Multi-wavelength photo-acoustic microscopy in the frequency domain for simultaneous excitation and detection of dyes

: An optical-resolution photoacoustic microscope with modulated CW laser diodes allowing multi-channel imaging is presented that can be used for both imaging biological tissues and for targeted photo-dynamic therapy (PDT) varying the optical power and exposure time. The effects of this therapy are immediately monitored in order to optimize the time of irradiation. After the description of the experimental setup, in vitro and in vivo applications are presented on a synthetic sample and on the mouse ear using hemoglobin as endogenous and methylene blue as exogenous dye for imaging and PDT, respectively.


Introduction
In clinic, there is an important need for handheld imaging equipment for the detection of microscopic effects of anti-cancer therapies. These have multiple effects on tissue parameters, concerning tumor (micro)-vessels, different cell types, and inflammation processes. These effects are either anti-or pro-tumoral and often stimulate recurrence in time [1]. Until now, there are no clinical data available on these processes, because low cost imaging techniques for frequently monitoring in a personalized treatment planning and follow-up are absent. Most conventional medical imaging techniques are expensive, unable to resolve tissues at the microscopic level.
Recent clinical studies have shown that microscopic details of the tumor vessels in human breast cancer can be screened by photo-acoustic (PA) imaging to a depth of a few centimeters [2]. The technique may use endogenous dyes e.g. hemoglobin (Hb) or exogenous clinical dyes. PA-imaging has the potential to be low cost, and handheld for in vivo imaging with either a microscopic optical resolution or with a mesoscopic acoustic resolution.
In this preclinical proof of principle study, we focus on optical-resolution photoacoustic microscopy (OR-PAM) using modulated CW laser diodes. This allows simultaneous multichannel imaging with a simple device without a high bandwidth or temporal multiplexing in comparison to time domain (TD) systems [3,4]. In comparison to TD excitation, the generation of acoustic pressure waves is usually less effective for frequency domain (FD) excitation. However, narrowband signal detection may be very effective, especially by the use of a matched filter or a resonant detector. A recent study has shown that the signal-to-noise ratio in FD photo-acoustics can be equal to or even higher than that of TD photo-acoustics under certain conditions [5]. Furthermore, CW laser diodes are more cost effective, come with smaller footprints and require less maintenance compared to the Q-switched ns-lasers typically used in TD systems. After a short description and validation of our setup, we will demonstrate that simultaneous excitation and detection of multiple dyes is possible in monitoring acute effects of a photo-dynamic therapy.
In photo-dynamic therapy (PDT), photosensitizing agents are used for selectively killing cancer cells or tumor blood vessels after local exposure to light photons [6]. This therapy is tested in clinical trials for the treatment of residual tumor cells after resection [7]. Recently, the clinical dye methylene blue (MB) has been successfully used for PDT in preclinical studies [8][9][10]. Here, we demonstrated that photo-activation of MB in micro-vessels and simultaneous monitoring of the red blood cell distribution is possible on the PA-microscope. Different groups have proven the feasibility of PA microscopy and imaging in vivo, but only few [11] focused on low cost equipment with the use of different laser diodes for the simultaneous excitation of several endogenous and exogenous PA dyes.
The aim is to proof the principle that photo-dynamic therapy and in vivo monitoring of the effects on the microvasculature can be performed simultaneously when the intensities of CW laser diodes are frequency encoded on a photo-acoustic microscope.

Experimental setup
The experimental setup is described in Fig. 1. The collimated beams of two CW laser diodes (Omicron PhoxX laser diodes @ λ 1 = 415 nm and λ 2 = 660 nm) are combined with a dichroic mirror (Omicron LightHub). The laser diodes intensities are sinus modulated at f 1 = 5 MHz and f 2 = 5.3 MHz respectively, which ensures good channel isolation (see detection bandwidth below). The modulation signals are provided by a dual channel function generator (Tektronics, AFG3102C). Both beams have the same waist ω 0 = 350 µm and the same quality factor M 2 = 1.06 at the output of the laser diodes. The width of the beams is then increased by a factor of 10 by means of an afocal system in order to match the diameter of the entrance pupil of the microscope objective (Zeiss LD Epiplan 20X/0.4). This is the best compromise between resolution and loss of power. The resolution is of the order of 1 μm at 415 nm and 1.6 μm at 660 nm. The mean optical powers on the target are set between 2 mW and 6 mW for the two wavelength. A two-axis galvanometric mirror scanner makes it possible to move the beams on the surface of the target to build an image, with a pixel dwell time of 150 µs. The position of the target is set by a motorized three-axis translation stage. The PA signal is collected through an ultrasound transmission chamber by a focused piezoelectric transducer (NDT Systems, IBMF054) in a confocal arrangement. The acoustic probe has a focal distance of 1", which is large enough that heating of the transducer by direct laser irradiation can be neglected. The transmission chamber is filled with a mixture of 50% ultrasonic gel (modul diagram, UG0260A2) and deionized water to achieve a compromise between ease of handling and acoustic impedance matching with the transducer. The electric signal from the acoustic probe is finally amplified (R&K, LA110-0S) and demodulated by two lock-in amplifiers (Stanford Research Systems, SR844), respectively running at f 1 and f 2 in a bandwidth of 3.2 kHz, to recover both the amplitudes of the PA signals generated at λ 1 and λ 2 .

Animal model
PA experiments were performed on the ears of white CD-1 IGS mice (Charles River, Écully, France). All efforts were made to minimize their number. They were housed in ventilated cages with food and water ad libitum in a 12 h light/dark cycle at 22 ± 1°C. For the experience, mice were anesthetized by an intraperitonial injection of ketamine (0.1 mg/g, Imalgene 1000, Boehringer Ingelheim) / xylazine (0.01 mg/g, ROMPUN Bayer) mixture. Their body temperature was maintained at 36 -37° C using an electric heating pad with a feedback system. In experiments with dye infusion, mice were cannulated in a caudal vein (BD NeoflonTM, 16GA 0.6 x 19mm, Becton Dickinson, Helsingborg Sweden) and connected to a 1 ml syringe on an infusion pump with a target volume of 0.2 ml/min. In photo-dynamic therapy experiments, the concentration of the infused methylene blue solution was 0.2 mg/ml (M9140-100g, SIGMA Alderich France). Before positioning the ear on the lid of a petri dish (Ø = 10mm) in the focal plane of the objective with ultrasound gel, ears were cleaned with a hear-removing cream (VEET, France) and rinsed with 70% ethanol.
In accordance with the policy of Clinatec (permit number: B38-185 10 003) and the

Image processing and analysis
All PA images were processed and analyzed using Image-J software: version ImageJ 1.51 u, Wayne Rasband, NIH, USA, http://imagej.nih.gov/ij, Java 1.8.0_66 (64 bit). Composite images are used to highlight changes in the hematocrit distribution in the whole vessel network. The LUT of the sum of reference Hb PA images before a change is set to red and the LUT of the sum of Hb PA images after a change is set to green. The number of images for the calculation of the sum before and after a change is equal. In this case, a composite image indicates an increase of hematocrit in green and a decrease is coded red. Vessels are coded yellow ( = red + green) when no change in hematocrit occurs and become orange for transient modulations of the Hb PA signal. This is perfectly illustrated in Fig. 6(d) and 6(e). Most PA images in Figs. 4, 6 and 7 had a background correction using a rolling ball algorithm with a radius of 20 pixels.
All graphs in the paper, statistics and linear regression analysis were generated by GraphPad PRISM version 3.02, 2000 (Graphpad Software, Inc). In Figs. 4, 5 & 6, signal to noise ratios of Hb PA signals in vessels were obtained after dividing the PA signal amplitudes by the mean signal amplitude of the background. For all regions of interest (ROI), signal to noise ratio analyses in time were performed with similar areas.

Results
The first section describes an in vitro validation of the multi-wavelength PA microscope followed by a proof of principle in vivo PDT study on the mouse ear.

In vitro validation of the PA microscope
Nowadays, there are no regular test samples to measure and validate signal to noise ratios of the PA signals as a function of the dye concentration for different excitation wavelengths. Therefore, we present here a simple and reproducible test sample in Fig. 2(a) and 2(d), which contains the primary colors of a laser printer: yellow, cyan and magenta. The surface concentration of the dyes (number of printed dots per unit area) was varied using different transparencies of a colored pattern, printed on a transparency for overhead projector (see Fig.  2(a)). The normalized concentration is determined by taking the highest concentration as the reference. A normalized concentration of 100% corresponds to a transparency of 0% and vice-versa. The PA signal amplitudes were linearly related to the dye concentration as depicted in Fig. 2(c). Similar results were obtained for cyan (λ 2 = 660 nm), but not shown.  Hb  . 3(b)). t can be seen is less than the rea of Fig. 2 Fig. 3(a) In summary, real simultaneous excitation and detection of different PA dyes is possible and the surface average PA-signal amplitudes are linearly related to the surface concentration of the dyes at constant laser irradiance and dye temperature.

Photo-acoustic microscopy and photo-dynamic therapy
The current PA microscope can be used simultaneously for PA imaging and PDT. In our setup, the same laser diodes are used for both applications. In the imaging mode, an overview image may indicate the ROIs for PDT using single point excitations. Before combining PA microscopic imaging and PDT, it is important to analyze possible tissue laser damage during imaging at the absence of photo sensitive dyes. In the imaging mode, the maximum 415nm laser diode power is set to 2.3 mW at the entrance of the skin and the dwell time is 150 µs. The average speed of red blood cells (RBC) in the capillaries is 130 µm/s [12]. During the pixel dwell time, the RBCs have moved 0.02 µm, which is small compared with their size (6-8 µm). One can then assume that the RBCs are stationary during a pixel acquisition. Therefore, the average of the PA amplitude in a capillary is proportional to the hematocrit. On  Fig. 4(c), yellow coded vessels indicate no change in RBC distribution. We can see that there is no tissue damage within 10 minutes in the imaging mode, although we are a priori above the American National Standards Institute (ANSI) laser safety limit (0.8 kW/cm 2 for an exposition time of 150 µs) if we consider the irradiance at the optical focus in air (150 kW/cm 2 ). Actually, this safety limit needs clarification for the present problem. Indeed, the laser beam is focused under the surface of the skin. First, the radius of the exposed area on the surface of the skin is much larger than the beam waist because of convergence. On this basis, our estimate shows that the irradiance at the surface is below the safety limit. Then, some of the light is reflected by the surface of the skin (up to 30%, depending on the skin color). The rest undergoes scattering through the tissue before reaching the focal plane, where only a fraction of the initial light is concentrated. Finally, the local heat in the functional vessels is rapidly dispatched in comparison to the therapy mode with higher incident power and exposure time. In this paragraph, we analyzed the effect of a 1 minute single point laser irradiation for the 415 nm laser diode at 2.3, 3.6 and 6.1 mW (therapy mode). The effect on blood clotting was

Simultaneous PA microscopy and PDT of methylene blue for the monitoring of the acute effects on the RBC distribution in microvessels
The final in vivo application is a simultaneous photo-activation of MB and monitoring the effects on the RBC distribution. The dye MB is used as agent in preclinical and clinical photo-dynamic therapy studies [8,9]. Here, MB was intravenously infused, photo-activated with an optical power of 6.1 mW at 660 nm in the whole image during scanning and detected inside the mouse ear (micro)vessels. In Fig. 6, the MB dye caused blood coagulation in mainly arteries and arterioles immediately after infusion, see high intensity spots in 6b and cyan signals in 6f. Indeed, image 6c, which is the sum of slices after photo-activation, show an important decrease of the RBC concentration with regions of less Hb PA signals, see white arrows. In Fig. 6(d), the composite images of the sum slices before activation (red signals) and the sum of slices after photo-activation (green signals) clearly indicate vessels with a strongly reduced RBC concentration (red vessels), increase of RBC concentration (green vessels) or vessel areas with no change in the RBC concentration (yellow areas). The Hb PA signal to noise ratios changes are shown for the different ROIs in Fig. 6(e) and confirm the color codes of image 6d. Only the curve of the yellow ROI-2 in Fig. 6(d) is made orange in Fig. 6(e) for a better visibility. The overall normalized Hb and MB PA signals of all vessels after multiplying the whole times series with a binary mask (vessel = 1, background = 0) is shown in Fig. 6(h). The overall RBC flux reduction was immediate when the MB infusion started (blue arrow). Indeed, an orthogonal view of X in time (see Fig. 6(g) and horizontal yellow line in Fig. 6(f)) show at the time of MB infusion (cyan signal) that the MB PA signals are hardly mixed with red Hb PA signals, which should have become violet signals otherwise. In a control experiment (see Fig. 7), the laser diode at 660 nm was set to 4.2 mW at the presence of MB, which induced not the effects as described above. Thus, photo-activation of MB caused blood coagulation and not the infusion of MB only.
The simultaneous detection of the endogenous dye Hb and photo-activation of MB is important in photo-dynamic therapy for on-line activation and monitoring of PDT effects. This permits to stop PDT as soon as maximum effects are reached in order to spare normal adjacent tissues.

Conclusio
We designed modulating th be imaged sim mW), it is po can then be u monitoring of temporal colo hematocrit af configuration mode can be device [3]. F implemented photoacoustic 7

Disclosures
The authors declare that there are no conflicts of interest related to this article.