1.

Introduction

The photoacoustic (PA) effect was first observed in the year 1880 by Alexander Graham Bell.¹ The PA effect initially was used in spectroscopy studies of materials throughout the mid-1900s, but it was not until invention of the laser that significant biomedical imaging applications using the PA technique became possible due to the improved sensitivity provided by laser excitation.^2–12 The low-power PA technique has been predominantly utilized in the field of spectroscopy,^11–18 with substantial applications in trace gas detection.^5,19–23 It exhibited significantly higher sensitivity than most other techniques,^24–27 with concentration sensitivity in the parts per billion.^3–5,25,27

PA spectroscopy is a powerful analytical tool for examining the optical absorption properties of solids as it directly measures the energy absorbed by the material on exposure to light.²⁸ Conventional optical absorption/transmission spectroscopy requires a sample with a specialized surface as the scattering would significantly affect the accuracy of the measurement of optical absorption coefficient. It cannot be used to study highly scattering samples.²⁹ The only other method of obtaining spectroscopic information from opaque samples is diffused optical reflection spectroscopy, where special handling is required for sample surface preparation. In the case of PA spectroscopy, spectral information can be obtained from a variety of samples, including opaque materials such as powders, metals, and semiconductors.^2,12

The principle and theory of the standard PA technique were well established by Rosencwaig, who described the fundamental principles and investigated several possible applications of the PA technique.³⁰ Patel and Tam reviewed the physics of the technique and also extended the technique to other applications such as liquids.^31–33

Several types of PA cells have been proposed by various investigators.^{4,19,31,34–37} PA cells made of a kilohertz microphone and piezoelectric transducers (PZT) have been widely studied.^31,34,37,38 Gas-microphone PA cells were generally used to study powder and samples with large surface to volume ratio, whereas PZT PA cells were used for liquid and bulk solids.² The gas-microphone PA cell possesses a relatively better sensitivity at low modulation frequencies.² The sample in the gas-microphone PA cell is held in the air tight sample metal chamber and is excited through the optical window. Light absorption by a material leads to two types of de-excitation processes: radiative and nonradiative. The nonradiative de-excitation results in heating the sample. Since the incident radiation is intensity modulated by the chopper, the sample gets heated repetitively. The resulting periodic heat flow from the solid absorber to the surrounding gas creates pressure fluctuations in the cell, which are detected by the microphone. Therefore, the depth of specimen responsible for the PA signal is restricted within a thermal diffusion length defined by³⁰

PAM is a hybrid imaging modality that uses an optical technique for excitation and an acoustic technique for detection. In the conventional approach, the acoustic wave is generated by transient pressure variation caused by the absorption of a nanosecond light pulse. High-resolution deep tissue imaging is possible using acoustic resolution PA imaging.^46–49 PA imaging has been used to image anatomical,⁵⁰ functional,⁴⁶ molecular,⁴⁶ flow dynamic,⁴⁷ and metabolic contrasts in vivo.⁴⁷ Microscopy applications can use either acoustic resolution PAM (AR-PAM) or optical resolution PAM (OR-PAM), or a combination of both, depending on the application and imaging target.⁵¹ The lateral resolution of AR-PAM depends on the acoustic focus of the transducer, which would be on the order of tens of micrometers⁴⁹ with a maximum imaging depth of several millimeters. The lateral resolution of OR-PAM depends on the diffraction limited spot size produced by the lens with a maximum depth imaging of 1.2 mm.^52,53 OR-PAM was used to image small animal models and biological samples with various endogenous or exogenous contrasts.⁵² AR-PAM is a promising imaging technique for both preclinical and clinical applications as deep imaging of biological tissues possible with lower frequency transducers.⁵³

Conventional PA imaging systems require expensive ultrafast lasers with nanosecond pulse widths, an ultrasound transducer, and an aqueous coupling medium for the PA waves to propagate through.^49,51 Clinical adoption has been hindered due to its expense and cumbersome size. Considerable efforts have been made to use low-cost continuous wave (CW) lasers instead of ultrafast lasers for PA imaging.^54,55 Petschke and La Rivière theoretically investigated the possibility of using chirped CW diode lasers,⁵⁵ which would exhibit 20 to 30 dB lower SNR than the typical pulsed laser-based systems, but its compactness and relatively low cost could potential outweigh the lower SNR in selected applications.^54–56

Other noncontact PA systems based on the interferometric principle have been developed.^57–61 However, interferometric techniques were complex, bulky, expensive, and difficult to integrate with the existing optical microscopes.^57,62 A Fabry–Perot sensor is another option, but it is still a contact technique as the Fabry–Perot sensor has to be placed in contact with the sample.^10,63 Researchers have investigated using a fiber sensor, which also required expensive focusing and collection optics. Optical beam deflection techniques are another simple method for PA wave detection. However, this approach requires two lasers for pumping and probing and a large footprint area for the set up because the probe beam needs to travel a relatively long distance (for increasing deflection angle) before detection to improve sensitivity.^64,65 There are also other air couple transducers available, but air-coupled transducers are unfocused, bulky, and exhibit poor SNR, which makes imaging more complicated.⁶⁶

This study is aimed at developing a low-cost air-coupled PA imaging system using common low-power CW lasers and a kilohertz-range microphone, using the PA spectroscopy principle. In the PA spectroscopy, a wide area of the sample is illuminated to get spectroscopy information of aggregates. The typical PA cell used for spectroscopy cannot be used for high-resolution OR-PAM application as the sizes (in the range of a few millimeter to a few inches) and design of the chamber limits the focusing of light.^12,23,35,36 We also developed PA cell for spectroscopy studies. The PA cell was made using a bulk aluminum rod to avoid vibrational noise. The powdered sample was taken in a sample holder made of copper and enclosed in an air tight metallic PA chamber with a thick optical window above the sample. The developed-PA cell was suitable for an epi-illumination excitation, where the sample was excited over the surface by passing the light through an optical window and sample chamber (SC) air column. The sample was excited over a wide area (6 mm) to get the PA spectrum. A thick optical window was used to dampen the vibration noise. A high-magnification objective could not be used in this configuration for a high-resolution PA image due to the short working distance.

In this study, we designed a PA chamber suitable for a low-power CW laser-based high-resolution PAM. The PA cell was designed to fit into existing optical microscopes for multimodal imaging. As the PA spectroscopy technique can be used to study samples in any state such as solid, liquid, and gas,^19,33,34 numerous applications exist, such as coagulation tests to study clotting disorders like thrombophilia and hemophilia, blood pattern analysis (BPA) in forensics,⁶⁷ and monitoring the penetration of drugs into membrane acceptor systems.⁶⁸

2.

Materials and Methods

2.1.

Experimental Setup

A dual modal optical/acoustic microscope system was developed as shown in Fig. 1(a). The optical system is built as an inverted microscope configuration consisting of a halogen lamp, an infinity-corrected $40\times$ and 0.6 NA Olympus objective (O), a mirror (M5), and a Ximea xiQ camera (Ximea). The system consists of motorized linear $X\text{-}Y$ scanning stages (Zaber Technologies) for raster scanning and manual $X\text{-}Y\text{-}Z$ stages (Thorlabs) for manual adjustment. A diode pumped solid state green laser (532 nm, CrystaLaser, Nevada) was used as a source of PA excitation. The laser beam diameter was 0.36 mm and operated at 3 mW power. The beam waist size at the sample surface was $1.2\mu \mathrm{m}$. The beam was intensity modulated for PA generation using a mechanical chopper. The laser path in the system was described by the solid green line in Fig. 1(a). The light beam at the beam splitter is split into two beams, one for the sample and the other for the reference. The reference beam was monitored by a photodiode and used to normalize the PA signal to avoid any change due to laser fluctuations. The reflected light from the beam splitter was focused on the sample using the objective. As the light beam was periodically intensity modulated using a chopper, the periodic pressure wave was created in the PA cell which was measured using a lock-in amplifier. All components were computer controlled through custom MATLAB software (Mathworks). The lock-in amplifier is operated at low-pass filter with a single RC filter ($6\mathrm{dB}/\mathrm{oct}$ roll off) with a time constant of 300 ms. An equivalent bandwidth ($\mathrm{\Delta }f$) of a lock-in is $1/4T$, where $T$ is the time constant ($R\times C$). In this study, $\mathrm{\Delta }f$ is equal to 2.5 Hz. To scan the image of size $0.3\mathrm{mm}\times 0.3\mathrm{mm}$ by $60\times 60$ steps would take up to 20 min. This long scan time is due to the slow serial communications interface used. Future enhancements use a USB interface to increase the scan time by $10\times$.

Fig. 1

(a) A Schematic diagram of a dual model system that contained optical and PA microscopes. (The solid green line represents the beam propagation direction), (b) the schematic diagram of the PA sensor, and (c) photograph of the PA sensor. M1, M2, M3, and M4 are mirrors, M is the microphone, SC is the sample chamber, W is the window, and T is the resonator tunnel.

3.

Results and Discussion

Figures 2(a)–2(c) show the optical image, the PA image, and the corresponding line spread function (LSF), respectively, of a single element in the USAF test target.

Fig. 2

Spatial resolution measurement: (a) optical image of group-6 of the USAF-1951 chart, (b) PA image of element-2 of group-6 (marked by red square in optical image), (c) ESF (black line) and the LSF (blue dots) calculated from PA signal in the rectangular box of the PA image. The solid red line represents the Gaussian fit to the LSF.

The LSF of the system was obtained by numerically differentiating an edge spread function (ESF, solid black line). The ESF represents an average value obtained from the line strip marked by a yellow rectangular box in the PA image 2(b). The lateral resolution limit of the system was calculated from the full width half maximum (FWHM), which was obtained using a Gaussian fit to the LSF. The solid red line in Fig. 2(c) is the Gaussian fit to the LSF. From this measurement, the system is determined to have a lateral resolution of $2.1\mu \mathrm{m}$.

Measurements were performed on a common blood smear on a glass slide to probe single RBC. The system was operated with the settings described in Sec. 2.1. Figure 3 shows the optical and PA images of the RBCs.

Fig. 3

(a) Optical and (b) PA images of RBC (marked in a black rectangular box in the optical image) smeared on a glass substrate.

The MTF of the system was calculated using the USAF chart and an RBC monolayer. The MTF measurement from element 2 of group 6 of USAF chart demonstrated a 76% contrast ratio. Figure 4(a) shows the PA image of the RBCs. Line profiles AB, CD, and EF were drawn across RBCs that exhibit various separations between them. The PA line profiles were obtained [Figs. 4(b)–4(d)] along these lines. The valley (1 to 5) in the line profile represents the separation whose value was obtained from FWHM of the Gaussian fit. The MTF was calculated along those lines by measuring the signal from two neighboring RBCs and the signal from the gap between them. The signal from the RBC was obtained by averaging signals measured across RBCs, whereas the signal from the gap was obtained by averaging signal above the FWHM. Table 1 shows the FWHM of separation of RBC pairs 1 to 5 and their contrast ratios, which were calculated using the expression 2.

Table 1

MTF calculation of system at various RBC aggregates.

Fig. 4

(a) PA image of the RBC monolayer, (b) PA line profile measured along a line AB marked in PA image, (c) PA line profile measured along a line CD, and (d) PA line profile measured along a line EF.

The contrast of the PA image as optical image decreased with an increase in spatial frequency (Table 1). The RBCs at spacing $2.16\mu \mathrm{m}$ exhibited contrast close to PA image of element 2 group 6 of the USAF chart even though its physical properties such as shape, optical, and thermal were completely different. The lowest separation that the system could resolve was $1.37\mu \mathrm{m}$, which was close to the interrogation spot size ($\text{waist size}=1.2\mu \mathrm{m}$) of the laser beam. The separation of the RBCs at position 1 could not be resolved as the spacing was $<1.2\mu \mathrm{m}$. The resolution measurement using the USAF 1951 test target was higher, probably due to the larger thermal diffusivity ($29\times {10}^{-6}{\mathrm{m}}^2/\mathrm{s}$) and effusivity of chrome.

Next, the noise equivalent sensitivity (equivalent to molecular sensitivity) of the system was calculated. The average PA signal obtained from the RBC due to a point excitation by a $40\times$ objective was $9.66\times {10}^{-5}\mathrm{V}$, whereas the standard deviation of the background noise was $9.54\times {10}^{-7}\mathrm{V}$. Therefore, the SNR of this measurement is $\sim 100$. As the signal is generated from $\sim 7\times {10}^6$ hemoglobin molecules (see Sec. 2.7), the NEM is $\sim 693\times {10}^2$.

The NEM per unit bandwidth ($\mathrm{NEM}/\sqrt{\mathrm{\Delta }f}$) is $\sim \mathrm{43,887}{\mathrm{Hz}}^{-1/2}$ or $72.88\times {10}^{-21}\mathrm{mol}$ or 72.88 zeptomol. The system exhibited a sensitivity of nearly half of the value reported by Winkler et al.⁷² The sensitivity was due to an increase in the PA energy due to integration time of the lock-in amplifier. The sensitivity can be improved further by increasing the integration time, but would increase the image acquisition time by reducing the scanning rate.

Next, the MPDL of the sensor was estimated. The average PA signal generated by $7\times {10}^6$ hemoglobin molecules was $9.66\times {10}^{-5}\mathrm{V}$, and, in the same measurement, the standard deviation of background noise was $0.954\mu \mathrm{V}$. The sources for this noise were the amplifier input noise, the Johnson noise of the resistor, and shot noise due to nonuniform flow of current charge carrier. If we assume that all the sources of noise are incoherent and random in nature, then the total noise can be estimated using the square root of the sum of the squares of all the noise. Since the sensitivity of the microphone is $\sim 50\mathrm{mV}/\mathrm{Pa}$, we estimate the MPDL to be $19.1\mu \mathrm{Pa}$.

Amplifier input noise and Johnson noise were Gaussian nature; the amount of noise measured was proportional to the square root of the bandwidth of the lock-in amplifier. Hence, the MPDL per unit bandwidth, $\mathrm{MPDL}/\sqrt{\mathrm{\Delta }f}=7.64\mu \mathrm{Pa}/{\mathrm{Hz}}^{1/2}$.

Other PAM systems require a fluid coupling medium for PA signal detection.^{46–49,51,61,63} This system uses air as a coupling medium and, thus, can be used to image samples that cannot normally be imaged using PAM, such as samples that would be altered or destroyed through contact with fluids, including powders or water-soluble dyes. There are other noncontact techniques based on optical interferometers such as Mach–Zehnder interferometer,^57,58 low-coherence Michelson interferometer,⁵⁹ and Fabry–Perot interferometer.^61,63 These systems are typically complex, expensive, and difficult to integrate with another system. In most cases, the sample surface also required special treatment as the scattering and diffuse surface of the sample would affect the reflection of the probe light (sample arm) of the interferometer system and thereby its SNR. The Fabry–Perot sensor can also be used, but it is a contact technique as the sensor has to be placed in contact with the sample.^57,62 Our sensor is inexpensive, simple, and easy to integrate with existing optical microscopes. The sensor can be used for opaque and highly scattering samples. As the microphone is located far from the laser beam direction, optical imaging techniques such as epi-fluorescence, confocal, Raman, phase contrast, bright, and dark field microscopy can be performed on the sample.

4.

Conclusions

This study demonstrates a compact noncontact PAM system using a low-power CW laser source for PA measurements. Air can be used as the coupling fluid, enabling PA measurements of any sample type. The system exhibits a lateral resolution of $1.37\mu \mathrm{m}$ and a noise equivalent hemoglobin molecular sensitivity of 43,887 molecules ($72.88\times {10}^{-21}\mathrm{mol}$ or 72.88 zeptomol). Future work will focus on improving the sensitivity and the resolution using high numerical aperture objectives and measurements using different CW laser wavelengths, enabling an inexpensive, high-resolution, multispectral PAM system. The compactness of the PA sensor will enable easy integration with existing commercial optical microscopes for multimodal imaging, with numerous biomedical and forensic applications.