Melanomas are among the most commonly occurring cancers and are a clinical challenge to diagnose [1]. Presently, the best method for the clinical evaluation of melanoma is visual inspection by dermoscopy; however, this technique is still far from a reliable diagnosis [2]. The optical diagnosis of melanoma is accessible because the two dominant types of melanin (eumelanin and pheomelanin) have a specific and intrinsic molecular contrast when imaged [3,4]. Several forms of nonfluorescent optical contrast mechanisms can be used to image melanoma without any staining, such as pump–probe [1,3], photoacoustic [5–8], and photothermal [7,8] microscopy. Photothermal microscopy has been used widely in material and biological research [7–11] because of the high sensitivity and lack of staining requirements. Compared with fluorescence microscopy, photothermal imaging does not suffer from photobleaching and blinking, which makes it attractive for robust single molecule detection [12–14]. Besides the commonly used linear photothermal (LPT) mechanism, nonlinear photothermal (NLPT) microscopy has been used to image silica windows and gold pads with enhanced imaging contrast [15], in which the pump heating beam is chopped at $f$ and the temperature variation of the sample is detected at $2f$ with a linear temperature sensor. Nonlinear phenomena have also been observed in photothermal-optical-deflection imaging experiments on samples of both high-purity aluminum and aluminum alloys [16]. The NLPT response of tungsten [17], thin solid films and coatings [18], carbon/epoxy composite materials [19], and cracks in solid [20] have also been investigated. The resolution of nonlinear thermal wave microscopes has been discussed, and it has been suggested that nonlinear microscopy is superior to linear microscopy [21]. However, to the best of our knowledge, all of the research on the effect of NLPT has been performed on bulk materials, thin films, and layered-structure samples; there are no experimental results about the NLPT imaging of biological tissues.

There are two possible explanations for the mechanism of the NLPT process. The first attributes the effects to thermal-wave second harmonic generation (SHG) related to the temperature-dependent heat capacity $C(T)$ and thermal conductivity $k(T)$ of the materials [22,23]. In the second explanation, a diffraction theory of continuous-wave photothermal deflection spectroscopy with fundamental and harmonic responses is presented, in which the displacement of the probe beam centroid leads to a set of analytical solutions to the fundamental and SHG, and the harmonics are caused by the diffraction of the probe beams in the mirage range [24]. In this Letter, we give a brief explanation of NLPT using the first mechanism. The conventional form of the heat conduction equation in a medium with temperature dependent heat capacity $C(T)$ and thermal conductivity $k(T)$ is as follows [25]:

The relationship is also proved in [16] experimentally. Compared with the LPT signal, which can be written as $S\propto I_{\text{pump}}·I_{\text{probe}}$ [26], the cubic dependence of the NLPT signal on the laser intensity makes the spatial resolution of NLPT microscopy higher than that of the LPT one. If we assume that the pump and probe beams have the same wavelength and intensity distribution, this spatial resolution enhancement can be estimated to be $\sim 18\%$ using scalar diffraction theory. Compared with conventional optical microscopy, the resolution enhancement of NLPT microscopy is 42%.

Figure 1 shows the schematic of the experimental setup. Both the pump and probe pulses, with a 10 nm spectral width, are spectrally filtered from a commercial supercontinuum (SC) fiber laser (WL-SC400-4, Fianium, Southampton, UK) using band-pass filters. The pulse duration of a 10 nm spectral width component of the SC is measured as $\sim 100\mathrm{ps}$ and the repetition frequency is tunable in the range of 0.1–40 MHz. The central wavelengths of the pump and probe beams are 488 and 632 nm, respectively. After collinear combination by a short-pass dichroic mirror (DM) with a cutoff wavelength of 560 nm, the two beams are focused into the specimen by a microscope objective with an NA of 0.9 and amplification of $60\times$. The forward propagating beam is collected and collimated by a condenser lens (Olympus) with an NA tunable from 0 to 1.4. Then, the probe beam is spectrally filtered out by a band-pass filter with a central wavelength of 632 nm. To eliminate the effects of the laser noise of the probe beam, an auto-balanced detector (BD) (Nirvana 2007, Newport) is used to accumulate the signals. The intensity of the pump heating beam is modulated at $f$ with an electro-optic modulator (EOM) (LM0202P, Qioptiq Photonics, Germany), creating an LPT signal at $f$ and an NLPT signal at $2f$, as depicted in Fig. 1(b). Two lock-in amplifiers (LIA) (SR844, Stanford Research System, US; Model 7265, Signal Recovery, US) are applied to extract the LPT and NLPT signals from the probe simultaneously, with reference frequencies of $f$ and $2f$, respectively. The sample is mounted on a set of piezo stages (P-622.2CL and P-622.ZCL, Physik Instrumente (PI), Germany), with a resolution of better than 1 nm in all three directions.

 

Fig. 1. (a) Schematic diagram of the experimental setup. EOM, electro-optic modulator; DM, short-wavelength transmitting dichroic mirror with a switching wavelength of 560 nm; BS, beamsplitter; MMF, multi-mode fiber; Filter, bandpass filter with a central wavelength of 632 nm and a bandwidth of 10 nm; BD, auto-balanced detector with a bandwidth of 125 kHz; LIA1,2, lock-in amplifiers. (b) Temporal intensity modulation behavior of the pump and probe pulse trains before and after interacting with the sample. The intensity of the pump pulse is modulated at $f$ with the EOM and the output probe pulse is modulated at both $f$ and $f$ with a modulation depth of $\mathrm{\Delta }{I}_f$ and $\mathrm{\Delta }{I}_{2f}$, respectively, because of the generation of the second-harmonic of the thermal wave. The LPT and NLPT signals are demodulated from the output probe beam with two LIAs referring to $f$ and $2f$, respectively.

Download Full Size | PDF

The sample of mouse skin melanoma is prepared as follows. B16 melanoma cells RCB-1283 (Riken Bio-Resource Center) are cultured in RPMI-1640 supplemented with 10% FBS, 100 units/ml of penicillin, and $100\mathrm{\mu g}/\mathrm{ml}$ of streptomycin (Meiji Seika Kaisha Ltd.) at 37°C in a humidified atmosphere of 95% air and 5% ${\mathrm{CO}}_2$. A 50 μl suspension containing 0.5 million melanoma cells for each head is subcutaneously inoculated into female nude mice (six weeks old) as five aliquots on the dorsal sides from the base of the tail to the neck. Tissues containing the inoculated cells were fixed with 4% paraformaldehyde for three days, embedded in melted paraffin, sliced at 15–20 μm, extended on glass slides, and enclosed with a coverslip.

First, we optimize the whole experimental setup and check the linearity of the detection system. The repetition frequency of the pump and probe pulses is set to its maximum value (40 MHz) for a high signal-to-noise ratio (SNR). The LPT and NLPT signals decrease with the increase of the modulation frequency of EOM, as is shown in Fig. 2(a). The optical noises of the pump and probe pulses increase sharply with the frequency decreases to $<20\mathrm{kHz}$. The modulation frequency of the EOM is set to 30 kHz for the consideration of both the signal and the noise level. The auto-balanced detector is critical for imaging, as it can help to reduce the laser noise of probe pulse by 20–30 dB and obtain images with a higher SNR. The NA of the condenser lens is set within the range of 0.4–0.6 for the maximum LPT and NLPT signals. The time delay between the pump and probe pulses is also adjusted carefully for the highest LPT and NLPT signals (within 1 ns). After optimizing the experimental setup, we check the linearity of the detection system and ensure that the NLPT signal is not due to the saturation and nonlinearity of the detector. As shown in Fig. 2(b), both the LPT and NLPT signals increase linearly with the probe power and no obvious saturation is observed. The LPT signal also shows a linear response to the pump power, shown as a black line in Fig. 2(c). The best fit to the data of NLPT signal is as an equation of the form $y=a{x}^2+bx$, shown as a blue line in Fig. 2(c). The term $bx$ is because of imperfection of the square wave modulator (EOM in the experiment), which may have generated a small spurious NLPT signal. The corrected curve (without the term $bx$) for the response of the NLPT signal to the pump power is shown as a red line in Fig. 2(c). To show the fitting more clearly, Fig. 2(d) is shown with vertical ($x$) and longitudinal ($y$) axes in the logarithmic scale to the base 10 of pump power and NLPT signal, respectively. The slope of the fitting curve increases from 1 to 1.75 with the pump power increased from 0.4 to 1.0 mW, which means that the NLPT signal is mainly the result of the term $bx$ with low pump power, and the effect of the term $a{x}^2$ becomes significant at high pump rate.

 

Fig. 2. (a) Dependence of the LPT and NLPT signals on the modulation frequency of the EOM. Responses of the LPT and NLPT signals to (b) the probe power and (c) the pump power. The pump power is 1 mW in (b) and the probe power is set to 0.5 mW in (c). In (d), the $x$ and $y$ axes are the logarithm to the base 10 of pump power and NLPT signal, respectively. The slopes of the red, green, and blue parts of the fitting curve are 1.00, 1.52, and 1.75, respectively.

Download Full Size | PDF

The LPT and NLPT images of mouse melanoma are obtained simultaneously in the experiment. The pump and probe powers are set to 0.62 and 0.45 mW, respectively, which are in the linear range of the detection system shown in Fig. 2. The pump power is also far below the damage threshold of the sample. To obtain images with a high SNR, the time constant of the LIA and the dwell time of the imaging are set to 10 and 20 ms, respectively. The LPT and NLPT imaging of the mouse melanoma is then performed in a range of $10\mathrm{\mu m}\times 10\mathrm{\mu m}$. The melanin in the melanoma is clearly imaged with $\mathrm{SNR}>10$, as shown in Figs. 3(a) and 3(b). The NLPT signal is about 100 times smaller than the LPT signal, although the ratio is variable with different pump rate and different sample position. As shown in Fig. 3(c), the cross section at the same position [the red line in Figs. 3(a) and 3(b)] indicates that the resolution of LPT and NLPT can achieve 212 and 174 nm, respectively, with a resolution enhancement of 18% for the NLPT microscopy compared with the linear microscopy. The contrast is also higher for the NLPT imaging, as the obscure structures in Fig. 3(a) no longer exist in Fig. 3(b). The contrast enhancement can also be seen in Fig. 3(c), in which the maximum to minimum value of the NLPT curve is larger than that of the LPT curve. The maximum SNR of Fig. 3(b) is $>150$, indicating that a high-quality image is still possible, even if the dwell time is decreased 10 times [27].

 

Fig. 3. (a) LPT and (b) NLPT imaging of mouse melanoma with the range of $10\mathrm{\mu m}\times 10\mathrm{\mu m}$ and whole pixels of $300\times 300$. The powers of the pump and probe pulses are 0.62 and 0.45 mW, respectively. The line profiles in (c) indicated by the red line in (a) and (b) show that the spatial resolution for the LPT and NLPT imaging are 212 and 174 nm, respectively. The NLPT signal is about 100 times smaller than the LPT signal.

Download Full Size | PDF

The high harmonic demodulation technology is also applied in saturated excitation (SAX) microscopy to obtain sub-diffraction limited resolution [28,29], although the saturated fluorescence in SAX microscopy may induce photobleaching of the fluorescent dye, which will not happen in the label-free NLPT microscopy. The sensitivity of photothermal microscopy is also higher compared with fluorescence microscopy, as discussed in the introduction.

In summary, we demonstrated the label-free imaging of mouse melanoma with both LPT and NLPT microscopy simultaneously. The spatial resolution and contrast of NLPT microscopy are higher than those of LPT microscopy. The pump and probe pulses are filtered from a supercontinuum fiber laser, which makes the experimental setup more compact and cheaper than the Ti:sapphire laser and OPA system. The technique opens further possibilities for the clinical evaluation of stainless melanoma and other bio-tissues with a high resolution and contrast.