system was 79.17%, as shown in Fig. 2(e), demonstrating the optimal feasibility and superior performance of the BRS system in detecting the chemicals. For the sample of DMSO, the maximum and mean LERP values observed by the BRS system were 6 cm^-1 and 3 ± 2.06 cm^-1, respectively; whereas, these values provided by the GRS system were 8 cm^-1 (close to the system spectral resolution) and 3.11 ± 2.32 cm^-1, as shown in Figs. 2(c) and 2(f). For the acetaminophen in Figs. 2(d) and 2(f), the maximum LERP value in all the identified Raman peaks was 11 cm^-1 for both the BRS and GRS system, greater than the system spectral resolution. However, the LERP at only one Raman peak was higher than the system spectral resolution for the BRS system, while the GRS system had two (664 cm^-1 and 2934 cm^-1). This also results in a mean LERP of 2.24 ± 2.83 cm^-1 for the BRS system, which is lower than that for the GRS system (2.33 ± 2.90 cm^-1). These results confirmed the validity of the developed BRS system for chemical sensing of samples.

 

Fig. 2. Feasibility validation of the Bessel beam Raman spectroscope. (a) and (b) Raman spectra of DMSO and acetaminophen respectively; (c) and (d) the number of Raman characteristic peaks of DMSO and acetaminophen samples respectively, and the associated LERP value at each peak position; here, magenta lines indicate the system spectral resolution; (e) and (f) the recognition rate of the Raman peaks (RRRP) and the mean localization error of the Raman peaks (LERP) for DMSO and acetaminophen samples. In these figures, the red symbols (solid lines, solid dot lines, bars) indicate the results of the BRS system, whereas the blue symbols (solid lines, solid dot lines, bars) represent those of the GRS system.

Download Full Size | PDF

We also investigated the direct comparison between the BRS and GRS systems by selecting the GRS system as a reference. The detailed results are presented in Supplementary Note 4 and Supplementary Fig. S4. For both DMSO and acetaminophen, highly consistent measurements were obtained between the BRS and GRS systems (Supplementary Figs. S4(a)-4(d)). To further quantify the differences between the Raman spectra measured by the BRS and GRS systems, we calculated the ratio of Raman signal intensities between them at each Raman characteristic peak (I_BRS/I_GRS), as plotted in Supplementary Figs. S4(e) and S4(f). The closer this ratio is to approximately 1, the better the fit between the BRS and GRS systems, indicating that the performance of the BRS system is closer to that of the GRS. In Supplementary Fig. S4(e), most of the ratios are close to 1 (0.9893 ± 0.0658), representing the BRS system can well detect the Raman signals of DMSO sample. This ratio was off by a bit from 1 when we focused on the Raman peaks at 335 cm^-1 (0.9182) and 699 cm^-1 (0.9118). This can be caused by the deviation of the detector response. The results confirmed the feasibility and good performance of the BRS system in detecting pure samples. Supplementary Fig. S4(f) depicts the I_BRS/I_GRS ratio for the acetaminophen sample, which was more concentrated around 1 (0.9623 ± 0.0779). Owing to the limited spectral resolution of the spectrometer and the related noise interference, the ratios are a little bit off from 1 (±0.15) at Raman peaks of 466 cm^-1, 630 cm^-1, and 1518 cm^-1. There results collectively demonstrated the high accuracy and effectiveness of the developed BRS system, providing Raman spectra that are almost identical to those of the standard GRS system.

3.2 Chemical sensing of drugs in the scattering media

To prove the capability of the BRS system to chemically sense drugs in scattering media, the Raman spectra of acetaminophen were detected at different thicknesses of the scattering samples, and then compared with those of the GRS system. Figures 3(a) and 3(b) show the detected Raman spectra by the BRS and GRS systems respectively. A few valuable observations could be drawn. Firstly, the thicker the scattering medium was, the greater the impact on the Raman signal. On increasing the thickness of the scattering medium, the intensity of the Raman signal decreased, which was consistent in both BRS and GRS schemes. Secondly, on enhancing the scattering medium thickness, the trend of reduction in the Raman signal intensity was different. The Raman signal intensity detected by the BRS system attenuated approximately linearly (Fig. 3(a)), whereas the intensity measured by the GRS scheme was observed to decrease sharply as the scattering medium thickness was increased to 2 mm (Fig. 3(b)). It also confirmed that the scattering had a significant influence on the Gaussian beam. As the thickness of scattering medium was increased to 4 mm, it became difficult for the GRS scheme to recognize the Raman peaks of acetaminophen (Fig. 3(b)). This is because the scattering medium destroyed the focusing status of the Gaussian beam, resulting in the Gaussian beam not having sufficient energy density to excite the Raman scattering signal. In contrast, the BRS system resolved most of the Raman peaks of acetaminophen clearly, even as the thickness of the scattering medium was increased to 5 mm.

 

Fig. 3. Chemical sensing of the drugs in the scattering media. (a) and (b) The Raman spectra of acetaminophen in the scattering media with varying thickness, collected by the BRS system and GRS scheme respectively; (c) and (d) the intercepted spectra of [760, 900] cm^-1 region from (a) and (b), which covered the Raman characteristic peaks of 800 cm^-1, and 861 cm^-1. The spectra in (c) and (d) were baseline corrected; (e) the average intensity of the Raman signals at the selected peaks as a function of the medium thickness; and (f) the attenuation rate of the Raman signal intensity at each thickness, where the black solid square lines present the results generated by the BRS system and the red solid dot lines indicate the results obtained from the GRS scheme.

Download Full Size | PDF

The spectrum from 760 cm^-1 to 900 cm^-1 was further intercepted for quantitative analysis, which covered the Raman characteristic peaks at 800 cm^-1 and 861 cm^-1, as shown in Figs. 3(c) and 3(d). The earlier observations could be observed more clearly and intuitively in these figures, which underlined the superiority of the developed BRS system as compared with the GRS scheme. The signal intensities averaged over all the Raman characteristic peaks in the selected region were calculated, and then plotted as a function of thickness in Fig. 3(e). The attenuation rate of the Raman signal intensity at each thickness was plotted in Fig. 3(f). It was calculated as follows. The difference between the previously calculated signal intensities in the presence and absence of the scattering medium was first calculated, and then the ratios of these difference values to the signal intensities in the absence of the scattering medium were further calculated. From these quantitative analyses, the following observations were made. First of all, the Raman signal intensity attenuated as the thickness of the scattering medium increased, where the attenuation of the GRS scheme was much higher than that of the BRS system. For the medium with thickness smaller than 2 mm, the Raman signal intensity obtained by the GRS scheme was observed to be stronger than that of the BRS system. This was due to the fact that the focusing energy of the Gaussian beam was much higher than the energy at the Bessel beam central lobe, which was responsible for the excitation of Raman signals. However, due to the self-reconstructing property of the Bessel beam and susceptibility of the Gaussian beam to scattering, the Raman signal intensity of BRS became higher than that of GRS when the thickness exceeds 2 mm. Secondly, the Raman signal intensity is linearly attenuated over a specific thickness range, while the BRS provided a larger region of linear attenuation of the signal, as presented in Fig. 3(e). On the other hand, the GRS scheme provides a linear attenuation region of the signal within 2 mm thickness, and for the thickness larger than 2 mm, the signal strength and attenuation rate tended to stabilize. However, the BRS system provided a linear attenuation region of up to 4 mm. Thirdly, the intensity attenuation rates of Raman signals based on the Gaussian beam were noted to be always larger than those in the case of the Bessel beam (Fig. 3(f)). The experimental details were described in Supplementary Note 5.

In order to avoid the influence of other unwanted factors and to ensure the reproducibility of the analysis results, the Raman peaks at 1239 cm^-1, 1280 cm^-1 and 1326 cm^-1 were also selected for analysis, as presented in Supplementary Fig. S5. The same conclusions, as observed earlier, were drawn. In addition, the size of the polystyrene microspheres used to produce the scattering medium was also changed from 2 µm to 10 µm, followed by an experimental analysis similar to the one mentioned earlier. The detailed results are presented in Supplementary Fig. S6, which revealed the similar results and arrived at the same conclusions.

Collectively, the findings from these experiments validated that the BRS system exhibited superior capability in detecting the chemicals in scattering media, thus, exhibiting a potential of in vivo drug detection.

3.3 Quantification of different concentration of drugs

The capability of the BRS system to quantify different concentrations of drugs in scattering media was investigated with the aim of demonstrating its superiority over the conventional GRS system in terms of quantification. The Raman spectra of different concentrations of acetaminophen were detected, and the characteristic spectrum from 760 cm^-1 to 900 cm^-1 was intercepted for quantitative analysis, as shown in Fig. 4. Figures 4(a) and 4(b) show the Raman spectra of different concentrations of acetaminophen detected by the BRS and GRS system in the absence of scattering media, respectively. We found that the Raman signal intensities of acetaminophen detected by the GRS system were higher than those of the BRS system at the same concentration. However, the Raman signal intensity detected by the GRS system did not show a linear decay with decreasing acetaminophen concentration. For example, the Raman signal intensity detected at 90% concentration was significantly higher than that detected at 95% concentration. This may be due to the relatively small focusing volume of the Gaussian beam, and in Raman spectroscopy acquisition, this relatively small focusing volume does not necessarily act exactly on the drug due to the presence of drug excipients. This can also be seen from Fig. 4(c) that plots the Raman signal intensity at 856 cm^-1 as a function of drug concentration (red line). Here, the mean and variance of the Raman signal intensity from multiple measurements were calculated. We found that firstly, the linearity between the Raman signal intensity detected by the GRS system and the drug concentration was relatively poor, with an R² value of 0.8; and then the fluctuations between experimental results of multiple measurements near the same detection site were also relatively large. These problems are well avoided by the BRS system. Firstly, the signal intensity of the Raman spectra detected by the BRS system decays linearly with the decrease of drug concentration, where the R² value of the fitted curve is 0.98 (the black line in Fig. 4(c)). Secondly, the fluctuation and the variability of the Raman spectral signals detected for the same sample at different detection points are low. These results validated the good stability and quantitative capability of our BRS method in detecting non-pure drugs, even in the absence of scattering media. The experimental details were presented in Supplementary Note 5.

 

Fig. 4. Quantitative sensing of different concentrations of drugs. (a)-(b) Raman spectra of different concentrations of acetaminophen detected by the BRS system (a) and the GRS system (b) in the absence of a scattering medium respectively; (c) The corresponding Raman signal intensity as a function of the concentration of acetaminophen, where the red dot line indicates the results of the GRS system, and the black square line presents those of the BRS system. (d)-(e) Raman spectra of different concentrations of acetaminophen detected by the BRS system (a) and the GRS system (b) in the presence of a scattering medium respectively; (f) The relevant Raman signal intensity as a function of the concentration of acetaminophen, where the red dot line indicates the results of the GRS system, and the black square line presents those of the BRS system.

Download Full Size | PDF

Next, we further examined the Raman spectra of different concentrations of acetaminophen buried under 1.8 mm scattering samples. Figures 4(d) and 4(e) show the detected Raman spectra by the BRS and GRS systems, respectively. Unlike the case without a scattering medium, the signal intensity of the Raman spectra detected by the BRS system was higher than that detected by the GRS system in the presence of a scattering medium. This difference in signal intensity was particularly pronounced with decreasing drug concentration. For example, when the concentration of acetaminophen was reduced to 20%, the BRS system detected Raman signals with intensity values up to 1000, as shown in Fig. 4(d); while the GRS system detected signals with an intensity of only 150, as shown in Fig. 4(e). This also indicated that the Bessel beam is much less affected by scattering. Furthermore, in the presence of scattering medium, there was a better linear relationship between the Raman signal intensity and the drug concentration detected by the BRS system compared to the GRS one. As shown in Fig. 4(f), the R² value for BRS is 0.98, while the corresponding value for GRS is only 0.59. We found that the linearity between the Raman signal intensity detected by the BRS system and the drug concentration was well maintained regardless of the presence of the scattering medium; however, for the GRS system, the linearity deteriorates dramatically (0.80 to 0.59) due to the introduction of the scattering medium. Moreover, the fluctuation of experimental measurements at different detection points detected by the GRS system remained large. Although the experimental measurements of a given drug concentration detected by the BRS system fluctuated widely from one detection point to another due to the influence of the scattering medium, they were generally stable and less volatile than those of the GRS system.

As mentioned above, this relatively large fluctuation should be caused by the inability of a Gaussian beam with a relatively small focusing volume to act accurately on the drug; a problem that is perfectly solved by the Bessel beam with a long focusing distance and self-reconstructing properties. To ensure the reliability of the experimental results, we conducted reproducibility experiments. Samples of all concentrations were evenly divided into three groups, and each group was measured three times from different detection points. The average of multiple different measurements was calculated as the final results. These repeated experiments yielded highly similar conclusions. The other two experimental results are shown in Supplementary Figures S7 and S8.

In summary, these experimental results collectively demonstrated the higher stability and superior quantitative capability of our BRS method over the conventional GRS one for non-pure drug detection. Such an advantage is especially evident and outstanding in the presence of scattering media. Penetration depth, stability, and quantitative capability are the key elements for realization of in vivo analysis based on RS. The outstanding advantage of our BRS method could achieve these keys to a certain degree, thus exemplified the potential for the quantitative detection of drugs in vivo.

3.4 Applications of drug sensing in biological tissue slices

As an example to illustrate the potential of the proposed BRS system in biological applications, we detected the Raman spectra of acetaminophen buried in the biological tissue slices of different thicknesses. The biological tissue slices include mouse ear, mouse inner skin, mouse liver, and lamb rolls. In order to ensure the accuracy and stability, we measured Raman spectra of acetaminophen at three points on the sample, and then calculated their average values. These Raman spectra in the range of 760 cm^-1 to 1400 cm^-1 were further selected for analysis, as shown in Fig. 5. Some interesting phenomena can be observed. Firstly, the Raman spectra behaved differently in various biological tissue slices, especially between 900 cm^-1 to 1100 cm^-1, where the variation was most pronounced. This may be due to the response of different biological tissues to laser excitation. Secondly, the attenuation of the Raman signal intensity was greatest for the mouse liver slice (1 mm in thickness), even though it was not the thickest tissue slices. This is because the liver is the largest absorber of light among these tissue slices. Figures 5(a) and 5(b) present the measured Raman spectra of acetaminophen buried under 0.44 mm mouse ear slice and 1 mm mouse liver slice, respectively. We found that our BRS system could detect distinct Raman characteristic peaks of acetaminophen at 802 cm^-1, 843 cm^-1, 865 cm^-1, 1175 cm^-1, 1244 cm^-1, and 1332 cm^-1 in both case, and these were also consistent with the Raman spectra of pure acetaminophen. Moreover, our BRS system can also detect the Raman characteristic peak of acetaminophen at 971 cm^-1 from under the mouse liver slice. Furthermore, we selected mouse inner skin and lamb rolls to make tissue slices and used them as scattering media for Raman spectroscopy of acetaminophen. The thickness of the mouse inner skin and lamb rolls slices were 1.3 mm and 2 mm respectively. The relevant Raman spectra were shown in Fig. 5(c) and 5(d). The results showed that the Raman peaks of acetaminophen at 802 cm^-1, 843 cm^-1, 865 cm^-1, 971 cm^-1,1175 cm^-1, 1244 cm^-1, and 1332 cm^-1 can be observed in both cases. We found that the Raman peaks measured by the BRS system differed slightly in shape and intensity when different biological tissue slices were used as scattering media, but the positions of these Raman peaks were essentially fixed. The differences between peaks (Raman spectra) can be induced by the interference of stray signals from biological tissues and different absorption of light by different tissue slices. These biological experiments demonstrated that the BRS system provided promising performance in detecting the chemicals under the biological tissue slices, thus, exhibiting a high potential of in vivo drug detection.

 

Fig. 5. Raman spectra of acetaminophen buried in the different biological tissue slices, including mouse ear slice of 0.44 mm (a), mouse liver slice of 1 mm (b), mouse inner skin slice of 1.3 mm (c), and lamb rolls slice of 2 mm (d).

Download Full Size | PDF

4. Discussion

The demonstration of the proof of concept of the Bessel beam Raman spectroscopy (BRS) for chemical sensing of drugs in handmade scattering samples and biological tissue slices has been presented. The BRS scheme has been confirmed to be capable of accurately detecting the Raman spectra of chemicals buried in the scattering media, especially exhibiting superiority in quantitative analysis. Based on the concept, the BRS system was first built, followed by the demonstration of the feasibility and accuracy by measuring Raman spectra of DMSO and acetaminophen. Compared with the spectra acquired by the Gaussian beam Raman spectroscope, the BRS system exhibited an optimal ability in both Raman spectrum detection and Raman peak recognition. By carrying out the experiments on the detection of buried acetaminophen in scattering media, the potential of the developed technology in detecting and quantitating the chemicals in the scattering media were exemplified. Finally, the potential of BRS for chemical sensing of acetaminophen in biological tissue slices was also explored, indicating significant development towards to the evaluation of drugs in vivo. However, there is plenty of room for further improvement of the developed BRS technology.

First, the BRS system requires a long acquisition time to obtain high-quality Raman spectra, owing to the low detection sensitivity of the system. In this study, a portable spectrometer was employed to separate and collect the Raman spectra. The photodetector used was of industrial grade, with poor detection capability in low light. In this regard, the use of highly sensitive photodetectors can further improve the signal-to-noise ratio as well as reduce the signal acquisition time. In addition, high-performance spectrometers can also enhance the detection sensitivity of the system by optimizing the associated optics, while improving the spectral resolution of the system, thereby, collectively improving the spectral detection ability and peak recognition accuracy of the BRS system.

Second, the detection sensitivity of the BRS system can be further improved by enhancing the excitation power, which was limited by the laser system used in this study. It should be emphasized that due to the particularity of the Bessel beam, e.g., the energy of the Bessel beam is evenly distributed in its central lobe and side rings, enhancing the excitation power suitably does not cause light damage to the sample [45]. In the experiments, the total power of the Bessel beam at sample was set equal to the focused power of the Gaussian beam. However, only the Raman signal generated at Bessel beam central lobe was probed. As the power density at Bessel beam central lobe is much lower than that of the focused Gaussian beam, thus, it results in the weak Raman signal, as presented in Fig. 3. To prove this hypothesis, a set of experiments were carried out by reducing the power of the Gaussian beam, as shown in Supplementary Fig. S9. On decreasing the beam power, the intensity of the Raman signal generated by the GRS scheme decreased significantly (Supplementary Figs. S9(c) and S9(d)). In the case where the power of the Gaussian beam at the focus is equal to the power of the Bessel beam central lobe, the GRS scheme can detect only very weak Raman peaks even with a 1 mm thick scattering medium. Even with no scattering medium, the detected Raman peaks were still very weak. These provided a more direct proof of the advantages of the developed BRS system. We also theoretically calculated the focal volume of these two beams. The focal volume of the Gaussian beam was 0.0005 mm³, while that of the Bessel beam was 0.0676 mm³. Therefore, when the power the Gaussian bema at focus is equal to the power of the Bessel beam central lobe, the laser power density decreases because the focused volume of the Bessel beam was approximately 130 times larger than the Gaussian beam, so that no photodamage is caused to biological tissues. Thus, a more powerful laser with high output power would enhance the detection sensitivity, thus, in turn, improving the spectral quality and reducing the acquisition time.

Third, in the current system, the excitation and detection arms are not in the coaxial optical path, which necessitates the sample fixation during the experiment and also affects the efficient collection of the Raman signals. On the other hand, the system also takes up a lot of space, thus, losing viability for portable applications. Thus, the future study is planned with a focus on optimizing the system design, making the detection and excitation optical path coaxial, reducing the occupied space and improving the sample compatibility.

Fourth, there is no doubt that the excitation efficiency of the Raman signal is inversely proportional to the fourth power of the excitation light wavelength. Therefore, theoretically the shorter the wavelength of the excitation light is, the stronger the Raman signal is generated. In this work, aiming at verifying the feasibility and the application potential of the BRS scheme, we used a 532 nm laser to generate the Bessel beam in the homebuilt BRS system, which provided high intensity and satisfactory signals. However, the 532 nm laser is not a perfect choice for probing Raman signals from a living organism since it receives interference from strong autofluorescence of tissues. In the ongoing work, we are developing a Bessel beam Raman spectroscope based on a 785 nm laser with a view to gaining more applications in deep tissues of living organisms.

In perspective of application, with the increased development of various types of drugs, how to detect drug rapidly, simply, and efficiently are required in qualitative and quantitative analysis both in vivo and in vitro [46]. Raman spectroscopy is experiencing a surge in interest for pharmaceutical applications. However, the applications were mainly focused on detecting the minimum concentrations of drug in mixture samples or distinguish compounds with similar structure. Commonly, chemometric techniques were chosen as the assistant method to achieve the quantitative analysis of Raman spectroscopy, and many efforts were taken on it [47]. Our results have led to an alternative research approach to innovative Raman spectroscopy through excitation light source innovation, thus facilitating the potential for quantitative analysis, rather than just through algorithms. Our results confirmed that high precision and low error drug quantification in the scattering media can be achieved potentially using Bessel beam Raman spectroscopy. Future innovation in systems and algorithms should be enable more accurate in vivo Raman detection of drugs qualitatively and quantitatively.